AI isn’t failing because leaders lack vision. It’s failing because work hasn’t been redesigned.
I walked into the Microsoft Garage in January for our Work Design Collaborative meetup with a very specific goal. Reduce the pressure enterprise leaders are carrying from their CEOs and boards.
Not with reassurance. With something they could actually execute.
Every organisation in that room is being pushed to redesign for AI, fast. Yet they’re all blocked on the same question.
What exactly do we redesign first?
There are no credible enterprise benchmarks. No shared examples of AI changing work at scale. No agreed unit of progress.
Not because leaders aren’t trying.
But because almost no one has actually reinvented work yet.
Until tasks, workflows, and roles are explicit, AI has nothing to attach to. It floats above the organisation as tools, pilots, and intent. Never hardening into production reality.
That’s what leaves executives exposed in boardrooms and operators stuck in the middle.
AI adoption is not the problem. Accountability is.
AI is already in inboxes, workflows, and meetings. But most organisations are still studying work instead of changing it. Insight accumulates. Execution stalls.
What’s missing isn’t data or ambition.
It’s ownership.
And this is where I’ll be direct.
HR has to go first.
Not because HR owns people.
Because HR understands work.
If AI is going to scale responsibly, it has to start where work is designed and governed. Anything else becomes IT theatre or shadow AI.
If AI isn’t changing live workflows, it isn’t transformation.
That’s why we shared the Reinvention Flywheel, the operating model we use to take AI from experimentation into execution.
The Reinvention Flywheel
This isn’t transformation theatre.
It’s how work actually evolves.
Trying to AI everything is how teams lose credibility.
We only reinvent four categories of work:
If it doesn’t fit one of these, it’s noise.
This filter alone removes most AI ideas. That’s a feature, not a bug.
This wasn’t a discussion. We built a real workflow.
Every table was given real AI workflow options and one rule.
Pick something you could actually deploy.
The clear winner was an Offer Package Sense Check, an AI agent that provides instant guidance on compensation offers.
It won because:
We built it live in Microsoft Copilot Studio. End to end.
No someday. No slideware.
If you can’t prove impact, it doesn’t matter.
We measure AI by what changes in work.
Our core execution metrics
This is the language executives understand.
And it’s how reinvention gets funded.
Tools don’t change behaviour. Design does.
The strongest HR-led AI transformations follow the same pattern:
If you can’t point to what changed, it didn’t.
Why should HR lead AI work redesign?
Because HR understands how work is structured, governed, and absorbed into roles. AI cannot scale without that foundation.
Where should enterprises start with AI execution?
At the task level, inside real workflows, not at the tool or job title level.
How do boards measure AI progress?
By changes in work. Time freed, risk reduced, capacity created.
In 2026, most organisations will still be talking about AI.
A few will be quietly redesigning work.
The difference won’t be technology.
It will be ownership.
When HR leads through work design, AI stops being a side project and becomes a capability.
That’s when things finally move.