• AI STRATEGY · 14 MIN READ
AI is not a tool rollout. It is a workplace rewrite.
Most organisations have mistaken access for adoption – and adoption for transformation. The technology is ready. The question is whether the institution around it is.
By Suchetana Bauri · Published 8 September 2026
Conceptual Framework
Access → Adoption → Transformation
Access
Surface Layer
Adoption
Process Layer
Transformation
Structural Layer
We bought licences
People are using it
Work changes. Value becomes visible.
THE PRACTICAL TEST
Days 1-30
Find work worth changing
Days 31-60
Redesign before automating
Days 61-90
Launch, learn and publish the evidence
IN THIS ARTICLE
01. Why tool rollouts fail
02. The organisational readiness gap
03. What to measure instead
04. How to communicate change honestly
05. Governance people can actually use
06. A practical 90-day plan
• 01 / The Rollout Illusion
A tool can be deployed in a day. Trust cannot.
That distinction sounds obvious, yet it is routinely ignored. Organisations treat AI as if it were the latest software upgrade: procure it, configure it, train users, track log-ins, declare progress. This approach produces activity. It does not reliably produce value.
Kyndryl’s 2026 People Readiness Report, based on responses from 1,100 business and technology leaders across eight countries, captures the contradiction neatly. Seventy-seven per cent of organisations said they had scaled generative AI across multiple functions, but workforce readiness had moved in the other direction. The report argues that the central question is no longer whether a company can deploy AI, but whether it can prepare people and build enough trust to get value from it.
“AI is available to everyone’ is not an achievement. ‘People are using it’ is not an achievement either.”
– rollout versus value metrics
The better questions are less comfortable:
Which work has actually changed because of AI?
Which steps in a workflow no longer need to exist?
Where has decision-making become faster, better or safer?
What has been deliberately kept human?
Do employees know what they are allowed to do, expected to do and protected from doing?
If AI saves time, where does that time go?
If nobody can answer these questions, the organisation does not have an AI strategy. It has a collection of subscriptions and a growing sense of unease.
This is especially visible in knowledge work. A person uses an AI assistant to summarise meetings, draft a first-pass briefing, produce options for a campaign, clean a spreadsheet or generate a presentation outline. Individually, these are helpful gains. But the meeting still happens. The briefing still goes through three unnecessary approval layers. The campaign still begins with a vague brief. The spreadsheet is still passed across departments by email. The presentation still exists mostly to prove that work took place.
That is the rollout illusion: confusing accelerated output with meaningful change.
• 02 / The Real Bottleneck Is Organisational
The real bottleneck is organisational
The debate about AI often gets stuck at the level of the individual. Can people write better prompts? Are they sufficiently AI-literate? Will they embrace the tools or fear them?
These are reasonable questions, but they are not the whole story. Employees cannot prompt their way out of a broken operating model.
McKinsey’s 2026 research draws a useful line between personal readiness and organisational readiness. Personal readiness is whether employees feel able, supported and willing to use AI. Organisational readiness is whether the company is prepared to change workflows, leadership behaviour, roles, resource allocation and cultural norms around the technology.
The findings are revealing. Seventy per cent of respondents said they were personally ready to adopt and use AI, while only 27% of leaders believed their organisations were ready for the changes required by a more agentic future. Organisational readiness was also more strongly associated with enterprise value than individual readiness: 48% of the difference between organisations capturing value and those that were not, compared with 25% for personal readiness.
“People may be ready before their employer is. Training without work redesign asks employees to become more efficient inside a system that wastes their efficiency.”
– McKinsey 2026 Research Insights
It is like giving everyone in a company a faster car, then keeping the same traffic jam.
The organisations getting somewhere are not simply asking, ‘How can AI help employees do their current jobs faster?’ They are asking a much more useful question:
‘If we were designing this work today, with AI available, would we do it this way at all?’
That question is threatening because it exposes organisational habits that have little to do with customer value and a great deal to do with hierarchy, caution and inherited bureaucracy.
• 03 / Metric Redesign
Stop measuring log-ins
A high adoption rate can be the most misleading metric in the room. It tells you that people opened a tool. It does not tell you whether they used it well, whether the use was safe, whether it improved outcomes or whether the benefit has travelled beyond the individual user.
Rather than measuring use in isolation, measure changed work. A good AI programme should be able to identify a small number of priority workflows and show what has improved.
Weak Measure
Better Measure
Number of AI licences issued
Time from source material to approved briefing
Number of prompts submitted
Reduction in duplicated research or rewriting
Monthly active users
Percentage of priority workflows redesigned
Training attendance
Employee confidence in safe and useful use
AI-generated outputs produced
Quality, accuracy and stakeholder usefulness of outputs
Estimated hours saved
Hours deliberately redirected to higher-value work
The final row matters most. Time saved is only valuable if someone decides what happens next. Without an answer, ‘productivity’ becomes another way of saying: do the same job, but faster, with fewer people and less room to think.
• 04 / Trust & Communication
People do not need reassurance. They need honesty.
The worst AI communication is a polished message telling employees not to worry. People are not naive. They can see that AI may change tasks, responsibilities, career paths, team structures and, in some cases, headcount.
5 Pillars of Credible AI Communication
4.1. Explain the purpose in human terms
Do not begin with ‘We are embracing innovation.’ Say what problem the organisation is trying to solve. Employees can judge whether a change is sensible when they understand the work behind it.
4.2. Name the trade-offs
AI may make some tasks easier while creating new checking, oversight and coordination work. Name these tensions. Do not hide them beneath the word ‘opportunity’.
4.3. State what is decided and what is still open
Be specific. Tell people where they have real influence, and where hard decisions have already been set.
4.4. Give managers a script – and permission to say ‘I don’t know’
The most trustworthy line in an AI transformation may be: ‘I do not know the answer yet. Here is what we know, what we are deciding, and when I will update you.’
4.5. Close the loop
If employees raise concerns, show what happened next. Publish changes. Explain decisions. Listening without response is just another form of broadcast.
• 05 / Governance & Agency
Governance should help people move
A policy that says ‘use AI responsibly’ is not governance. It is a motivational poster with legal exposure. A workable governance model should answer the questions employees actually ask at 4.45pm on a Friday:
Can I put this document into the tool?
Can I use AI to write a first draft for this client or stakeholder?
What must I check before I send an AI-assisted output?
When must I disclose that AI was used?
Which decisions can AI support but not make?
Who do I contact if the tool produces an inaccurate, biased or unsafe result?
What happens if I make a good-faith mistake while following the guidance?
• 06 / Action Roadmap
A better 90-day plan
You do not need to redesign the entire organisation before doing anything. But you do need to stop calling random experimentation a strategy.
1-30
6.1. Find the work worth changing
Choose three to five workflows where there is a clear business or service problem. Include the people who actually do the work. If a process disappeared tomorrow and nobody noticed, do not automate it. Stop doing it.
31-60
6.2. Redesign before you automate
For each selected workflow, decide what should be removed, simplified, or automated, and where human editors focus on critical judgement. Build the new workflow with the people who will live inside it.
61 – 90
6.3. Launch, learn and show the evidence
Run a limited implementation with clear guardrails and proper feedback loop. Compare actual test data against pre-agreed operational thresholds, then make an objective decision to scale, redesign, or stop.
• 07 / Strategic Choice
The choice is not human or AI
The tired framing of ‘human versus machine’ is a distraction. The real choice is between organisations that use AI to deepen old habits and organisations that use it to question those habits.
AI will not automatically make work better. It can just as easily make bad work faster, surveillance easier, expectations more punishing and decisions less accountable. But it can also remove pointless friction, and create space for better human judgement.
employees, like audiences, can tell when they are being sold a story that does not match the evidence.
Footnotes & References
[1] Kyndryl, ‘People Readiness Report 2026’ – kyndryl.com
[2] McKinsey, ‘From adoption to impact: Three horizons of AI transformation’ – mckinsey.com
AI GOVERNANCE · STRATEGY CONSULTING
Your AI rollout is not an adoption strategy.
I help organisations identify the workflows worth redesigning, establish practical governance and build communication that earns trust.
Not ready for a conversation? Start with an AI readiness audit.