Some organizations are turning off AI features users can access when the cost of operating them outweighs their value. In Applause’s 2026 survey, 44.1% of respondents said their organization had deactivated live AI features in the prior year for that reason. That is evidence of post-launch deactivation—not proof that 44.1% of AI features were retired, or that every affected feature had shipped exactly a year earlier.
What the numbers do—and don’t—say
Applause’s 2026 survey included 1,097 respondents working in software development, quality assurance, data science, AI research, and product management. In that group, 54.5% said their organization had released AI features, while 44.1% said their organization had deactivated live AI features in the prior year because operational costs outweighed user value. These are respondent-reported, organization-level findings; they are not a feature-by-feature retirement rate, nor a measure limited to engineering teams. Applause’s 2026 survey
As an Amazon Associate I earn from qualifying purchases.
Other figures about AI projects describe different stages and should not be added to the deactivation figure. S&P Global Market Intelligence reported in 2025 that the share of companies abandoning most AI initiatives before production rose from 17% to 42% year over year. Respondents also estimated that an average of 46% of projects were scrapped between proof of concept and broad adoption. Those findings concern pre-production attrition, not the removal of features already available to users. The survey covered 1,006 midlevel and senior IT and line-of-business professionals in North America and Europe. S&P Global Market Intelligence
Longer production duration is another, distinct measure. In Gartner’s 2025 survey of 432 respondents across the U.S., U.K., France, Germany, India, and Japan, 45% of leaders in high-AI-maturity organizations said their AI initiatives remained in production for at least three years, compared with 20% in low-maturity organizations. This is a comparison by reported organizational maturity, not a controlled demonstration that maturity causes longevity or a direct measure of how many features were retired. Gartner’s production-duration findings
#1 Best Overall
Why a launched AI feature can stop earning its place
The running cost outlasts the user benefit
A feature that attracts interest at launch may be expensive to serve or support once real usage begins. The Applause respondents who reported deactivation cited operational costs outweighing user value. That trade-off is not just the model’s inference bill: teams should also account for infrastructure, engineering maintenance, quality work, and support. S&P Global likewise identified cost as a concern in AI projects, but its project-abandonment findings are not evidence that production features were retired for the same reason. S&P Global on AI-project costs
Integration and data are harder than the demo
An AI capability has to fit the application’s workflows and data, not merely produce a plausible answer in a prototype. Gartner found that 77% of engineering leaders surveyed described building AI capabilities into applications as a significant or moderate pain point. The survey covered 400 U.S. and U.K. software engineering and application-development leaders and was conducted from October through December 2024. Gartner analyst Jim Scheibmeir said, “Even with business leaders focusing more on this technology and despite the growing hype, execution is not easy.” Gartner’s integration findings
Rank #2
Applause also described teams encountering internal workflows and data that were not ready for agent-based features. Its EVP of High Tech and AI, Chris Sheehan, said, “Right now, many teams are struggling with internal workflows or finding that their data isn’t in the proper format for an agent,”. The quotation’s comma reflects that the original sentence continues; the point is that data and workflow readiness can constrain a feature even after the model itself has been selected. Applause on AI quality and readiness
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Quality problems erode trust and adoption
Testing can reveal that a system does not handle context or user intent reliably enough for the workflow. Applause described cases in which teams found AI systems not ready after testing, including a chatbot delayed following tester feedback. A feature that users cannot trust may fail to produce the adoption needed to justify its costs.
Gartner’s Birgi Tamersoy put the relationship plainly: “Trust is one of the differentiators between success and failure for an AI or GenAI initiative.” She also said, “Building trust in AI and GenAI solutions fundamentally drives adoption, and since adoption is the first step in generating value, it significantly influences success.” These are Gartner analyst observations, not a quantified threshold for whether an individual feature should remain live. Gartner on trust and production duration
Value is difficult to demonstrate
If a team cannot connect a feature to a meaningful customer outcome, it is hard to distinguish useful automation from novelty. In a separate Gartner survey of 644 respondents in the U.S., Germany, and U.K., conducted in the fourth quarter of 2023, 49% cited difficulty estimating and demonstrating AI-project value as a primary adoption barrier. Gartner on AI-project value barriers
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to decide whether to retain, improve, or retire a feature
There is no universal retirement formula established by these surveys. A practical review should identify what is failing, whether it can be fixed, and whether the benefit justifies the full cost and risk of keeping the feature available.
- Measure user value. Define the user problem the feature is meant to solve, then look for evidence of recurring use and customer impact. Gartner reported that 63% of leaders in high-maturity organizations measured customer impact and conducted financial analysis of risk factors and ROI. That is a reported practice in this group, not proof that the practice alone causes better outcomes. Gartner on measurement in high-maturity organizations
- Calculate the total operating burden. Include inference and infrastructure costs alongside engineering upkeep, evaluation, incident response, and user support. Compare those costs with the value you can actually observe; do not treat a lower model bill as sufficient if quality or support work remains high.
- Check quality in real workflows. Examine whether the feature understands relevant context, behaves consistently, and fails safely. Use observed user feedback and evaluation results to identify whether the problem is fixable or inherent to the current implementation.
- Audit integration and data readiness. Confirm that the required data is available, suitable, and handled securely, and that the feature fits the workflow users follow. Gartner has identified data availability and quality, and security, among implementation barriers; S&P Global has also cited privacy and security challenges. Gartner’s reported implementation practices and barriers S&P Global’s reported AI-project challenges
- Choose a reversible path where practical. If a specific defect has a credible remedy, improve the feature and define what evidence would justify keeping it live. If value remains unproven, costs remain disproportionate, or risks cannot be managed, disable it and communicate the change clearly to affected users. The threshold should be set by the organization; the cited surveys do not provide a universal cutoff.
Why organizational maturity matters
Gartner’s duration comparison suggests that AI initiatives in higher-maturity organizations are more often reported as remaining in production for at least three years than those in lower-maturity organizations. Its findings on customer-impact measurement and financial analysis offer useful examples of practices associated with that group, but do not establish a causal recipe or guarantee that a particular feature will last.
DORA’s 2025 report drew on more than 100 hours of qualitative data and nearly 5,000 technology-professional survey responses. It describes AI in software development as an amplifier of organizational strengths and weaknesses: “The research reveals a critical truth: AI’s primary role in software development is that of an amplifier. It magnifies the strengths of high-performing organizations and the dysfunctions of struggling ones.” A feature’s fate can therefore depend on the surrounding engineering system—its data, workflows, evaluation, ownership, and ability to respond to user feedback—not only on the model. DORA’s 2025 report
What a responsible retirement looks like
Turning off a live feature is not the same as quietly abandoning a project before launch. The evidence supports reported deactivation, but it does not establish that teams concealed those decisions. If an organization disables a feature, it should make the change legible to users, explain what functionality is changing, and provide an alternative or next step when possible. Internally, preserve the evaluation results and cost or risk findings behind the decision so the team can learn from the launch rather than repeat it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
