Sixty-seven percent of the barriers to AI adoption are organizational, not individual. That number comes from Microsoft’s 2026 Work Trend Index, a survey of 20,000 people. It is the most useful sentence I have read about AI transformation this year, because it moves the problem from where everyone is looking to where it actually lives.
Most stalled AI programmes are diagnosed as a skills problem. The tools are bought, the licences are issued, the training is booked — and adoption still flatlines. The conclusion drawn in the steering committee is almost always the same: our people need more enablement. The data says otherwise. Two thirds of the time, the person is not the constraint. The structure around them is.
We strapped jet engines to horse-drawn carriages
Here is the image I keep coming back to. We strapped jet engines to horse-drawn carriages — then acted surprised when the carriage shook itself apart.
The engine works. Nobody seriously disputes that the models are capable. What fails is everything the engine was bolted onto: approval layers designed for a slower machine, unclear ownership, fragmented data, processes that assume a human bottleneck at every step. You can install a technology that operates in seconds inside a governance system that operates in weeks, and the governance system will win. It always wins. That is what a system is for.
This is why the training-first response is not just insufficient — it is quietly demoralising. You are telling capable people that the reason nothing moves is them.
Blocked agency
The report gives the failure a name I find precise: blocked agency. Capable people, trapped by permissions they cannot get, data they cannot reach, and processes nobody can explain. They can see the better way to work. They cannot get there from where they are standing.
Three further findings from the same research make the shape of it clear:
- 65% fear falling behind. The anxiety is real and it is widespread.
- 45% feel safer doing the work manually. Not slower — safer. That is a statement about psychological safety and accountability, not about skill.
- Only 13% are rewarded for reinventing how they work. This is the one that should stop a leadership team in its tracks.
Read those three together and the behaviour becomes entirely rational. People are frightened of being left behind, they are not protected when the machine gets it wrong, and almost none of them are recognised for changing anything. Under those conditions, doing it the old way by hand is the correct individual strategy. You have not got an adoption problem. You have got an incentive system doing exactly what it was designed to do.
Audit your systems, not your people
The path forward is not more tools, and it is not more training. Three moves matter more:
1. Move authority closer to the work
Every approval layer between a person and a decision is a place where an AI-speed process reverts to human speed. The question is not “can we automate the approval?” but “why does this decision need to travel at all?” Most approval layers are fossilised information scarcity — they exist because, once, the person at the edge genuinely could not see enough to decide well. That constraint has largely dissolved. The layer usually has not.
2. Remove obstacles for the people already innovating
In every organisation I have worked in, someone has already built the thing. They did it in a spreadsheet, or in a personal account, or at the weekend. They are not asking for training. They are asking for a permission, a data connection, or for someone senior to stop treating their initiative as a compliance risk. Find those people first. They are your fastest evidence and your cheapest change agents.
3. Match efficiency gains with meaning
If every hour saved is immediately re-filled with more of the same work, people learn quickly that efficiency is a trap. The organisations that get real adoption are explicit about what the saved time is for — better thinking, closer customer contact, actual recovery. Efficiency without a stated purpose reads as extraction, and people withdraw from it.
The question to sit with
If your AI rollout has stalled and you are being told it is a training issue, run one test before you approve the next enablement budget. Take a single stalled use case and trace it end to end: who had to approve what, how long each step waited, which data was unreachable, and who would have been blamed if the model had been wrong.
You will almost certainly find that the person at the centre of it did nothing wrong at all.
Audit your systems, not your people.
Source: Microsoft Work Trend Index 2026 (survey of 20,000 people).
📚 Start here: AI-Native Organization Design — the full research hub, with every article and video in one place.
More in this series
- $2M Revenue Per Employee: The Case for the Living Organization
- The Living Organization: Five Pillars of an AI-Native Operating Model
- Why Hierarchy Existed — and Why AI Just Removed the Reason
- Is HR Ready to Design Human and AI Agent Productivity Together?
- How Much Does a CEO Really Need to Understand About AI?
- The Five Steps of AI Adoption: From Gated to AI-Native
- Your AI Isn’t Failing — Your Specification Is
- Stop Measuring Tomorrow’s Talent With Yesterday’s Rulers
Watch the full series
This article accompanies a video from AI-Native Operating Model and Org Design — a series on how organisations actually absorb AI, and where they break.
- ▶️ Watch the full playlist — AI-Native Operating Model and Org Design
- 📺 Subscribe on YouTube — @BreezeDONG
Chunfeng “Breeze” Dong is an executive coach (ICF PCC, CPCC) and founder of Springbreeze Ventures, with twenty years in organisational development inside Fortune 100 companies — Roland Berger, Siemens, ABB and Roche. She writes on AI-native organisation design, human–agent governance and change leadership.
📘 The Living Organization · 📘 A Soulful Transition · 🔗 LinkedIn
