84% of CIOs Have Canceled an AI Project Over Legacy Systems: What to Fix First

If your AI pilot stalled, the model may not be the problem. A new survey of 945 technology leaders says the old systems underneath it are.

On September 29, GFT Technologies published a global survey run by Wakefield Research. It asked CIOs and CTOs at companies with at least $500 million in annual revenue, across 19 countries, about AI and legacy infrastructure. This post covers what they found, how to read it, and what to do first if you are planning an AI digital transformation.

What Did the GFT Survey Find?

The survey ran from August 11 to 31, 2026. These are GFT’s published figures:

• 84% say limits in their legacy systems have caused their organization to cancel an AI pilot or project.
• 93% believe that running AI on old infrastructure without modernizing will eventually trigger an enterprise-wide security crisis.
• 89% worry that global AI investment is growing faster than the business value it can deliver.
• 99% say possible government restrictions on AI access make it more important not to depend on a single AI provider.
• 20% say the other executives and board members fully understand the security risks of running AI on legacy systems.

GFT reports a margin of error of plus or minus 3.2 percentage points at the global level. The full question wording and the breakdown by country are in GFT’s report, not in the press release.

How Should You Read These Numbers?

Two cautions. First, GFT sells digital transformation and modernization services, so it has a stake in this finding. Second, this is a survey of opinions and self-reported experience at large companies. It does not measure how many AI projects fail overall.

Still, the direction matches other reporting. CIO Dive covered the same study and noted that nearly half of respondents have started modernizing legacy systems, about a quarter feel behind, and only 15% say they are nearly or fully done. CIO Dive also cites a July Google report in which only 17% of IT leaders were confident their stack could support mission-critical AI agents.

GFT’s U.S. CEO, Rishi Chohan, told CIO Dive that legacy bottlenecks can often be managed for small productivity gains, but become a real barrier as AI connects to more of the business, as with agentic AI.

Why Do Legacy Systems Stop AI Projects?

GFT’s release does not list the technical causes, so this part is our view from project work, not the survey’s. The common blockers are:

• Data is locked in old systems. AI needs access to current data. If it sits in a system with no API, the pilot cannot reach it.
• No safe way to connect. Agents that act on your systems need controlled, logged access. Old systems often only offer shared logins.
• Slow release cycles. AI features need frequent changes. A quarterly release train does not fit.
• Security gaps. Connecting AI to an unpatched system widens the attack surface.

What Should You Modernize First?

You do not need to rewrite everything. CIO Dive quotes Chohan on the same point: define what you want AI to do for the business, find the systems that block it, and modernize those. We agree, and here is how we would sequence it:

1. Pick one business process. Choose a workflow where AI would save real time, such as support triage, document processing or reporting.
2. Map the systems it touches. List where the data lives and how each system can be reached.
3. Fix the connection first. Add an API layer or integration in front of the legacy system instead of replacing it.
4. Set access and logging rules. Decide what the AI may read and change, and record every action.
5. Run a small pilot, then expand. Move to the next process only when the first one is stable.
6. Avoid single-provider lock-in. Keep the AI model behind a thin layer so you can switch providers.

How Incresco Helps

Incresco does AI digital transformation. Our AI transformation work covers the pieces above:

• AI strategy and consulting: roadmaps, readiness assessments and the business case, so you pick the right process first.
• AI integration: connecting AI to your existing systems through workflow automation, API integration and legacy system modernisation.
• Custom AI development: AI agents, copilots and intelligent automation built around your workflows.

If an AI pilot of yours is stuck on an old system, or you want to check readiness before you start one, talk to our team.