Skip to content

2026 05 06 aibom platform observability control comparison

Jump to bottom
github-actions[bot] edited this page May 6, 2026 · 1 revision

What introspection, export, and control surfaces actually exist across production agentic Artificial Intelligence (AI) platforms: a comparative analysis of Amazon Web Services (AWS) Bedrock Agents, Microsoft 365 Copilot, Salesforce Agentforce, and ServiceNow Now Assist?

Research Question

What logs, traces, audit Application Programming Interfaces (APIs), Artificial Intelligence Bill of Materials (AIBOM) export capabilities, version-pinning mechanisms, allowlists, and policy hooks actually exist in production agentic AI platforms, specifically Amazon Web Services (AWS) Bedrock Agents, Microsoft 365 Copilot, Salesforce Agentforce, and ServiceNow Now Assist, and where does each platform remain opaque even with full observability enabled?

Scope

In scope:

  • Inventory of observability surfaces per platform, including trace APIs, log formats, audit logs, and agent or configuration export capabilities.
  • Control surfaces, including version pinning or version snapshots, tool allowlists or denylists, policy hooks, guardrails, and content filters.
  • AIBOM export, specifically whether the platform provides any native mechanism to export an AIBOM or equivalent structured inventory of agent dependencies.
  • Opaque zones, specifically which aspects of agent execution remain hidden even with logging enabled, such as vendor-side prompt augmentation, internal routing, and model-side reasoning internals.
  • Managed versus open comparison, specifically Microsoft 365 Copilot, Salesforce Agentforce, and ServiceNow Now Assist as managed platforms versus AWS Bedrock Agents as a more API-addressable orchestration surface.
  • Feasibility of automated AIBOM generation in Continuous Integration and Continuous Deployment (CI/CD) pipelines for each platform.

Out of scope:

  • Deep implementation of AIBOM construction, which is covered elsewhere in the corpus.
  • Theoretical AIBOM schema design.
  • Platforms outside the four named products.

Constraints:

  • Use publicly available documentation only.
  • Apply one comparison rubric across all four platforms.
  • Prefer 2024 or later sources where the capability is changing quickly.

Context

[fact; source: https://davidamitchell.github.io/Research/research/2026-04-26-vendor-platform-governance-constraints-compensating-controls.html; https://davidamitchell.github.io/Research/research/2026-04-26-ai-lowcode-observability-telemetry-governance.html] Prior completed work in this repository already found that vendor platforms expose governance unevenly and that audit evidence is usually split across traces, payload logs, lifecycle audit records, and policy systems rather than presented as one complete inventory.

[inference; source: https://davidamitchell.github.io/Research/research/2026-05-02-ms-copilot-vs-aws-bedrock-enterprise-ai-capability-model.html; https://davidamitchell.github.io/Research/research/2026-05-02-vendor-lock-in-portability-multi-platform-ai.html] This item narrows that broader platform-governance question to the specific evidence surfaces an enterprise would need to build or automate an AIBOM, because portability, assurance, and incident reconstruction all depend on what the platform exposes rather than what an AIBOM schema could theoretically represent.

Approach

  1. Evaluation rubric design: compare each platform on trace completeness, audit coverage, configuration export, version control, tool and policy controls, native inventory or export, and opaque zones.
  2. AWS Bedrock Agents assessment: review trace events, invocation logging, guardrails, agent configuration APIs, and alias or version management.
  3. Microsoft 365 Copilot assessment: review Microsoft Purview audit coverage, Microsoft Graph reporting, Microsoft 365 agent registry and tool approval, and Microsoft Copilot Studio governance and export surfaces.
  4. Salesforce Agentforce assessment: review Agentforce Observability, trusted-services security controls, session-tracing export, and public evidence for agent versioning and metadata portability.
  5. ServiceNow Now Assist assessment: review AI Control Tower and Now Assist governance material for inventory, oversight, monitoring, and export or lifecycle controls.
  6. Cross-platform synthesis: determine which platform provides the strongest AIBOM substrate, which remains most opaque, and which compensating controls are required.

Sources

Research Skill Output

(Full output from running the research skill, retained verbatim in the completed item. Sections 0 to 5 are the investigation, and section 6 seeds the Findings section below.)

§0 Initialise

§1 Question Decomposition

  • Root question: Which platform exposes the most decision-useful substrate for an AIBOM, and where does managed-platform opacity prevent complete inventory reconstruction?
    • A. Trace and log completeness
      • A1. What runtime trace does each platform expose?
      • A2. Does the runtime trace include prompts, tool calls, retrievals, rationale, and version identifiers?
    • B. Audit and analytics
      • B1. What tenant or admin audit surfaces exist?
      • B2. Are those surfaces operational traces or only usage and lifecycle records?
    • C. Export and configuration inventory
      • C1. Can the agent configuration or manifest be exported natively?
      • C2. Does the platform expose enough configuration fields to automate a declarative inventory?
    • D. Version control and policy hooks
      • D1. What immutable version, alias, or packaging controls exist?
      • D2. What tool allowlists, policy hooks, guardrails, or approval flows exist?
    • E. Opaque zones
      • E1. Which runtime behaviors remain hidden from the customer?
      • E2. Does "full observability" still omit model, routing, augmentation, or transcript details?
    • F. AIBOM feasibility
      • F1. Is there a native AIBOM or Bill of Materials export?
      • F2. If not, how feasible is automated AIBOM construction from the documented surfaces?

§2 Investigation

Comparison rubric

  • [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185] The most decision-useful comparison rubric is: runtime trace fidelity, tenant audit coverage, configuration exportability, version and change control, tool and policy control, native inventory or export, opaque zones, and AIBOM automation feasibility.

AWS Bedrock Agents

  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html] AWS Bedrock Agents expose step-level runtime traces that include agentId, agentAliasId, agentVersion, sessionId, callerChain, preprocessing, orchestration, postprocessing, guardrail, failure, routing, and custom-orchestration events.
  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html] Bedrock trace payloads include prompt text, inference configuration, prompt-override state, parser state, rationale, action or knowledge-base invocation inputs, observations, raw model output, parsed output, and token-usage metadata.
  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html] Model invocation logging can export full request data, full response data, and metadata to Amazon CloudWatch Logs or Amazon Simple Storage Service (Amazon S3), but it is disabled by default.
  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html] Model invocation logging only captures calls made through the bedrock-runtime endpoint and does not capture traffic sent through other endpoints such as bedrock-mantle.
  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html] The GetAgent API exposes a machine-readable configuration surface that includes foundation model, instruction, orchestration type, prompt overrides, inference configuration, memory configuration, and guardrail identifier and version.
  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/agents-version-view.html] Bedrock versions are immutable snapshots created through aliasing, and aliases can be redirected or paused without changing application integration code.
  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails.html] Bedrock Guardrails expose configurable content filters, denied topics, word filters, sensitive-information filters, contextual grounding checks, and automated-reasoning checks, and the guardrail can be invoked by identifier and version.
  • [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html] Bedrock provides the strongest documented substrate for automated AIBOM generation because both declarative state and runtime evidence are exposed through APIs and exportable logs rather than only through tenant analytics.
  • [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html] Bedrock still has opaque zones because provider-side model internals, underlying foundation-model weights, and any traffic outside the documented logging path remain outside customer-visible evidence even when tracing is enabled.

Microsoft 365 Copilot and Microsoft Copilot Studio

  • [fact; source: https://learn.microsoft.com/en-us/purview/audit-copilot] Microsoft Purview audit records for Copilot include accessed resources, agent identifiers, agent names, agent versions, plugin identifiers and versions, host application, application identity, message identifiers, and policy details on blocked or restricted resources.
  • [fact; source: https://learn.microsoft.com/en-us/purview/audit-copilot] For Microsoft 365 Copilot scenarios, the Purview ModelTransparencyDetails field includes model provider name, but model name and model version are not available.
  • [fact; source: https://learn.microsoft.com/en-us/graph/api/reportroot-getmicrosoft365copilotusageuserdetail?view=graph-rest-beta] Microsoft Graph exposes tenant usage export for Microsoft 365 Copilot, including per-user last-activity dates across Copilot surfaces and comma-separated values export via a preauthenticated download URL.
  • [fact; source: https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio] Copilot Studio logs both agent authoring and usage events into Microsoft Purview, including agent creation, publishing, sharing, App Insights logging configuration updates, plugin operations, environment-variable changes, and CopilotInteraction usage records.
  • [fact; source: https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio] The Audit solution in Purview does not include the full text or transcript of Copilot Studio interactions, only transcript thread identifiers, while transcript text is handled separately through Data Security Posture Management for AI.
  • [fact; source: https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-data-loss-prevention; https://learn.microsoft.com/en-us/microsoft-copilot-studio/security-and-governance] Copilot Studio provides strong policy hooks through Power Platform data-loss-prevention controls, including blocking knowledge sources, Hypertext Transfer Protocol (HTTP) requests, connectors used as tools, skills, channels, and event triggers, with real-time enforcement.
  • [fact; source: https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-solutions-import-export; https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-solutions-overview] Copilot Studio supports solution-based export and import of agents across environments, but not all components and properties transfer, and knowledge-source or connector dependencies can require reconfiguration after import.
  • [fact; source: https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide] The Microsoft 365 agent registry provides tenant-wide inventory, comma-separated values export, and management fields that include status, channel, platform, version, owner, description, and instructions, and it also supports packaged upload of custom agents.
  • [fact; source: https://learn.microsoft.com/en-us/microsoft-365/admin/manage/manage-tools-for-agent?view=o365-worldwide; https://learn.microsoft.com/en-us/microsoft-agent-365/bring-your-own-mcp] Microsoft 365 provides a tool-governance surface for MCP servers, including admin review, approval, blocking, publisher visibility, and permission consent before tools become available to agent-building surfaces.
  • [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/graph/api/reportroot-getmicrosoft365copilotusageuserdetail?view=graph-rest-beta; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio; https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-solutions-import-export] Microsoft exposes strong governance surfaces for tenant inventory, tool approval, lifecycle audit, and Copilot Studio packaging, but the core Microsoft 365 Copilot runtime remains only partially inspectable because the standard audit stream does not expose full prompt or response content or model-version details.
  • [assumption; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide] Search note: query site:learn.microsoft.com "Microsoft 365 Copilot" "model version" returned no reviewed Microsoft Learn page describing customer-configurable model-version pinning, so this item treats that control as not publicly documented here.
  • [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-solutions-import-export] Automated AIBOM generation is feasible for custom Copilot Studio agents and tenant agent inventory, but only as a partial inventory for the managed Microsoft 365 Copilot runtime because significant orchestration and model details remain hidden.

Salesforce Agentforce

  • [fact; source: https://www.salesforce.com/agentforce/observability/] Salesforce Agentforce Observability presents a centralized mission-control surface that includes deep observability, session-level tracing, health monitoring, adoption analytics, and consumption tracking.
  • [fact; source: https://www.salesforce.com/agentforce/observability/] Salesforce states that administrators can inspect every Agentforce session, review turns in the conversation, analyze subagents, segment sessions by intent and sentiment, and configure near-real-time alerts on key metrics.
  • [fact; source: https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/] Salesforce documents near-real-time Shield Event Monitoring for Agentforce, including tracking data access, auditing Agent Apex and API activity, applying transaction-security policies, and surfacing agent metrics through Security Center.
  • [fact; source: https://www.salesforce.com/blog/best-practices-for-secure-agentforce-implementation/] Salesforce documents role scoping, data minimization, private versus public actions, runtime guardrails, supervisory Large Language Models (LLMs), and the Einstein Trust Layer as core control surfaces for Agentforce.
  • [inference; source: https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://www.salesforce.com/agentforce/observability/] Public Salesforce documentation reviewed in this session indicates that Agentforce can export session tracing in OpenTelemetry format, which makes it the closest managed-platform analogue to a runtime AIBOM feed among the three more managed platforms in scope.
  • [inference; source: https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/] Public Salesforce material also indicates agent-version management and observability access exist, but some of the more detailed versioning and export documentation was not directly retrievable in this runtime, so those claims should be treated as medium-confidence rather than high-confidence evidence.
  • [assumption; source: https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://www.salesforce.com/agentforce/observability/] Search note: query site:salesforce.com Agentforce DX metadata Salesforce official found official Agentforce DX pages, but the developer pages were not directly retrievable in this runtime, so this item uses them as medium-confidence support only.
  • [inference; source: https://www.salesforce.com/agentforce/observability/; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/; https://www.salesforce.com/blog/best-practices-for-secure-agentforce-implementation/] Agentforce appears to provide richer native runtime observability than Microsoft 365 Copilot, but it still leaves opaque zones around the underlying model internals, full declarative export semantics, and the precise completeness of portable configuration outside the documented observability and governance flows.

ServiceNow Now Assist and AI Control Tower

  • [fact; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185] ServiceNow describes AI Control Tower as a centralized governance layer that tracks model inventory, lifecycle status, data touchpoints, drift, compliance mapping, and workflow-triggered remediation.
  • [fact; source: https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] ServiceNow describes AI Control Tower as a single authoritative source for AI initiatives that inventories systems, datasets, prompts, evaluations, applications, risks, and value outcomes across the lifecycle.
  • [fact; source: https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753] ServiceNow also describes dashboards for monitoring agents, workflows, key performance indicators, governance automation, and audit trails for stakeholders.
  • [assumption; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] Access note: the reviewed public ServiceNow docs URLs were not directly readable through the fetcher in this session, so the assessment relies on accessible official ServiceNow community material on the same domain.
  • [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753] ServiceNow's public evidence is strongest on centralized governance and inventory but weaker on concrete customer-visible trace-export schemas, per-agent version pinning, or portable configuration manifests than the AWS and Microsoft materials reviewed here.
  • [assumption; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753] Search note: query site:servicenow.com "AI Control Tower" export configuration API found no accessible official technical page in this session that exposed a concrete AI Control Tower export or trace API.

Cross-platform synthesis

  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html] Bedrock exposes the deepest documented combination of configuration API, immutable runtime versioning, and step-level execution traces.
  • [fact; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-data-loss-prevention; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide] Microsoft exposes the broadest tenant-governance and admin-inventory surface among the managed platforms, including agent registry export, tool approval, lifecycle audit, and strong policy controls.
  • [fact; source: https://www.salesforce.com/agentforce/observability/; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/] Salesforce exposes a stronger native runtime-observability story than Microsoft 365 Copilot through session tracing and observability tooling, but the publicly accessible documentation reviewed here is less transparent about portable configuration detail than AWS.
  • [fact; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] ServiceNow emphasizes governance, inventory, and executive oversight, but the public evidence reviewed here is comparatively weak on detailed operational trace export.
  • [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185] None of the four platforms provides a native AIBOM export, so every platform still requires an enterprise-side normalization layer if the goal is a portable, verifiable inventory artifact.

§3 Reasoning

  • [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html] Bedrock's high ranking is grounded in directly documented runtime trace fields plus directly documented configuration API fields, not only in marketing language about observability.
  • [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide] Microsoft's ranking depends on the distinction between administrative governance and runtime introspection, because Microsoft documents strong inventory and approval surfaces while also documenting missing model-name and model-version detail in core Copilot audit events.
  • [inference; source: https://www.salesforce.com/agentforce/observability/; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/] Salesforce's ranking depends on accessible public evidence for runtime observability and security controls, not on inaccessible help-center content alone.
  • [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] ServiceNow's lower confidence is driven by the level of detail in the accessible evidence, not by proof that the platform lacks deeper internals.
  • [inference; source: https://davidamitchell.github.io/Research/research/2026-04-26-vendor-platform-governance-constraints-compensating-controls.html; https://davidamitchell.github.io/Research/research/2026-04-26-ai-lowcode-observability-telemetry-governance.html] The correct conclusion is therefore comparative rather than absolute: the platforms differ mainly in how much evidence they externalize for customer-controlled assurance, not in whether governance matters.

§4 Consistency Check

  • [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html] No contradiction remains in the Bedrock assessment, because strong trace coverage coexists with logging gaps outside bedrock-runtime and does not imply full visibility into provider internals.
  • [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio] No contradiction remains in the Microsoft assessment, because Microsoft simultaneously offers strong tenant audit and governance while still omitting full transcript content and model-version detail from the main audit stream.
  • [inference; source: https://www.salesforce.com/agentforce/observability/; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html] The Salesforce assessment is internally consistent if session-trace export and observability are treated as documented but medium-confidence because the public material leaves export-schema detail under-specified.
  • [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] The ServiceNow assessment is internally consistent if it is framed as "public evidence is high-level" rather than "the platform definitively lacks detailed controls."

§5 Depth and Breadth Expansion

  • [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] From a technical lens, Bedrock is the easiest platform for machine-generated evidence collection, Microsoft is strongest for tenant governance over agents and tools, Salesforce is strongest for managed-platform runtime session visibility, and ServiceNow is strongest for centralized governance framing.
  • [inference; source: https://davidamitchell.github.io/Research/research/2026-05-02-vendor-lock-in-portability-multi-platform-ai.html; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://learn.microsoft.com/en-us/purview/audit-copilot] From a portability lens, externalized control planes and normalized telemetry remain necessary because every reviewed vendor still leaves some of the actual execution substrate inside vendor-owned boundaries.
  • [inference; source: https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/; https://learn.microsoft.com/en-us/microsoft-agent-365/bring-your-own-mcp; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185] From a governance lens, tool approval, transaction-security hooks, and workflow-triggered remediation are at least as important as raw trace export, because AIBOM usefulness depends on policy context as well as on event capture.
  • [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html] From a behavioral lens, managed platforms are designed to reduce customer responsibility for orchestration internals, which also reduces customer evidence depth; the best a customer can usually get is a curated accountability surface rather than total introspection.

§6 Synthesis

Executive summary:

AWS Bedrock Agents provide the strongest documented substrate for automated Artificial Intelligence Bill of Materials (AIBOM) generation among the four reviewed platforms because AWS exposes both a machine-readable configuration API and detailed step-level runtime traces with immutable version snapshots. [inference; source: https://owaspaibom.org/; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html] Microsoft 365 Copilot provides a strong tenant-governance surface through audit, inventory, and tool-approval controls, but its main audit stream omits critical runtime detail such as model name and model version for Microsoft 365 Copilot and does not provide full transcript content in the standard audit record. [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide] Salesforce Agentforce documents extensive managed-platform runtime observability through session-level tracing and observability tooling, but the reviewed public evidence leaves export semantics and versioning detail under-specified, so confidence is lower than for AWS. [inference; source: https://www.salesforce.com/agentforce/observability/; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/] ServiceNow Now Assist and AI Control Tower emphasize centralized governance, inventory, and oversight, but the accessible public evidence reviewed here is more strategic than technical, which makes detailed AIBOM automation claims the weakest of the four. [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] The reviewed public documentation for none of the four platforms describes a native AIBOM or equivalent Bill of Materials export, so an enterprise that wants portable inventory assurance still needs an external normalization and evidence-binding layer. [inference; source: https://owaspaibom.org/; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-runtime-generation-divergence-theory.html; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-identity-delegation-trust-theory.html]

Key findings:

  1. AWS Bedrock Agents are the strongest documented Artificial Intelligence Bill of Materials (AIBOM) substrate in this comparison because AWS exposes agent configuration through the GetAgent API, immutable alias-based versions, and step-level runtime traces that include prompts, rationale, action invocations, observations, and version identifiers. ([inference]; high confidence; source: https://owaspaibom.org/; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html)
  2. AWS Bedrock still appears not to provide a native AIBOM export, and its observability remains incomplete when traffic bypasses the documented bedrock-runtime logging path or when the enterprise needs provider-side model internals rather than customer-visible orchestration evidence. ([inference]; high confidence; source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html)
  3. Microsoft 365 Copilot exposes strong tenant-governance evidence through Purview audit, the Microsoft 365 agent registry, and tool-approval workflows, but the standard audit schema explicitly omits model name and model version for Microsoft 365 Copilot scenarios and does not function as a full runtime trace. ([fact]; high confidence; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/manage-tools-for-agent?view=o365-worldwide)
  4. Microsoft Copilot Studio adds real-time data-loss-prevention controls, agent export and import through solutions, and detailed authoring and usage audit events, which strengthens Microsoft’s AIBOM readiness even though its export surface remains partial because some components and properties do not transfer cleanly. ([inference]; medium confidence; source: https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-data-loss-prevention; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio; https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-solutions-import-export)
  5. Salesforce Agentforce publicly documents mission-control observability, session-level tracing, and near-real-time security telemetry, including Event Monitoring and transaction-security controls around agent activity, which makes the reviewed public material look richer on runtime observability than the standard Microsoft 365 Copilot audit surface even though some of that gap may reflect how the vendors publish their capabilities. ([inference]; medium confidence; source: https://www.salesforce.com/agentforce/observability/; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/; https://www.salesforce.com/blog/best-practices-for-secure-agentforce-implementation/; https://learn.microsoft.com/en-us/purview/audit-copilot)
  6. Salesforce Agentforce also appears to support session-trace export and agent version management using OpenTelemetry, the vendor-neutral observability framework, which would make automated AIBOM generation more feasible than on Microsoft 365 Copilot, but the public evidence leaves the exact export schema and version workflow under-specified and therefore reduces confidence. ([inference]; medium confidence; source: https://opentelemetry.io/docs/; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5; https://help.salesforce.com/s/articleView?id=005237036&language=en_US&type=1)
  7. ServiceNow Now Assist and AI Control Tower publicly emphasize centralized inventory, lifecycle governance, compliance mapping, drift detection, and workflow-triggered remediation, but the accessible public evidence is too high-level to show a concrete per-session trace-export or configuration-export surface comparable to AWS or Salesforce. ([inference]; medium confidence; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753)
  8. Across all four platforms, the main opaque zones are the vendor-owned parts of orchestration, specifically hidden prompt augmentation, internal routing or planning logic, model-side reasoning internals, and platform-private policy engines that customers can observe only indirectly through summarized metadata or outcomes. ([inference]; medium confidence; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-runtime-generation-divergence-theory.html])
  9. An enterprise that wants reliable AIBOM automation across these platforms still needs an external control plane that normalizes tenant inventory, exported traces, approval records, and identity or tool-governance evidence into one portable schema rather than trusting any one platform’s native view to be complete. ([inference]; medium confidence; source: https://davidamitchell.github.io/Research/research/2026-04-26-vendor-platform-governance-constraints-compensating-controls.html; https://davidamitchell.github.io/Research/research/2026-04-26-ai-lowcode-observability-telemetry-governance.html; https://davidamitchell.github.io/Research/research/2026-05-02-vendor-lock-in-portability-multi-platform-ai.html; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-identity-delegation-trust-theory.html; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-runtime-generation-divergence-theory.html)

Evidence map:

Claim Source Confidence Notes
[fact] AWS Bedrock exposes configuration APIs, immutable agent versions, and step-level runtime traces suitable for automated inventory reconstruction. https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html ; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html ; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html high strongest documented API depth
[inference] AWS Bedrock logging remains incomplete outside bedrock-runtime and Bedrock does not appear to provide a native AIBOM export. https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html ; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html high exportable logs, no native bill-of-materials export
[fact] Microsoft 365 Copilot exposes strong audit, registry, and tool-governance surfaces but omits model name and model version from core Copilot audit records. https://learn.microsoft.com/en-us/purview/audit-copilot ; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide ; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/manage-tools-for-agent?view=o365-worldwide high strong admin evidence, partial runtime detail
[inference] Microsoft Copilot Studio strengthens policy and export coverage, but its packaging surface remains partial because some components and properties do not transfer. https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-data-loss-prevention ; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio ; https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-solutions-import-export medium strongest Microsoft authoring surface
[inference] Salesforce Agentforce documents session-level tracing, observability mission control, and near-real-time security telemetry that make the reviewed public material look richer on runtime observability than the standard Microsoft 365 Copilot audit surface, even though some of that gap may reflect publication differences. https://www.salesforce.com/agentforce/observability/ ; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/ ; https://www.salesforce.com/blog/best-practices-for-secure-agentforce-implementation/ ; https://learn.microsoft.com/en-us/purview/audit-copilot medium official public product and blog evidence
[inference] Salesforce likely supports OpenTelemetry, the vendor-neutral observability framework, session-trace export and agent version management, but evidence confidence is reduced because the public material leaves the exact export schema and version workflow under-specified. https://opentelemetry.io/docs/ ; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html ; https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5 ; https://help.salesforce.com/s/articleView?id=005237036&language=en_US&type=1 medium official URLs surfaced by search
[inference] ServiceNow public evidence is strongest on governance inventory and weakest on concrete trace-export and configuration-export detail. https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185 ; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104 ; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753 medium evidence quality limit
[inference] The reviewed public documentation for none of the platforms describes a native AIBOM export, so external evidence normalization remains necessary. https://owaspaibom.org/ ; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html ; https://learn.microsoft.com/en-us/purview/audit-copilot ; https://www.salesforce.com/agentforce/observability/ ; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185 medium cross-platform synthesis
[inference] Portable AIBOM automation still requires an enterprise-side control plane that binds traces, audit records, approval context, runtime-generation drift, and delegated identity evidence together. https://davidamitchell.github.io/Research/research/2026-04-26-vendor-platform-governance-constraints-compensating-controls.html ; https://davidamitchell.github.io/Research/research/2026-04-26-ai-lowcode-observability-telemetry-governance.html ; https://davidamitchell.github.io/Research/research/2026-05-02-vendor-lock-in-portability-multi-platform-ai.html ; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-runtime-generation-divergence-theory.html ; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-identity-delegation-trust-theory.html medium repository synthesis qualifier

Assumptions:

  • [assumption; source: https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5] The official Salesforce URLs surfaced by search reflect current product capability even though the accessible public material leaves some schema and workflow detail unspecified. Justification: the URLs are on official Salesforce domains and were corroborated by accessible Salesforce product and blog material.
  • [assumption; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] The accessible ServiceNow community articles are materially representative of the platform’s public governance surfaces even though the public material does not expose deeper technical export detail. Justification: the articles are on official ServiceNow domains and describe the same AI Control Tower feature set consistently.

Analysis:

The evidence separates into two clusters. AWS Bedrock exposes both declarative and runtime state through customer-addressable APIs and logs, which makes it the most suitable platform for building a machine-generated AIBOM from native evidence. [inference; source: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html] Microsoft, Salesforce, and ServiceNow are more managed, so the relevant question is not whether they provide observability at all, but whether their observability is operationally deep enough for portable inventory reconstruction rather than only for tenant governance. [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] Microsoft is strongest when the enterprise needs registry, approval, and policy control over agents and tools, but weaker when it needs transcript-complete and model-complete runtime evidence for the core Microsoft 365 Copilot experience. [inference; source: https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/manage-tools-for-agent?view=o365-worldwide; https://learn.microsoft.com/en-us/purview/audit-copilot] Salesforce appears to move furthest toward runtime introspection in the reviewed public materials because it explicitly frames session tracing and observability as operational products, yet the evidence still does not show a native AIBOM export or enough publicly accessible schema detail to treat it as equivalent to AWS API-level introspection. [inference; source: https://www.salesforce.com/agentforce/observability/; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html] Part of the apparent Salesforce-versus-Microsoft runtime gap may reflect differences in what each vendor exposes in public documentation and product marketing, rather than a fully measured difference in underlying platform capability. [inference; source: https://www.salesforce.com/agentforce/observability/; https://learn.microsoft.com/en-us/purview/audit-copilot] ServiceNow’s public story is governance-first rather than trace-first, which is useful for enterprise oversight but insufficient by itself for proving that a portable, field-level AIBOM can be generated from native product surfaces. [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753] The main trade-off is therefore between runtime depth and platform abstraction: the more the platform centralizes orchestration on the vendor side, the more governance becomes curated metadata rather than complete customer-visible evidence. [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/] Alternative remedies such as adding more human reviewers, tightening model-quality gates, or relying only on richer tenant governance interfaces can improve oversight, but they do not by themselves create the missing portable configuration and trace exports needed for repeatable AIBOM generation. [inference; source: https://davidamitchell.github.io/Research/research/2026-04-26-vendor-platform-governance-constraints-compensating-controls.html; https://davidamitchell.github.io/Research/research/2026-04-26-ai-lowcode-observability-telemetry-governance.html; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html]

Risks, gaps, uncertainties:

  • Public Salesforce material reviewed here confirms observability and security surfaces, but it does not expose enough detailed schema documentation to verify the exact customer-visible mechanics of session-trace export and agent version management with high confidence. [inference; source: https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5; https://www.salesforce.com/agentforce/observability/]
  • Public ServiceNow material reviewed here confirms governance, inventory, and remediation framing, but it does not expose enough technical detail to verify a customer-visible AI Control Tower export interface or per-session trace schema with high confidence. [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753]
  • The reviewed public docs do not prove whether hidden vendor-side prompt augmentation or internal routing is never logged anywhere; they only show which surfaces are customer-visible through documented interfaces. [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/]
  • Microsoft 365 Copilot’s deeper transcript access can depend on additional Purview or security products, so the standard audit stream should be treated as the default documented evidence path rather than the only possible path. [inference; source: https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio; https://learn.microsoft.com/en-us/purview/audit-copilot]

Open questions:

  • What exact field schema does Salesforce’s OpenTelemetry session-trace export expose, and is it complete enough to populate a runtime AIBOM without auxiliary APIs?
  • Does ServiceNow expose a programmatic AI Control Tower export or trace API in licensed documentation that is not visible from public community material?
  • Can Microsoft 365 Copilot’s newer agent-registry and tools-governance surfaces be joined reliably with Purview audit and Data Security Posture Management for AI to create a tenant-wide partial AIBOM?
  • What normalization schema best reconciles Bedrock trace events, Microsoft Purview audit records, Salesforce OpenTelemetry exports, and ServiceNow AI inventory records into one portable evidence model?

§7 Recursive Review

  • Outcome: pass
  • Confidence: medium
  • [fact; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185] Notes: Bedrock and Microsoft claims rest on directly specified official documentation, while Salesforce and ServiceNow findings rely more heavily on higher-level public material and therefore remain medium-confidence.

Findings

Executive Summary

AWS Bedrock Agents provide the strongest documented substrate for automated Artificial Intelligence Bill of Materials (AIBOM) generation among the four reviewed platforms because AWS exposes both a machine-readable configuration API and detailed step-level runtime traces with immutable version snapshots. [inference; source: https://owaspaibom.org/; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html] Microsoft 365 Copilot provides a strong tenant-governance surface through audit, inventory, and tool-approval controls, but its main audit stream omits critical runtime detail such as model name and model version for Microsoft 365 Copilot and does not provide full transcript content in the standard audit record. [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide] Salesforce Agentforce documents extensive managed-platform runtime observability through session-level tracing and observability tooling, but the reviewed public evidence leaves export semantics and versioning detail under-specified, so confidence is lower than for AWS. [inference; source: https://www.salesforce.com/agentforce/observability/; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/] ServiceNow Now Assist and AI Control Tower emphasize centralized governance, inventory, and oversight, but the accessible public evidence reviewed here is more strategic than technical, which makes detailed AIBOM automation claims the weakest of the four. [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] The reviewed public documentation for none of the four platforms describes a native AIBOM or equivalent Bill of Materials export, so an enterprise that wants portable inventory assurance still needs an external normalization and evidence-binding layer. [inference; source: https://owaspaibom.org/; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-runtime-generation-divergence-theory.html; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-identity-delegation-trust-theory.html]

Key Findings

  1. AWS Bedrock Agents are the strongest documented Artificial Intelligence Bill of Materials (AIBOM) substrate in this comparison because AWS exposes agent configuration through the GetAgent API, immutable alias-based versions, and step-level runtime traces that include prompts, rationale, action invocations, observations, and version identifiers. ([inference]; high confidence; source: https://owaspaibom.org/; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html])
  2. AWS Bedrock still appears not to provide a native AIBOM export, and its observability remains incomplete when traffic bypasses the documented bedrock-runtime logging path or when the enterprise needs provider-side model internals rather than customer-visible orchestration evidence. ([inference]; high confidence; source: https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html)
  3. Microsoft 365 Copilot exposes strong tenant-governance evidence through Purview audit, the Microsoft 365 agent registry, and tool-approval workflows, but the standard audit schema explicitly omits model name and model version for Microsoft 365 Copilot scenarios and does not function as a full runtime trace. ([fact]; high confidence; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/manage-tools-for-agent?view=o365-worldwide)
  4. Microsoft Copilot Studio adds real-time data-loss-prevention controls, agent export and import through solutions, and detailed authoring and usage audit events, which strengthens Microsoft’s AIBOM readiness even though its export surface remains partial because some components and properties do not transfer cleanly. ([inference]; medium confidence; source: https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-data-loss-prevention; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio; https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-solutions-import-export)
  5. Salesforce Agentforce publicly documents mission-control observability, session-level tracing, and near-real-time security telemetry, including Event Monitoring and transaction-security controls around agent activity, which makes the reviewed public material look richer on runtime observability than the standard Microsoft 365 Copilot audit surface even though some of that gap may reflect how the vendors publish their capabilities. ([inference]; medium confidence; source: https://www.salesforce.com/agentforce/observability/; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/; https://www.salesforce.com/blog/best-practices-for-secure-agentforce-implementation/; https://learn.microsoft.com/en-us/purview/audit-copilot)
  6. Salesforce Agentforce also appears to support session-trace export and agent version management using OpenTelemetry, the vendor-neutral observability framework, which would make automated AIBOM generation more feasible than on Microsoft 365 Copilot, but the public evidence leaves the exact export schema and version workflow under-specified and therefore reduces confidence. ([inference]; medium confidence; source: https://opentelemetry.io/docs/; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5; https://help.salesforce.com/s/articleView?id=005237036&language=en_US&type=1)
  7. ServiceNow Now Assist and AI Control Tower publicly emphasize centralized inventory, lifecycle governance, compliance mapping, drift detection, and workflow-triggered remediation, but the accessible public evidence is too high-level to show a concrete per-session trace-export or configuration-export surface comparable to AWS or Salesforce. ([inference]; medium confidence; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753)
  8. Across all four platforms, the main opaque zones are the vendor-owned parts of orchestration, specifically hidden prompt augmentation, internal routing or planning logic, model-side reasoning internals, and platform-private policy engines that customers can observe only indirectly through summarized metadata or outcomes. ([inference]; medium confidence; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-runtime-generation-divergence-theory.html])
  9. An enterprise that wants reliable AIBOM automation across these platforms still needs an external control plane that normalizes tenant inventory, exported traces, approval records, and identity or tool-governance evidence into one portable schema rather than trusting any one platform’s native view to be complete. ([inference]; medium confidence; source: https://davidamitchell.github.io/Research/research/2026-04-26-vendor-platform-governance-constraints-compensating-controls.html; https://davidamitchell.github.io/Research/research/2026-04-26-ai-lowcode-observability-telemetry-governance.html; https://davidamitchell.github.io/Research/research/2026-05-02-vendor-lock-in-portability-multi-platform-ai.html; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-identity-delegation-trust-theory.html; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-runtime-generation-divergence-theory.html)

Evidence Map

Claim Source Confidence Notes
[fact] AWS Bedrock exposes configuration APIs, immutable agent versions, and step-level runtime traces suitable for automated inventory reconstruction. https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html ; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html ; https://docs.aws.amazon.com/bedrock/latest/userguide/deploy-agent.html high strongest documented API depth
[inference] AWS Bedrock logging remains incomplete outside bedrock-runtime and Bedrock does not appear to provide a native AIBOM export. https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html ; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html high exportable logs, no native bill-of-materials export
[fact] Microsoft 365 Copilot exposes strong audit, registry, and tool-governance surfaces but omits model name and model version from core Copilot audit records. https://learn.microsoft.com/en-us/purview/audit-copilot ; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide ; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/manage-tools-for-agent?view=o365-worldwide high strong admin evidence, partial runtime detail
[inference] Microsoft Copilot Studio strengthens policy and export coverage, but its packaging surface remains partial because some components and properties do not transfer. https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-data-loss-prevention ; https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio ; https://learn.microsoft.com/en-us/microsoft-copilot-studio/authoring-solutions-import-export medium strongest Microsoft authoring surface
[inference] Salesforce Agentforce documents session-level tracing, observability mission control, and near-real-time security telemetry that make the reviewed public material look richer on runtime observability than the standard Microsoft 365 Copilot audit surface, even though some of that gap may reflect publication differences. https://www.salesforce.com/agentforce/observability/ ; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/ ; https://www.salesforce.com/blog/best-practices-for-secure-agentforce-implementation/ ; https://learn.microsoft.com/en-us/purview/audit-copilot medium official public product and blog evidence
[inference] Salesforce likely supports OpenTelemetry, the vendor-neutral observability framework, session-trace export and agent version management, but evidence confidence is reduced because the public material leaves the exact export schema and version workflow under-specified. https://opentelemetry.io/docs/ ; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html ; https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5 ; https://help.salesforce.com/s/articleView?id=005237036&language=en_US&type=1 medium official URLs surfaced by search
[inference] ServiceNow public evidence is strongest on governance inventory and weakest on concrete trace-export and configuration-export detail. https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185 ; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104 ; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753 medium evidence quality limit
[inference] The reviewed public documentation for none of the platforms describes a native AIBOM export, so external evidence normalization remains necessary. https://owaspaibom.org/ ; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html ; https://learn.microsoft.com/en-us/purview/audit-copilot ; https://www.salesforce.com/agentforce/observability/ ; https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185 medium cross-platform synthesis
[inference] Portable AIBOM automation still requires an enterprise-side control plane that binds traces, audit records, approval context, runtime-generation drift, and delegated identity evidence together. https://davidamitchell.github.io/Research/research/2026-04-26-vendor-platform-governance-constraints-compensating-controls.html ; https://davidamitchell.github.io/Research/research/2026-04-26-ai-lowcode-observability-telemetry-governance.html ; https://davidamitchell.github.io/Research/research/2026-05-02-vendor-lock-in-portability-multi-platform-ai.html ; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-runtime-generation-divergence-theory.html ; https://davidamitchell.github.io/Research/research/2026-05-06-aibom-identity-delegation-trust-theory.html medium repository synthesis qualifier

Assumptions

  • Assumption: The official Salesforce URLs surfaced by search reflect current product capability even though the accessible public material leaves some schema and workflow detail unspecified. Justification: the URLs are on official Salesforce domains and were corroborated by accessible Salesforce product and blog material. [assumption; source: https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5; https://www.salesforce.com/agentforce/observability/]
  • Assumption: The accessible ServiceNow community articles are materially representative of the platform’s public governance surfaces even though the public material does not expose deeper technical export detail. Justification: the articles are on official ServiceNow domains and describe the same AI Control Tower feature set consistently. [assumption; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104]

Analysis

The evidence separates into two clusters. AWS Bedrock exposes both declarative and runtime state through customer-addressable APIs and logs, which makes it the most suitable platform for building a machine-generated AIBOM from native evidence. [inference; source: https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html; https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://docs.aws.amazon.com/bedrock/latest/userguide/model-invocation-logging.html] Microsoft, Salesforce, and ServiceNow are more managed, so the relevant question is not whether they provide observability at all, but whether their observability is operationally deep enough for portable inventory reconstruction rather than only for tenant governance. [inference; source: https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/; https://www.servicenow.com/community/now-assist-articles/ai-control-tower-an-executive-view-on-ai-governance/ta-p/3378104] Microsoft is strongest when the enterprise needs registry, approval, and policy control over agents and tools, but weaker when it needs transcript-complete and model-complete runtime evidence for the core Microsoft 365 Copilot experience. [inference; source: https://learn.microsoft.com/en-us/microsoft-365/admin/manage/agent-registry?view=o365-worldwide; https://learn.microsoft.com/en-us/microsoft-365/admin/manage/manage-tools-for-agent?view=o365-worldwide; https://learn.microsoft.com/en-us/purview/audit-copilot] Salesforce appears to move furthest toward runtime introspection in the reviewed public materials because it explicitly frames session tracing and observability as operational products, yet the evidence still does not show a native AIBOM export or enough publicly accessible schema detail to treat it as equivalent to AWS API-level introspection. [inference; source: https://www.salesforce.com/agentforce/observability/; https://www.salesforce.com/blog/secure-agentforce-with-trusted-services/; https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html] Part of the apparent Salesforce-versus-Microsoft runtime gap may reflect differences in what each vendor exposes in public documentation and product marketing, rather than a fully measured difference in underlying platform capability. [inference; source: https://www.salesforce.com/agentforce/observability/; https://learn.microsoft.com/en-us/purview/audit-copilot] ServiceNow’s public story is governance-first rather than trace-first, which is useful for enterprise oversight but insufficient by itself for proving that a portable, field-level AIBOM can be generated from native product surfaces. [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753] The main trade-off is therefore between runtime depth and platform abstraction: the more the platform centralizes orchestration on the vendor side, the more governance becomes curated metadata rather than complete customer-visible evidence. [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/] Alternative remedies such as adding more human reviewers, tightening model-quality gates, or relying only on richer tenant governance interfaces can improve oversight, but they do not by themselves create the missing portable configuration and trace exports needed for repeatable AIBOM generation. [inference; source: https://davidamitchell.github.io/Research/research/2026-04-26-vendor-platform-governance-constraints-compensating-controls.html; https://davidamitchell.github.io/Research/research/2026-04-26-ai-lowcode-observability-telemetry-governance.html; https://docs.aws.amazon.com/bedrock/latest/APIReference/API_agent_GetAgent.html]

Risks, Gaps, and Uncertainties

  • Public Salesforce material reviewed here confirms observability and security surfaces, but it does not expose enough detailed schema documentation to verify the exact customer-visible mechanics of session-trace export and agent version management with high confidence. [inference; source: https://developer.salesforce.com/docs/ai/agentforce/guide/otel-api.html; https://help.salesforce.com/s/articleView?id=ai.agent_versions.htm&language=en_US&type=5; https://www.salesforce.com/agentforce/observability/]
  • Public ServiceNow material reviewed here confirms governance, inventory, and remediation framing, but it does not expose enough technical detail to verify a customer-visible AI Control Tower export interface or per-session trace schema with high confidence. [inference; source: https://www.servicenow.com/community/admin-experience-blogs/introducing-the-servicenow-ai-control-tower-from-intelligent/ba-p/3261185; https://www.servicenow.com/community/product-launch-forum/using-the-ai-control-tower-in-servicenow-for-enterprise-wide-ai/m-p/3421753]
  • The reviewed public docs do not prove whether hidden vendor-side prompt augmentation or internal routing is never logged anywhere; they only show which surfaces are customer-visible through documented interfaces. [inference; source: https://docs.aws.amazon.com/bedrock/latest/userguide/trace-events.html; https://learn.microsoft.com/en-us/purview/audit-copilot; https://www.salesforce.com/agentforce/observability/]
  • Microsoft 365 Copilot’s deeper transcript access can depend on additional Purview or security products, so the standard audit stream should be treated as the default documented evidence path rather than the only possible path. [inference; source: https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-logging-copilot-studio; https://learn.microsoft.com/en-us/purview/audit-copilot]

Open Questions

  • What exact field schema does Salesforce’s OpenTelemetry session-trace export expose, and is it complete enough to populate a runtime AIBOM without auxiliary APIs?
  • Does ServiceNow expose a programmatic AI Control Tower export or trace API in licensed documentation that is not visible from public community material?
  • Can Microsoft 365 Copilot’s newer agent-registry and tools-governance surfaces be joined reliably with Purview audit and Data Security Posture Management for AI to create a tenant-wide partial AIBOM?
  • What normalization schema best reconciles Bedrock trace events, Microsoft Purview audit records, Salesforce OpenTelemetry exports, and ServiceNow AI inventory records into one portable evidence model?

Output

Navigation

Home

By Tag

bureaucracy

change-management

coase

constraint-analysis

control-model

decision-rights

delegation

delivery-risk

demand-segmentation

enterprise

exception-handling

execution

flow

flow-design

flow-metrics

governance

governance-patterns

incentives

instability

institutional-economics

leading-indicators

operating-model

organisation

organisational-design

queue-design

queueing

regulated-enterprise

routing

throughput

throughput-risk

transaction-costs

triage

williamson

Clone this wiki locally