The AI Tightrope: What’s Really Keeping CTOs up at Night
By: Rahul Kumar, Regional Director, Experis Europe
Last month I sat in a room with a group of senior technology leaders. The topic was AI strategy. Every single person had a slide deck. Every slide deck had a roadmap. And somewhere between the third presenter and the fourth coffee, one CTO leaned over to me and quietly said: “Honestly, I have no idea how we’re actually going to deliver any of this.”
That one sentence says more about the state of enterprise AI than most research reports combined.
The pressure to “do AI” has never been higher. Boards are demanding strategies. CEOs are pointing at competitors. Vendors are promising transformation in 90 days. And CTOs are sitting in the middle of it all − trying to build something real, on infrastructure that wasn’t designed for it, with data that’s a mess, and a leadership team that’s only just starting to understand what they’ve signed up for.
Here’s my honest read on the five things that are actually making this hard.
The infrastructure isn’t ready − and everyone knows it
Two thirds of CTOs openly acknowledge their networks can’t fully support GenAI workloads. That’s not a fringe concern; that’s the majority experience. And rushing past it doesn’t help − 76% of CTOs have said that moving too fast on GenAI integration will create serious long-term infrastructure problems. The urgency to act is colliding head-on with the readiness to deliver.
The data foundation is broken
AI doesn’t fix bad data. It amplifies it. Yet nearly 70% of IT leaders describe their data as siloed, and the vast majority of organisations hit significant data quality problems the moment they try to build anything serious with GenAI. This isn’t a new problem − but AI makes it impossible to ignore any longer.
The C-suite doesn’t speak AI fluently enough
This is the one nobody wants to say out loud. According to the IBM CEO Study 2025, only 25% of AI initiatives have delivered the ROI they promised − and just 16% have scaled enterprise-wide. Part of the reason? Leaders are sponsoring programmes they don’t fully understand, setting expectations that the technology can’t currently meet, and losing confidence when reality doesn’t match the pitch.
At Experis, we see this constantly in our research. The IT World of Work 2025 Outlook found that 76% of IT employers worldwide are struggling to find the skills they need − and AI tops the list of critical shortages. That gap doesn’t just affect technical teams. It reaches into the boardroom too.
Shadow AI is already in your organisation
Right now, while you’re reading this, your employees are using AI tools you don’t know about. Feeding them documents you’d rather they didn’t. According to Menlo Security’s 2025 State of Shadow AI Report, 57% of employees are inputting sensitive data into free-tier AI tools. The risk isn’t theoretical. It’s operational.
Most AI projects never leave the pilot phase
Gartner predicted that 30% of GenAI projects would be abandoned after proof-of-concept by the end of 2025. The pattern is everywhere: brilliant demos that go nowhere, enthusiastic pilots that quietly die, and portfolios of “in progress” AI projects that somehow never reach production. McKinsey’s research confirms it − nearly two thirds of organisations are still mostly experimenting, not scaling. Only 6% qualify as genuine AI high performers.
Here’s the counter-intuitive point I’d add to all of this: the biggest risk facing most enterprise AI programmes right now isn’t too little ambition. It’s too much of it, applied too fast, without the foundations to support it.
I’ve watched organisations kick off fifteen AI initiatives simultaneously, each backed by genuine enthusiasm and real budget, and deliver meaningful results from none of them. The ones that consistently outperform have a different instinct − they do fewer things, properly. One or two use cases, owned end to end, taken all the way through to real business impact. Then they build from there.
The organisations delivering results treat AI as a business transformation programme, not a technology deployment. They fix the data before they build the model. They invest in leadership fluency before they approve the roadmap. They define what success looks like before the pilot starts.
And they’re honest − internally − about what they’re not yet ready for. That honesty is rarer than it should be. But it’s the starting point for everything that actually works.
I’ll be exploring each of these five challenges in depth over the coming weeks — infrastructure, data, talent, governance and scaling.
And if you’re working through any of these challenges right now − whether it’s building AI talent pipelines, finding the right technical skills, or thinking through workforce strategy for an AI-driven organisation − it’s exactly what we do at Experis. Reach out to us to find out how we can help. The organisations that will lead in AI aren’t the ones that started first. They’re the ones that started right. It’s still not too late to be one of them.
