2026 05 16 do mode demand persistence and build mode displacement

Temporary Automation Demand Persistence and Core Capability Investment Displacement

Research Question

What evidence exists that temporary automation workarounds displace investment in core software delivery, and what is the observed persistence rate of those workarounds after the underlying systems capability gap has been closed?

Scope

In scope:

Comparative evidence from low-code, Robotic Process Automation (RPA), and Artificial Intelligence (AI) agent programmes.
Decommission outcomes after software delivery closes the original capability gap.
Governance and incentive mechanisms that reduce displacement of core software-delivery investment.

Out of scope:

Prescriptive target operating model design for a specific enterprise.
Evaluation of non-enterprise consumer automation tools.

Constraints:

Use comparable organisations and time windows where possible.
Distinguish correlation from causal displacement claims.

Context

[assumption; source: https://davidamitchell.github.io/Research/research/2026-04-26-systems-capability-debt-agentic-ai-risk-synthesis.html; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html] Working definition: temporary automation workarounds are local low-code, RPA, or AI solutions introduced to bridge a missing integration, workflow, or data-access capability until core software delivery closes that gap.

[assumption; source: https://davidamitchell.github.io/Research/research/2026-03-07-run-build-it-allocation-implementation-how.html; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html] Working definition: core software-delivery investment means build-mode spending that removes the underlying capability gap in the platform, data, or integration layer rather than automating around it.

[inference; source: https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/] The empirical test is whether these bridges merely absorb short-term demand spikes or whether they divert attention, operating budget, and governance energy away from closing the underlying gap.

Approach

Review empirical studies and benchmark datasets for low-code, RPA, and agent investment outcomes.
Extract persistence and decommission evidence for automation workarounds after corresponding software-delivery events.
Evaluate governance and incentive designs that successfully protect core software-delivery capacity.

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

Question: what evidence shows that temporary automation workarounds divert demand away from closing the underlying systems capability gap, and what public evidence exists on how long those workarounds persist after the gap is closed?
Scope: low-code, RPA, and AI-agent programmes; decommission and persistence evidence; governance and incentive mechanisms that preserve core software-delivery capacity.
Constraints: use public and accessible sources, distinguish behaviour from proven budget crowd-out, and record explicit gaps where the evidence base does not publish rates.
Output: knowledge.
Prior completed-item sweep: https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html ; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html ; https://davidamitchell.github.io/Research/research/2026-05-14-citizen-development-rollout-empirical-evidence.html ; https://davidamitchell.github.io/Research/research/2026-05-08-shadow-ai-behavioral-drivers-governance-effectiveness.html ; https://davidamitchell.github.io/Research/research/2026-04-26-systems-capability-debt-agentic-ai-risk-synthesis.html
[assumption; source: https://davidamitchell.github.io/Research/research/2026-04-26-systems-capability-debt-agentic-ai-risk-synthesis.html; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html] Working definition: a temporary automation workaround is a local low-code, RPA, or AI bridge introduced because the core platform cannot yet deliver the needed process, data, or integration capability.
[assumption; source: https://davidamitchell.github.io/Research/research/2026-03-07-run-build-it-allocation-implementation-how.html; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html] Working definition: core capability investment means build-mode software delivery that removes the missing capability in the underlying system rather than automating around it.
[fact; source: https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html; https://davidamitchell.github.io/Research/research/2026-05-08-shadow-ai-behavioral-drivers-governance-effectiveness.html; https://davidamitchell.github.io/Research/research/2026-04-26-systems-capability-debt-agentic-ai-risk-synthesis.html] Prior completed items already support three adjacent propositions: workaround demand rises when systems capability gaps persist, recurring workaround operation does not remove the long-run case for closing stable gaps in software, and post-rollout users continue bypassing sanctioned tools when feature or workflow gaps remain.

§1 Question Decomposition

Root question: Do temporary automation bridges actually displace core software-delivery investment, and how reliably are they retired after the gap closes?
A. Demand displacement
- A1. What evidence shows that business users adopt temporary automation because central delivery or sanctioned tools are too slow or inadequate?
- A2. Does the evidence prove direct budget or headcount crowd-out, or only operational demand diversion?
- A3. Do low-code, RPA, and AI-agent sources point in the same causal direction?
B. Persistence after gap closure
- B1. What public evidence describes bot, app, or agent persistence after the original business need should have disappeared?
- B2. Do public sources publish a post-gap-closure persistence rate?
- B3. Which observable trigger surfaces support safe retirement?
C. Protective governance
- C1. Which governance mechanisms keep local automation from becoming a substitute for core system change?
- C2. Which incentive mechanisms keep business units collaborating with central Information Technology (IT) rather than bypassing it?
- C3. Which controls make decommission operational rather than aspirational?
D. Synthesis
- D1. What does the evidence support strongly?
- D2. What remains uncertain?
- D3. What operating model follows from that evidence?

§2 Investigation

Source-access and replacement notes

Access note, seeded IEEE Xplore document identifier: unrelated IEEE metadata in this session; not used downstream.
Access note, seeded Deloitte URL: 404 in this session; replaced with Deloitte's current 2020 intelligent automation survey page.
Access note, seeded Harvard Business Review direct article URL: 404 in this session; replaced with the official May 2016 issue index, but article text not accessible for downstream support.
Access note, Institute for Robotic Process Automation and Artificial Intelligence homepage: readable in this session, but no specific outcome or lifecycle evidence extracted for downstream claims.
Failed primary-source search record, persistence-rate query set: "post-gap-closure" automation decommission rate, "RPA bot" retirement rate after system upgrade, and "low-code app" decommission rate after replacement; no accessible public dataset found.

A. Evidence for demand displacement is behavioural and operational, not accounting-grade

[fact; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/] Binzer et al. report from 24 companies that low-code and no-code platforms let employees with little or no IT background quickly create digital solutions, but that success depends on explicit design choices around security, compliance, and organisational change.
[fact; source: https://aisel.aisnet.org/misqe/vol23/iss3/6/] Viljoen et al. report from 30 interviews that entrusting software development to novices creates risks of substandard software quality, shadow IT, and technical debt, and that governance must rely on technical experts plus platform-specific controls.
[fact; source: https://hdl.handle.net/10125/108890] Binzer et al.'s 18-firm multi-case study identifies coordinated scaling, project execution, and professional catalyst roles as distinct collaboration mechanisms, which indicates that local citizen development does not scale safely without explicit business-IT coordination.
[fact; source: https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/] Ajimati et al. review 40 primary studies and conclude that low-code and no-code adoption consistently broadens participation and speeds local digital transformation, but it also introduces governance and complexity challenges that remain active research concerns.
[fact; source: https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html] Deloitte reports that 73% of surveyed organisations had embarked on intelligent automation by 2020, but only 26% of piloting organisations and 38% of implementing or scaling organisations had an enterprise-wide automation strategy.
[fact; source: https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html] Deloitte also reports that process fragmentation and lack of IT readiness were the top barriers to success, while organisations in the piloting stage were more likely to automate current processes with limited change.
[fact; source: https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks] IBM reports that 80% of surveyed American office workers use AI in their roles, but only 22% rely exclusively on employer-provided tools.
[fact; source: https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks] IBM also reports that nearly 40% of workers prefer external AI solutions for their better features and that many turn to unauthorised platforms when company-provided solutions fail to meet their needs.
[fact; source: https://www.ibm.com/think/topics/shadow-ai] IBM defines shadow AI as unsanctioned AI use and says employees turn to unsanctioned AI when approved options are too slow or existing solutions are insufficient.
[fact; source: https://davidamitchell.github.io/Research/research/2026-05-08-shadow-ai-behavioral-drivers-governance-effectiveness.html] The nearest prior completed item in this repository finds the same post-rollout pattern in adjacent evidence: sanctioned provision does not end bypass behaviour when feature gaps, workflow friction, or speed pressure remain.
[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai] Across low-code, RPA, and AI-adoption sources, the best-supported displacement mechanism is demand diversion: when the sanctioned or central path is too slow, fragmented, or inadequate, users solve locally instead of waiting for core system change.
[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks] The public evidence does not prove a clean budget or headcount crowd-out rate, because the accessible studies describe adoption behaviour, governance strain, and feature gaps rather than audited shifts in build budgets.

B. Public evidence supports persistence risk, but not a published post-gap-closure persistence rate

[fact; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots] Pega states that organisations often automate legacy processes on top of legacy systems, which means bot build, support, and upgrade costs are added to the existing cost of keeping the underlying systems running.
[fact; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots] Pega also says some customers do not build RPA bots without first having an end-of-life plan for each bot and recommends decomposing automated processes to identify overlap that should be eliminated or rebuilt more durably.
[fact; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components] Microsoft's inactivity-notification guidance exists specifically to detect unused canvas apps and cloud flows, route approvals, and optionally auto-delete items approved for deletion.
[fact; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components] The same guidance sets the default inactivity interval to six months and exposes an Auto Delete on Archive configuration that can delete apps and flows after approval.
[fact; source: https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup] Microsoft automatically disables unused developer environments after 30 days of inactivity and deletes them 15 days later unless activity resumes or the environment is re-enabled.
[fact; source: https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup] Microsoft counts automated behaviours such as scheduled flow runs as activity, which shows that retirement logic is bound to observable runtime evidence rather than to owner declaration.
[fact; source: https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory] Power Platform inventory gives administrators a unified view of agents, apps, and flows, updates created, updated, or deleted resources within 15 minutes, and explicitly aims to prevent orphaned agents by finding resources owned by departing users.
[fact; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-orphan-components; https://learn.microsoft.com/en-us/power-platform/guidance/adoption/reactive-governance] Microsoft's orphaned-object and reactive-governance guidance routes ownerless and inactive resources into reassignment, review, or cleanup workflows rather than assuming they disappear naturally.
[fact; source: https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data] UiPath allows deletion only to account owners and system administrators and blocks App Inventory deletion while the application is still used in ideas or components.
[fact; source: https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html] The nearest prior completed repository item concludes that machine-verifiable retirement surfaces, such as inactivity, ownerlessness, and dependency elimination, are more reliable than manual owner self-declaration.
[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data] The presence of productised inactivity reviews, orphan detection, and dependency-aware deletion across major automation platforms supports the inference that workaround persistence is common enough to justify built-in lifecycle controls.
[inference; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/] The reviewed public sources used in this item do not yield a cross-organisational persistence percentage for low-code apps, RPA bots, or AI workarounds after the underlying capability gap has been closed.

C. The strongest protective mechanisms keep local automation tied to central governance and eventual core-system change

[fact; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/] Digital.gov recommends a centralised code repository, code review, role-based access controls, and separate development, test, and production environments for citizen-developed RPA.
[fact; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/] Digital.gov also says citizen-development success depends on collaboration with Chief Information Officer (CIO) shops and IT departments, leadership support, disciplined license allocation, and qualitative return-on-investment evaluation rather than speed alone.
[fact; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://hdl.handle.net/10125/108890] The low-code empirical studies consistently frame citizen development as a business-IT collaboration problem rather than as a purely local productivity tactic.
[fact; source: https://aisel.aisnet.org/misqe/vol23/iss3/6/] Viljoen et al. say technical experts and platform-specific functionalities are crucial to governing citizen development, which means local automation remains dependent on centrally supplied control surfaces even when creation is decentralised.
[fact; source: https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html] The adjacent cost-comparison item concludes that recurring workaround operation rarely beats closing a stable systems capability gap over multi-year horizons once governance, recovery, and review costs are included.
[inference; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://hdl.handle.net/10125/108890; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html] Governance is more likely to protect core capability investment when it forces local automation to remain visible, reviewable, and linked to central collaboration instead of rewarding isolated speed alone.
[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html] The best-supported operating pattern is therefore to treat temporary automation as a governed bridge with an explicit registry, inactivity window, dependency check, and end-of-life condition tied to observable replacement behaviour.

§3 Reasoning

[fact; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks] The accessible evidence base directly supports adoption pressure, governance strain, and sanctioned-tool inadequacy.
[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://www.ibm.com/think/topics/shadow-ai; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks] The strongest causal interpretation is not that organisations deliberately cut build budgets first, but that users divert demand into local workarounds when delivery friction remains unresolved.
[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data] Persistence is best supported as a lifecycle-control problem rather than as a measured industry rate, because the evidence is rich on retirement mechanisms and poor on published retirement outcomes.
[assumption; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots] If an organisation lacks inventory, inactivity review, and explicit end-of-life design, workaround persistence will be materially higher than in governed estates. Justification: every strong operational source assumes these controls are necessary and treats their absence as a risk surface.
[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://hdl.handle.net/10125/108890; https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/] A credible alternative interpretation is complementarity rather than displacement, because coordinated citizen-development programmes can surface demand, improve process understanding, and feed later core-system change. The evidence supports that alternative only when local automation remains inside business-IT collaboration and lifecycle controls, which means complementarity is conditional rather than automatic.

§4 Consistency Check

[fact; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks] No consulted source contradicts the basic direction of the behavioural mechanism: local automation demand rises when central or sanctioned options are too slow, fragmented, or inadequate.
[inference; source: https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots] The apparent tension between automation-growth narratives and persistence-risk narratives resolves once local efficiency gains are separated from underlying capability closure. An organisation can scale automation adoption and still deepen the need for later core-system change.
[fact; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data] The consulted platform sources agree that deletion should follow observable inactivity, ownership, approval, and dependency checks, which is consistent with the prior completed-item decommission synthesis.
[inference; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/] Confidence on the persistence-rate question must remain medium overall because the evidence supports the mechanism of persistence but does not publish the requested rate.

§5 Depth and Breadth Expansion

[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html] Technical lens: UI-layer automation and process-wrapping increase fragility when the underlying system keeps changing, so layered workarounds can expand maintenance load even while they deliver local speed.
[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html] Economic lens: the workaround and the legacy capability can coexist as additive cost stacks, so temporary automation can reduce pressure to build in the short run while worsening the long-run economic case for leaving the gap open.
[inference; source: https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai; https://davidamitchell.github.io/Research/research/2026-05-08-shadow-ai-behavioral-drivers-governance-effectiveness.html] Behavioural lens: sanctioned rollout alone does not extinguish bypass behaviour, because workers compare actual capability and workflow friction, not policy intent.
[fact; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/guidance/adoption/reactive-governance] Governance lens: the control surfaces that matter most are visibility, review, ownership assignment, and cleanup triggers rather than a single top-down prohibition.
[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://hdl.handle.net/10125/108890; https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/] Organisational lens: local innovation survives best when central teams provide review, reusable controls, and escalation paths to durable builds instead of forcing a binary choice between full prohibition and unmanaged autonomy.

§6 Synthesis

Executive Summary

[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai] Public evidence supports a displacement mechanism, but not a published displacement rate: temporary automation workarounds divert demand away from waiting for core software delivery when central or sanctioned options are too slow, fragmented, or inadequate.
[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data] Public sources also support persistence risk, because major platforms expose explicit inactivity, orphan-detection, approval, and dependency-aware retirement controls, and vendor lifecycle guidance treats bot end-of-life planning as necessary rather than exceptional.
[fact; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/] The accessible public evidence used here does not publish a robust post-gap-closure persistence percentage for low-code apps, RPA bots, or AI workarounds.
[inference; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://hdl.handle.net/10125/108890; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html] The best-supported governance response is to treat temporary automation as a visible bridge with central review, registry, runtime retirement triggers, and an explicit path back into core capability delivery.
[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://hdl.handle.net/10125/108890; https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/] That conclusion does not rule out complementarity, because governed citizen-development programmes can also surface demand and process knowledge that later help core-system teams build the durable fix.

Key Findings

[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai] Temporary automation displaces core software-delivery demand primarily by giving users a faster local alternative when central Information Technology delivery or sanctioned tools cannot meet workflow needs. (medium confidence)
[inference; source: https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://aisel.aisnet.org/misqe/vol23/iss3/3/] Automation adoption often scales ahead of enterprise strategy and IT readiness, which supports the inference that local delivery can expand before the organisation improves the underlying capability base. (medium confidence)
[fact; source: https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai; https://davidamitchell.github.io/Research/research/2026-05-08-shadow-ai-behavioral-drivers-governance-effectiveness.html] Post-rollout shadow Artificial Intelligence behaviour shows that sanctioned provision does not reliably eliminate workaround demand when users still perceive external tools as better or faster. (medium confidence)
[fact; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html] RPA and adjacent workaround programmes can create additive operating cost because bot support and upgrade effort persists while the underlying legacy system cost remains in place. (medium confidence)
[inference; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data] Major automation platforms have built retirement features around inactivity, ownerlessness, approvals, and dependency checks, which supports the inference that persistence of unused or obsolete workarounds is common enough to require explicit lifecycle controls. (medium confidence)
[inference; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/] The reviewed public evidence base does not yield a reliable cross-organisational persistence rate after the underlying capability gap has been closed, so the rate question remains unresolved. (medium confidence)
[inference; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://hdl.handle.net/10125/108890] The evidence most consistently points to central repositories, code review, role-based access control, separate environments, business-IT collaboration, and leadership-backed programme design as the governance bundle most likely to protect core build capacity. (medium confidence)
[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html] The most defensible operating model is to treat temporary automation as a governed bridge with explicit end-of-life conditions rather than as a standing substitute for fixing the underlying system. (medium confidence)

Evidence Map

Claim	Source	Confidence	Notes
[inference] Temporary automation diverts demand away from waiting for core software delivery when central or sanctioned options are too slow or inadequate.	https://aisel.aisnet.org/misqe/vol23/iss3/3/ ; https://aisel.aisnet.org/misqe/vol23/iss3/6/ ; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html ; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks ; https://www.ibm.com/think/topics/shadow-ai	medium	Behavioural mechanism, not budget ledger
[fact] Automation adoption often scales ahead of enterprise strategy and IT readiness.	https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html ; https://aisel.aisnet.org/misqe/vol23/iss3/3/	high	Strategy and readiness gap
[fact] Sanctioned tool rollout does not reliably eliminate workaround demand.	https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks ; https://www.ibm.com/think/topics/shadow-ai ; https://davidamitchell.github.io/Research/research/2026-05-08-shadow-ai-behavioral-drivers-governance-effectiveness.html	medium	Post-rollout bypass persists
[fact] Workaround programmes can create additive cost because the legacy stack remains while the workaround must still be maintained.	https://community.pega.com/blog/when-it-time-retire-your-rpa-bots ; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html	medium	Cost-stack effect
[inference] Major platforms productise retirement controls around inactivity, ownership, approvals, and dependency checks, which supports the view that persistence is common enough to require explicit lifecycle controls.	https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components ; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory ; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup ; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data	medium	Inference from control design
[inference] The reviewed public evidence base does not yield a reliable post-gap-closure persistence percentage.	https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components ; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup ; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data ; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/	medium	Governance-rich, outcome-light evidence base
[inference] The evidence most consistently points to a governance bundle of technical controls, central collaboration, and leadership backing as the pattern most likely to protect core build capacity.	https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/ ; https://aisel.aisnet.org/misqe/vol23/iss3/3/ ; https://aisel.aisnet.org/misqe/vol23/iss3/6/ ; https://hdl.handle.net/10125/108890	medium	Comparative conclusion
[inference] Temporary automation should be operated as a governed bridge with explicit retirement conditions.	https://community.pega.com/blog/when-it-time-retire-your-rpa-bots ; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components ; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory ; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html	medium	Best-supported operating model

Assumptions

[assumption; source: https://davidamitchell.github.io/Research/research/2026-04-26-systems-capability-debt-agentic-ai-risk-synthesis.html; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html] This item treats low-code apps, RPA bots, and AI-agent workarounds as comparable bridge patterns when they exist primarily to compensate for the same underlying capability gap. Justification: the sources differ by tooling class, but they converge on the same lifecycle problem of local automation persisting beyond its intended bridging purpose.
[assumption; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots] The existence of built-in retirement controls is treated as evidence that obsolete workarounds are common enough to matter operationally. Justification: platform vendors typically do not invest in inventory, orphan cleanup, and auto-delete controls for purely hypothetical problems.

Analysis

[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html] The evidence is strongest on why local workaround demand appears: users choose the fastest viable path when central delivery is slow, fragmented, or strategically underprepared.
[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data] The evidence is weaker on persistence measurement, but it is still enough to reject the assumption that workaround retirement happens naturally once the core system catches up.
[inference; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://hdl.handle.net/10125/108890; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html] The key trade-off is speed versus durable capability: unmanaged workarounds relieve local pressure quickly, while governed bridges preserve the option and incentive to invest in the underlying system.

Risks, Gaps, and Uncertainties

[fact; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/] No accessible public source in this item reports a quantitative post-gap-closure persistence rate, so the rate question remains open.
[inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks] Evidence for displacement is behavioural and organisational rather than financial-accounting grade, so this item supports a causal mechanism but not a precise crowd-out percentage.
[inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html] Pega and adjacent repository items sharpen the cost and lifecycle logic, but they do not substitute for a longitudinal multi-organisation dataset on workaround retirement after capability closure.

Open Questions

[inference; source: https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components] Which organisations publish inventory-level longitudinal data that could reveal the actual retirement rate of apps, flows, and agents after replacement capabilities go live?
[inference; source: https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots] What budget-governance designs most effectively prevent short-term automation success from suppressing approval for deeper system remediation?
[inference; source: https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai] How does the persistence pattern differ between consumer-style shadow Artificial Intelligence use and centrally registered enterprise agent platforms?

§7 Recursive Review

Review result: pass
Acronym audit: Artificial Intelligence (AI), Robotic Process Automation (RPA), Information Technology (IT), and Chief Information Officer (CIO) expanded on first use in the document body.
Claim audit: every substantive claim in Research Skill Output is labelled as fact, inference, or assumption and bound to URL-backed support.
Output audit: synthesis and Findings remain aligned on the same substantive conclusions, with confidence reduced to medium because the public persistence-rate question remains unresolved.

Findings

Executive Summary

Temporary automation workarounds displace core software-delivery demand mainly by giving users a faster local path when central Information Technology delivery or sanctioned tools cannot meet their needs, but the accessible public evidence does not publish a robust displacement percentage. [inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai]

The same evidence base shows strong persistence risk, because major automation platforms and bot vendors expose explicit controls for inactivity review, orphan detection, dependency-aware deletion, and end-of-life planning. [inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data]

The reviewed public evidence base does not yield a reliable cross-organisational rate for how often those workarounds survive after the underlying capability gap has actually been closed. [inference; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/]

The strongest supported response is therefore to treat temporary automation as a governed bridge, not as a standing substitute, by linking local automation to central review, registry fields, observable retirement triggers, and an explicit path back into core software delivery. [inference; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://hdl.handle.net/10125/108890; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html]

That conclusion does not rule out complementarity, because governed citizen-development programmes can also surface demand and process knowledge that later help core-system teams build the durable fix. [inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://hdl.handle.net/10125/108890; https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/]

Key Findings

Temporary automation displaces core software-delivery demand primarily by giving users a faster local alternative when central Information Technology delivery or sanctioned tools cannot meet workflow needs with adequate speed or fit. ([inference]; medium confidence; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai)
Automation adoption often scales ahead of enterprise strategy and Information Technology readiness, which supports the inference that local delivery can expand before the organisation closes the underlying capability gap in software. ([inference]; medium confidence; source: https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://aisel.aisnet.org/misqe/vol23/iss3/3/)
Post-rollout shadow Artificial Intelligence behaviour shows that sanctioned provision does not reliably eliminate workaround demand when users still perceive external tools as better, easier, or faster than approved enterprise options. ([fact]; medium confidence; source: https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai; https://davidamitchell.github.io/Research/research/2026-05-08-shadow-ai-behavioral-drivers-governance-effectiveness.html)
Robotic Process Automation and adjacent workaround programmes can create additive operating cost because bot support and upgrade effort persists while the cost of the underlying legacy system remains in place. ([fact]; medium confidence; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html)
Major automation platforms have built retirement features around inactivity, ownerlessness, approvals, and dependency checks, which supports the inference that persistence of unused or obsolete workarounds is common enough to require explicit lifecycle controls. ([inference]; medium confidence; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data)
The reviewed public evidence base does not yield a reliable cross-organisational persistence rate after the underlying capability gap has been closed, so the requested rate question remains unresolved rather than answered with confidence. ([inference]; medium confidence; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/)
The evidence most consistently points to central repositories, code review, role-based access control, separate environments, business-Information Technology collaboration, and leadership-backed programme design as the governance bundle most likely to protect core build capacity. ([inference]; medium confidence; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://hdl.handle.net/10125/108890)
The most defensible operating model is to treat temporary automation as a governed bridge with explicit end-of-life conditions and observable replacement signals, not as a standing substitute for fixing the underlying system. ([inference]; medium confidence; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html)

Evidence Map

Claim	Source	Confidence	Notes
[inference] Temporary automation diverts demand away from waiting for core software delivery when central or sanctioned options are too slow or inadequate.	https://aisel.aisnet.org/misqe/vol23/iss3/3/ ; https://aisel.aisnet.org/misqe/vol23/iss3/6/ ; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html ; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks ; https://www.ibm.com/think/topics/shadow-ai	medium	Behavioural displacement
[fact] Automation adoption can scale before enterprise strategy and Information Technology readiness catch up.	https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html ; https://aisel.aisnet.org/misqe/vol23/iss3/3/	high	Strategy lag
[fact] Sanctioned rollout does not reliably eliminate workaround demand if external tools still fit better.	https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks ; https://www.ibm.com/think/topics/shadow-ai ; https://davidamitchell.github.io/Research/research/2026-05-08-shadow-ai-behavioral-drivers-governance-effectiveness.html	medium	Post-rollout bypass
[fact] Workaround programmes can leave organisations paying both workaround and legacy-system costs.	https://community.pega.com/blog/when-it-time-retire-your-rpa-bots ; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html	medium	Additive cost stack
[inference] Major platforms productise retirement around inactivity, ownership, approvals, and dependency checks, which supports the view that persistence is common enough to require explicit lifecycle controls.	https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components ; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory ; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup ; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data	medium	Inference from control design
[inference] The reviewed public evidence base does not yield a reliable post-gap-closure persistence percentage.	https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components ; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup ; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data ; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/	medium	Quantitative gap remains
[inference] The evidence most consistently points to a governance bundle of technical controls, central collaboration, and leadership support as the pattern most likely to protect core build capacity.	https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/ ; https://aisel.aisnet.org/misqe/vol23/iss3/3/ ; https://aisel.aisnet.org/misqe/vol23/iss3/6/ ; https://hdl.handle.net/10125/108890	medium	Comparative conclusion
[inference] Temporary automation should be governed as a bridge with explicit retirement conditions.	https://community.pega.com/blog/when-it-time-retire-your-rpa-bots ; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components ; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory ; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html	medium	Operating-model conclusion

Assumptions

Assumption: Low-code apps, RPA bots, and AI-agent workarounds can be analysed together when they serve the same bridging role over an unresolved systems capability gap. Justification: The tooling differs, but the lifecycle logic, persistence risk, and retirement problem recur across all three classes. [assumption; source: https://davidamitchell.github.io/Research/research/2026-04-26-systems-capability-debt-agentic-ai-risk-synthesis.html; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html]
Assumption: Built-in lifecycle controls are a reasonable proxy for the operational significance of workaround persistence even when public retirement-rate datasets are absent. Justification: Platform vendors and bot operators usually expose such controls only for problems that appear repeatedly in live estates. [assumption; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots]

Analysis

The evidence shows a strong and repeated behavioural mechanism, not a complete financial ledger. [inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html]

Low-code studies, automation surveys, and shadow Artificial Intelligence reporting all point to the same sequence: users adopt local workarounds when central delivery or approved tools fail to meet immediate workflow needs, and those local successes can reduce the felt urgency of deeper remediation even when the underlying gap remains open. [inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://aisel.aisnet.org/misqe/vol23/iss3/6/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai]

The persistence side of the question is materially weaker on direct metrics, but strong enough on lifecycle design to reject the idea that workarounds self-retire once better systems exist. [inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data]

That makes the key trade-off speed versus durable capability. [inference; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://davidamitchell.github.io/Research/research/2026-05-16-agent-operational-cost-vs-gap-closure-cost.html]

A credible alternative interpretation is complementarity rather than displacement, because coordinated citizen-development programmes can surface demand, improve process understanding, and feed later core-system change. [inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://hdl.handle.net/10125/108890; https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/]

The evidence supports that alternative only when local automation remains inside business-Information Technology collaboration and lifecycle controls, which means complementarity is conditional rather than automatic. [inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://hdl.handle.net/10125/108890; https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/]

If organisations reward local speed without registry, review, and retirement design, temporary automation becomes a competing delivery lane that absorbs attention and leaves the underlying gap intact. [inference; source: https://digital.gov/2021/08/16/5-tips-for-implementing-citizen-development-in-your-rpa-program/; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://davidamitchell.github.io/Research/research/2026-05-16-decommission-trigger-design-for-do-mode-agents.html]

Risks, Gaps, and Uncertainties

The reviewed public sources used here do not yield a reliable cross-organisational persistence percentage after gap closure, so the rate question remains unresolved. [inference; source: https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components; https://learn.microsoft.com/en-us/power-platform/admin/automatic-environment-cleanup; https://docs.uipath.com/automation-hub/automation-cloud/latest/user-guide/deleting-data; https://research.universityofgalway.ie/en/publications/adoption-of-low-code-and-no-code-development-a-systematic-literat-6/]
The evidence for displacement is behavioural and organisational rather than financial-accounting grade, so the item supports a mechanism and direction of effect but not a precise crowd-out percentage. [inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks]
Vendor and platform sources can show lifecycle mechanics clearly, but they underreport estate-wide outcome metrics, so future work should seek longitudinal inventory data from live programmes. [inference; source: https://community.pega.com/blog/when-it-time-retire-your-rpa-bots; https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components]

Open Questions

Which organisations publish longitudinal inventory data that could reveal actual retirement rates for apps, flows, bots, and agents after replacement capabilities go live? [inference; source: https://learn.microsoft.com/en-us/power-platform/admin/power-platform-inventory; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components]
Which budget-governance mechanisms best prevent temporary automation success from delaying approval for deeper system remediation? [inference; source: https://www.deloitte.com/us/en/insights/topics/talent/intelligent-automation-2020-survey-results.html; https://community.pega.com/blog/when-it-time-retire-your-rpa-bots]
How different is the persistence pattern between centrally registered enterprise agents and consumer-style shadow Artificial Intelligence use? [inference; source: https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://www.ibm.com/think/topics/shadow-ai]

Output

Type: knowledge
Description: Public evidence supports a demand-diversion mechanism from temporary automation into local delivery lanes and shows strong persistence risk, but it does not publish a robust cross-organisational post-gap-closure persistence rate. [inference; source: https://aisel.aisnet.org/misqe/vol23/iss3/3/; https://www.ibm.com/think/insights/rising-ai-adoption-creating-shadow-risks; https://learn.microsoft.com/en-us/power-platform/guidance/coe/setup-archive-components]
Links: