How to build a ‘safe-to-fail’ culture for IT teams — and why you should
IT leaders are telling teams to experiment with AI. But most are still punishing failure in performance reviews.
Why it matters
Safe-to-fail culture is a necessary but underdeployed practice for AI adoption in enterprises. Without explicit permission, clear boundaries, and protected experimentation time, IT workers will avoid trying new tools—leaving companies' AI investments stranded. The article collects practitioner advice on how to unblock this.
The key facts
13 to knowTypeform allocates $1,000 per employee for training, tools, and certifications
Typeform delivered 9-month project in 3 months with AI tools and time allocation
Randstad Digital framework: every pilot needs business owner, hypothesis, measurable outcome, time limit, explicit decision
Smartsheet uses pre-approved sandboxes with DLP and access controls to allow same-day tool trials without security review
Equifax leadership actively commends teams for shutting down projects that no longer make sense
Lenovo uses staged testing: sandbox with synthetic data → test environment → small pilot with limited access
Typeform allocates $1,000 per employee annually for training, tools, or certifications; gives dedicated workday time for experimentation
Smartsheet enables pre-approved sandboxes for citizen-development projects with DLP and access controls pre-configured to eliminate security review delays
Lenovo built a governed AI sandbox on enterprise AI OS; successful prototypes inherit security and compliance controls by default for production
Typeform's Research Flow team delivered a working prototype in 3 months (vs. 9-12 months without AI tools) using clear goals and bonus incentives
Safe-to-fail culture requires: disciplined pilots with business owner, hypothesis, measurable outcomes, time limits, and explicit go/no-go decision
Clear boundaries approach: document what tools/data employees can use, which systems are accessible, when IT/security involvement is required
Staging approach: synthetic data in sandbox → test environment → small pilot, with audit trails and reversal plans before production
Go to the source
Computerworldcomputerworld.com
Publisher excerpt: Companies continue to invest in AI, automation, developer tools, and other new technologies. But they may not get the results they expect if IT workers don’t have the time, resources, or freedom to learn how to use them. IT workers may be afraid to try new tools if they think a failed experiment…