AI adoption fails in predictable ways. Pilots multiply. Ownership blurs. Value gets promised but never measured. Risk gets discussed but never decided.
The Six Lenses are a way for leaders to hold the whole picture at once, and to know which conversation to have, with whom, in what order.
You don't need to be expert in all six. You need to know that all six exist, and who in your organization owns each one.
Where AI should help first, and why.
The first question is never "what can AI do?" It's "where does this business need the most help, and is AI the right answer there?"
Most organizations get this backward. They start with the technology and look for a problem to apply it to. The leaders who get AI right start with the business and look for the highest-value work that AI is actually good at.
A useful filter: AI tends to create real value where the work is high volume, has a clear right answer, generates proprietary data, and has a verifiable outcome. It struggles where the work is ambiguous, deeply relational, or one-of-a-kind.
Three questions every leader should be able to answer:
- What are the top five business outcomes we're trying to move this year?
- For each one, is there a specific workflow where AI would create measurable lift?
- How will we know, in dollars, hours, or quality, whether it worked?
If you can't answer those, you don't have an AI strategy. You have an AI wish list.
What are the top five business outcomes we're trying to move this year, and where would AI create measurable lift?
How humans and AI share the work.
The biggest mistake in AI adoption is dropping AI into existing workflows and hoping productivity follows. It rarely does. The work itself has to be redesigned around what AI does well and what humans do better.
A simple way to think about it:
- Automate the work that is high-volume, rule-based, and low-empathy.
- Augment the work that requires judgment, where AI gives the human a better starting point.
- Preserve for humans the work where empathy, ethics, or relationship value cannot be replaced.
The unit of design is the task, not the role. A single role contains tasks that belong in all three categories. Mapping work at the task level is what separates real transformation from surface-level pilots.
For each major workflow, which tasks should AI own, which should AI assist on, and which should never leave human hands?
What gets decided centrally, what gets decided locally.
In a matrix organization, governance is where AI adoption either accelerates or stalls. Too little, and you get duplication, risk exposure, and fragmented data. Too much, and you smother the experimentation that creates value.
The principle is simple: Centralize what protects the enterprise. Decentralize what creates value.
Centralize: risk frameworks, model approval, data standards, vendor selection, security, ethics review, the inventory of what AI is in use across the company.
Decentralize: use case identification, workflow redesign, prioritization within a function, the rhythm of testing and iteration.
The job of central governance is not to approve every idea. It's to set the guardrails so the business units can move fast inside them.
What are the three or four decisions that must be made the same way across the whole organization, and what can be left to the people closest to the work?
What's ready, what isn't, and what to do about it.
Models are commodities. Compute is a commodity. The thing competitors cannot replicate is your data, and the workflows it sits inside.
But most organizations are not data-ready. The data exists, but it's locked in systems that don't talk to each other, governed by rules nobody wrote down, and described in language only one team understands.
The work here is unglamorous but decisive: make the data findable, usable, and trusted before scaling AI on top of it. Otherwise, every AI initiative becomes a data project in disguise, and timelines triple.
A useful sequence:
- Identify the data your highest-value use cases actually depend on.
- Make that data clean, accessible, and governed first.
- Expand outward from there.
You do not need to fix all your data. You need to fix the data that matters for the work that matters.
Of the data we'd need for our top three AI use cases, what percentage is ready to use today?
The skills, roles, and rhythm needed to execute.
Most organizations underestimate how much new capability AI adoption requires, and overestimate how much of it is technical.
The scarce skills are not data scientists. They are the people who can sit between the business and the technology and translate: the workflow designers, the product thinkers, the change leaders, the operators who know what good looks like.
Three capabilities matter most:
- Translators, people who can move between business problems and AI possibilities without losing meaning in either direction.
- Designers, people who can rebuild a workflow around a human-plus-AI model, not just bolt AI onto the old one.
- Operators, the leaders inside the business who own the outcome, not just the pilot.
The rhythm matters too. AI adoption is not a project with a finish line. It's a quarterly cycle of learning, deploying, measuring, and adjusting.
Do we have the right people, in the right roles, with the right cadence, or are we hoping the existing org chart will absorb this work?
How the organization actually changes.
This is the lens most leaders skip. They sponsor the strategy, fund the tools, approve the pilots, and then assume adoption will follow. It usually doesn't.
People do not adopt AI because they're told to. They adopt it when three things are true at the same time: they want to use it, they're able to use it, and they're prompted to use it at the right moment in their work.
Want is built through clear communication of why and what's in it for them. Ability is built through training, role clarity, and removing friction. The prompt is built into the workflow itself, AI that lives outside the tools people use every day will be ignored.
There's a sharper truth worth naming: most resistance to AI is not irrational. It's people protecting their judgment, their craft, or their job. The leaders who handle adoption well take that resistance seriously rather than trying to override it.
For each AI initiative, who has to change how they work, and what are we doing to make that change wanted, possible, and obvious?
Using the lenses together.
The six lenses are not a sequence. They are a system. Strong value strategy fails without governance. Good governance fails without capability. Capability without adoption is wasted hiring.
The work of an executive sponsor is to look across all six and find the weakest one, because that's where the next investment of attention should go.