You Deployed the Technology. Now You Have a People Problem You Did Not Plan For.
Last week, we looked at why AI programmes are not delivering: the strategy gap, the operating model that nobody redesigned, the 90% of senior executives reporting no measurable productivity improvement. If you missed it, the short version is that most companies are running new tools on old structures and wondering why the numbers are not moving.
This week, the story moves one step further.
Because for the organisations that did deploy AI at scale, something else is happening. Something that the strategy consultants did not put in the slide deck. The people are struggling. And in some cases, the people are gone.
The Deployment Gap
Kyndryl’s 2026 People Readiness Report, published this week and based on a global survey of 1,100 senior business and technology leaders across eight countries, has one figure that is hard to ignore.
AI is now deployed in 57% of enterprises. Yet only 11% of those organisations have achieved their top two AI objectives. Not 50%. Not 30%. Eleven per cent.
But the more revealing finding is what is happening to the workforce confidence of the people living inside those deployments. Only 23% of business leaders now say their workforce is fully prepared for AI. That figure is down six points from 2025. Deployment is accelerating. Readiness is going in the opposite direction. Only 19% of workers feel confident using AI tools. Only 33% of organisations have established clear policies defining which decisions AI can and cannot make.
The organisations that are actually converting AI investment into results, Kyndryl calls them Pacesetters, represent roughly 9% of those surveyed. They share three behaviours: they redesign roles around AI rather than layering it onto unchanged job structures; they invest in structured change management so employees understand the new operating model; and they prepare the workforce before scaling deployment, rather than afterwards. They are 1.5 times more likely to achieve AI-related revenue growth than their peers.
What sets them apart is how seriously they take the people side of AI.
The Cost of Moving Too Fast
Ford rehired 350 veteran engineers this week. The story, reported by HR Executive on 24 July, is a precise case study in what happens when AI restructuring uses what Peter D. Banko, president and CEO of Baystate Health, calls the ‘easy button’: targeting high-salary line items to cut costs rather than making deliberate decisions about which roles genuinely add value and which tasks are genuinely automatable.
Ford’s automated inspection systems missed design and quality defects that experienced staff would have caught. Charles Poon, Ford’s vice president of vehicle hardware engineering, said the company had let go of its most experienced people before their knowledge could be used to train the AI systems that were supposed to replace them. The institutional knowledge walked out the door before it could be captured. The AI was left working from an incomplete picture.
Daniela Seabrook, CHRO of the Adecco Group, whose research surveyed 2,000 C-suite executives across 13 countries, draws the lesson plainly:
‘The biggest risk for organisations today is not moving too slowly on AI; it is moving too quickly without a people strategy in place.’
Her data found that only 36% of leaders say their talent strategy clearly demonstrates how AI creates opportunity for employees, and just 39% are involving employees directly in redesigning their own jobs.
The Damage You Cannot See on a Headcount Report
The rehiring wave is visible. The damage beneath it is harder to measure but more consequential.
Research published in Harvard Business Review this week, by academics from ESSEC, INSEAD, and Nova SBE, documents something that most leaders have not yet factored into their AI strategy: the use of AI is actively eroding employees’ capacity for critical thinking.
Workers accept AI-generated outputs without interrogating them. A 2026 Wharton study on AI as a third cognitive system found that people using AI are disinclined to question or research what it produces. The cognitive muscle, when not used, weakens.
The real problem lies in how AI has been introduced. Many organisations treated it as a way to remove human effort, without changing how people are managed, trained or expected to think.
They are now finding that some employees have become less confident in their own judgement and less likely to question what the technology gives them. The efficiency gains are easy to measure. The loss of independent thinking is much harder to see, but may prove far more costly.
What the Pacesetters Do Differently
The AI-native companies highlighted by the Wall Street Journal this week, drawing on research from Harvard and INSEAD, show a different way of doing things. They tend to have leaner structures and fewer layers of management, but they have not got there simply by cutting jobs and replacing people with bots.
They have designed the work differently from the beginning.
At OffDeal, an AI-powered investment bank, engineers and bankers work side by side. Before writing any code, the engineers spent two weeks shadowing the bankers to understand how the work was actually done. The systems they built support human judgement rather than trying to replace it.
That is where the difference lies. The companies now rehiring, worrying about falling confidence and finding that people are becoming too reliant on AI are often the ones that saw it mainly as a way to cut costs. The companies getting better results treated it as a question of how the work itself should change.
Three TALENT Questions for Your Leadership Team This Week
1) When your company made AI-related workforce decisions in the past 18 months, did you document which tasks were genuinely automatable and what would happen to the rest of the role if only part of it was automated? If not, you may be building a rehiring case you have not yet recognised.
2) Do your managers know how to design work so that AI strengthens employee reasoning rather than replacing it? If the answer is no, the capability erosion the HBR research describes is already underway in your organisation.
3) Who in your organisation owns the question of workforce readiness, not AI adoption, but the confidence and capability of the people working alongside the systems you have deployed? If that question does not have a clear owner, the Kyndryl data suggests you are in the 89% that are not hitting their objectives
You Just Hired a Million Bad Employees
The transition from the C-Suite Radar to the talent implications is stark. If the management debt has come due, the people holding the bill are the employees tasked with making these systems work.
George Sivulka, the founder of enterprise AI startup Hebbia, published an essay this week that captures the current reality perfectly. His argument is blunt: companies raced to deploy AI workforces without building any of the management infrastructure to run them. The result? ‘You just hired a million bad employees.’
The models are rarely the main problem. The bigger issue is that very few people inside an organisation know how to brief and manage them properly.
Sivulka estimates that only one employee in 100 can describe a task clearly enough for an AI agent to carry it out well. He calls this ‘context engineering’, and argues that it is one of the scarcest skills in the market.
Without it, agents can get stuck in loops, repeatedly trying to correct poor instructions, using up tokens without producing anything useful.
The Rise of Context Hoarding
This lack of management capability is creating a toxic secondary effect: context hoarding. Employees are realising that their ability to provide context to AI systems is their last remaining form of job security. If they teach the AI their ‘secret sauce’, they make themselves redundant. So they are actively withholding their institutional knowledge.
This is a massive political and cultural problem that no governance framework can solve.
When employees do not trust the organisation’s AI strategy, they will sabotage it through ‘passive resistance’ (i.e. avoiding the tools, using them badly or simply carrying on as before).
This creates a political and cultural problem that governance alone will not fix.
Handling this well comes down to creating the right incentives, involving people properly and making AI feel like something employees can work with, rather than something they are being asked to compete against.
This piece first appeared in Talent Pools Nexus, our weekly read for leaders in research, insights and consulting. Subscribe on LinkedIn.
