The News
KPMG’s latest CEO Outlook survey finds that 86% of U.S. CEOs are actively redesigning roles and career paths for human-AI collaboration, while 53% are planning workforce reductions in some areas. Despite this broad restructuring activity, only 9% of those same CEOs describe AI agents as having been transformational to date. Laureen Knudsen, former Chief Transformation Officer at Broadcom and author of Becoming Essential, is raising a direct caution: organizations that automate without first identifying where human judgment is irreplaceable risk operational disruption and damaged customer relationships.
Analyst Take
The KPMG data captures a tension that’s becoming one of the defining strategic questions of 2025: how do you restructure for AI when the technology isn’t yet delivering the transformation you’re betting on? Nine percent calling AI agents transformational is not a rounding error. It’s a signal that most enterprises are mid-renovation, having torn out walls before deciding what the new floor plan should look like.
The Efficiency Narrative Has Outrun the Evidence
CEOs are reorganizing job functions at scale based on a productivity thesis that, by their own admission, has not yet materialized. That’s an unusual position. Knudsen’s warning about “serious operational problems and customer relationship damage” isn’t abstract caution — it’s a pattern she observed repeatedly across three decades of Fortune 1000 transformation work, including six years at Broadcom. The risk she’s naming is specific: companies are automating workflows without a disciplined analysis of which tasks require human judgment, institutional context, or ethical accountability. When those decisions go wrong, they tend to go wrong publicly.
For ITDMs, this is a board-level risk question dressed in an IT costume. The automation decisions being made now, at the role and workflow level, will determine which capabilities an organization retains in-house and which it effectively offloads to systems it doesn’t fully control. Getting that calculus wrong is recoverable in a pilot; it’s far more costly at enterprise scale.
What Developers Are Actually Experiencing
The gap between CEO ambition and ground-level transformation is consistent with what ECI Research’s survey data shows in adjacent sectors. According to the ECI Research Google GovTech Survey, 49.6% of respondents estimated that 26% to 50% of their organization’s code will be assisted or generated by AI within the next 12 months, which is a meaningful adoption horizon. Despite this, only 13.3% reported significant acceleration of greater than 30% speedup from AI-assisted coding tools already in use. The productivity gains are real but incremental, not the step-change that restructuring decisions at the CEO level seem to be pricing in.
That mismatch matters architecturally. Developers adopting AI tools in practice are finding them most useful for well-bounded tasks: generating boilerplate, writing tests, summarizing documentation. The autonomous, multi-step reasoning that would make an AI agent genuinely transformational in a complex business workflow is still the exception. Organizations that design human-AI collaboration models around the best-case ceiling of current AI capability, rather than its demonstrated floor, are building on unstable ground.
Where Knudsen’s Framework Intersects With Procurement Reality
One dimension of this problem that rarely surfaces in CEO-level surveys is the structural friction that slows human-AI collaboration models from even reaching production. ECI Research’s Google GovTech Survey found that 31.8% of respondents cited FedRAMP/compliance approval friction as the single largest blocker preventing widespread AI adoption in their developer workflows. In regulated environments, the question isn’t just “where does human judgment matter?” — it’s “which AI tools are even authorized to operate here, and on what timeline?”
That compliance overhead changes the transformation calculus significantly. A CEO redesigning roles for AI collaboration needs to account for the fact that in many enterprise and government contexts, the AI tools doing the collaborating may be months or years away from authorization. Knudsen’s point about identifying where human judgment is essential applies equally well before and after those tools arrive.
Looking Ahead
The 9% transformational figure from KPMG should be treated as a starting point, not a ceiling. As AI agents mature and more enterprise workflows are instrumented for automation, that number will climb. The organizations that will extract the most value from that shift are those doing the unglamorous work now: mapping decision points, documenting where institutional knowledge lives, and building human-AI handoff protocols before they’re under pressure to deploy them quickly.
Knudsen’s framework, and the broader discipline she represents, will gain relevance as CEOs move from the redesign phase to the accountability phase. When AI-assisted workflows produce errors or customer-facing failures, the question of who owned the decision and where the human was in the loop will not be abstract. Expect the next wave of enterprise AI governance investment to focus less on model selection and more on workflow architecture: who decides, who reviews, and who is responsible when things go sideways.
Stay Ahead of Application Development Trends
Get weekly analyst insights, research notes, event coverage, and AppDevANGLE updates delivered directly to your inbox.
Subscribe for Weekly Insights
Join technology leaders, practitioners, and GTM teams following the trends shaping modern software delivery.
Looking for deeper research access?
Explore ECI Research reports, survey insights, and market analysis through the ECI Research Portal.
