The story repeats itself across boardrooms and IT departments every quarter. A company rolls out a new AI capability – Copilot, an intelligent automation platform, an AI-powered analytics suite – with genuine fanfare. Early adopters evangelise. Productivity numbers tick up. Executives are pleased. Then, six months later, usage plateaus. The early adopters are still using it. Everyone else has drifted back to the way things were.
This is not a technology problem. It is a culture problem.
BCG’s 2026 AI at Work survey found that 74% of frontline white-collar employees now use AI regularly – up more than 20 percentage points over two years. But the same research revealed that 66% of those users receive little or no guidance on how to reinvest the time AI frees up. Adoption without direction is adoption without staying power.
The organisations that sustain and grow AI adoption after the initial rollout have built something the others have not: an internal operating model that treats human behaviour, not tool deployment, as the primary lever.
The Transformation Paradox Is Real
Microsoft’s 2026 Work Trend Index – developed in partnership with Harvard Business School – named it precisely: the Transformation Paradox. Employees feel genuine urgency around AI; 65% of AI users surveyed fear falling behind if they do not adapt quickly. Yet 45% say it feels safer to focus on current goals than to redesign work with AI. And only 13% say their organisations actually reward them for reinventing work with AI when short-term results fall short.
The paradox is structural. Metrics, incentives, and management norms still reflect the old way of working, even as leadership signals the new way. Until organisations close that gap, enthusiasm after a rollout is virtually guaranteed to fade.
Feedback Loops That Actually Close
The most overlooked failure point in AI programmes is not the launch – it is what happens in the weeks and months that follow. Most organisations collect feedback in theory. Few build feedback loops that actually close.
A closing feedback loop has four properties: it is continuous, not periodic; it is visible to the people who acted on it; it routes to someone with authority to make changes; and it produces a response fast enough that employees believe it matters.
In 2026, progressive leaders are using AI agents themselves to synthesise sentiment from collaboration channels, identifying bottleneck trends before they become disengagement. That early signal is only valuable if someone is assigned to act on it. Assigning a dedicated Adoption Lead – a named individual who owns the resolution of friction points – is increasingly the differentiator between programmes that adapt and those that stall.
The principle from change management is simple: feedback dies when no one owns the resolution.
What to Measure
Adoption metrics need to reflect depth, not just frequency. Organisations that sustain momentum track:
- Behaviour change - are people completing tasks differently, not just using the tool occasionally?
- Time reinvestment - where is the time AI frees up going? Is it redirected to higher-value work?
- Confidence signals - are employees self-reporting growing comfort, or stagnating uncertainty?
Deloitte’s 2026 State of AI in the Enterprise report found that only 30% of organisations are providing performance-based incentives for leveraging AI. The gap between deployment and reward structures is widening. Closing it starts with measuring the right things.
Recognition Systems That Reinforce the Right Behaviour
BCG’s 10-20-70 principle is now well established: 10% of AI value comes from algorithms, 20% from data and technology, and 70% from people, processes, and cultural transformation. Most organisations still invert this ratio in practice – spending the majority on tools while underinvesting in the human side.
Recognition is a concrete, low-cost mechanism that reshapes culture. But it has to recognise the right things. Rewarding volume of AI use is the wrong signal. Rewarding experimentation, knowledge sharing, and workflow redesign – even when the short-term result is imperfect – is the right one.
Research published alongside Microsoft’s 2026 Work Trend Index, drawing on a separate study of 1,800 workers globally, found that managers who create psychological safety around AI experimentation produce employees with up to 20 points higher AI readiness, and those employees are 1.4 times more likely to be high-frequency AI users. Recognition from a manager signals what is safe to try. That signal travels further than any internal communications campaign.
Publicly naming contributors in forums, weaving AI impact stories into performance reviews, and giving innovators face time with senior leaders are all low-cost mechanisms with measurable impact. They cost less than another round of tool licences and deliver more durable results.
The Champion Network Flywheel
The most scalable cultural mechanism for sustained AI adoption is the champion network – and it is gaining serious traction in 2026. Companies including Citi and PwC have used peer champion models to scale AI adoption across thousands of employees. Internal research from Worklytics suggests that 69% of employees cite colleagues as their primary source of AI skills learning, ahead of any formal training programme.
When champion networks function effectively, they create a flywheel: champions develop deep expertise, share it with peers, adoption climbs, early wins surface, leadership recognises the champions, more employees volunteer, and the network grows. The flywheel becomes self-sustaining.
Building a champion network that lasts requires three things that many programmes skip:
- Protected time - champions cannot champion if the role is piled on top of a full workload with no adjustment
- A clear mandate - champions need to know what decisions they can make, what problems they should escalate, and what success looks like
- Visible reward - public recognition, notes in performance reviews, and access to senior leaders are more effective than cash, and far cheaper
CSO Online put it plainly in 2026: “Sustained, distributed adoption doesn’t come from tool access. It comes from embedding AI capability inside how the organisation works.”
What Leaders Need to Do Now
For CIOs and CTOs reading this: the technology question is largely settled in your organisation. The human question is not. The InfoQ Culture and Methods Trends Report for 2026 was direct – the technology questions are increasingly settled, and the human questions are increasingly urgent.
The difference between organisations where AI sustains and grows versus those where enthusiasm fades comes down to three internal commitments:
- Build feedback loops with real owners and real response times
- Align recognition and incentives to the behaviours you want, not the outputs you measure today
- Invest in a champion network as infrastructure, not a nice-to-have
None of these require a new budget line. They require intention, accountability, and the willingness to treat culture as the product.
If your organisation is past the initial rollout and wondering why momentum has slowed, the answer is rarely the AI. It is the system around the AI.
That system starts with people readiness. In You Bought the AI. Now What? Why People Readiness Is the Real ROI Problem, we explore why deploying the technology is only the beginning – and why organisations need to prepare their people to change how work gets done.
From there, leaders need to understand where capability stands today. A Practical Framework for Assessing AI Literacy Across Your Workforce provides a way to identify gaps in AI knowledge, confidence and capability, while Building Organisation-Wide AI Literacy From the Ground Up looks at how to turn that understanding into broader workforce capability.
But awareness and literacy only create value when people start applying what they know. How Structured AI Experimentation Turns Passive Awareness Into Daily Practice explores how organisations can give employees the opportunity and confidence to experiment with AI in real work, moving from knowing about AI to using it as part of everyday workflows.
That experimentation also needs to reflect the realities of different jobs. Why AI Skill-Building Must Be Built Around Roles, Not Rolled Out Across Them examines why one-size-fits-all AI training falls short and how role-based capability building can make AI learning more relevant, practical and valuable.
And as AI becomes embedded in everyday work, responsible use cannot remain a policy employees read once and forget. How to Build Responsible AI Habits Into Everyday Work looks at how organisations can translate AI governance into practical behaviours and decisions employees can apply every day.
Together, these articles form A Leader’s Guide to Building an AI-Ready Organisation moving from people readiness and AI literacy through role-based skill-building and structured experimentation to responsible everyday use and, ultimately, a culture that sustains those behaviours.
That is where lasting AI value is created.
AI Momentum helps organisations turn these principles into practice. We work with leaders to understand workforce readiness, build role-relevant AI capability, enable responsible experimentation and create the feedback loops, recognition systems and champion networks that keep adoption moving.
The goal is not simply to get more people using AI. It is to build an organisation where people know when, where and how to use AI effectively and responsibly – and keep improving as the technology evolves.
