Why you need a business model canvas before you build an AI strategy

The failure mode we keep seeing

A leadership team commissions an AI strategy. Six weeks later a deck lands with a dozen use cases, some vendor logos, and a maturity radar chart. Nobody in the room can explain, in one sentence, how any of it changes the business. Everyone nods anyway, because the deck looks expensive and nobody wants to be the person who says the emperor has no clothes.

The missing artefact is almost always a shared business model. Without one, an AI strategy becomes a technology shopping list dressed up as a plan. It optimises for looking comprehensive rather than being true.

What the canvas gives you

The Business Model Canvas forces nine explicit answers: customer segments, value propositions, channels, customer relationships, revenue streams, key resources, key activities, key partnerships, and cost structure. None of this is new thinking — it is old thinking made visible, which is exactly the point.

For AI planning that matters because every serious initiative touches at least one block, and most touch two or three at once. Faster underwriting changes key activities and cost structure. A recommender system changes value proposition and customer relationships. A support copilot changes channels and cost structure together. If you cannot say which block an initiative moves, you have found a project looking for a business case, not the other way round.

A worked example

Take a mid-sized logistics firm. Its canvas shows revenue concentrated in two customer segments — retail distribution and cold chain — with a cost structure dominated by driver hours and fuel, and a value proposition built on reliability, not speed. Leadership wants to grow the cold chain segment over the next 18 months.

Run the AI brainstorm against that canvas and the noise falls away fast. A generic chatbot for the website barely touches any target block — cut it. Predictive maintenance on refrigerated units hits cost structure and directly protects the reliability promise underpinning the cold chain value proposition — keep it, and prioritise it. A route-optimisation model touches key activities and cost structure across both segments — keep it, but sequence it after maintenance since it is more disruptive to daily operations. Three initiatives went in; two survived, and both now have an owner who can explain why they matter in board language, not model language.

That is the entire value of the exercise: it turns "we should do something with AI" into "here is the one lever this segment needs, and here is what it changes."

How to run the exercise

1. Fill the canvas from the current state of the business, not the aspiration. Aspirational canvases produce aspirational, unfunded strategies.
2. Mark the two or three blocks where leadership genuinely wants change in the next 12 to 18 months. Not "all of them" — pick the ones tied to an actual growth or margin target.
3. Only then brainstorm AI initiatives, and tag each one to the block or blocks it affects.
4. Drop or defer anything that does not touch a target block, no matter how well the demo went.
5. Revisit the canvas when the business model itself shifts — a new segment, a pricing change, a channel you have exited. A canvas from two years ago is a historical document, not a planning tool.

This takes an afternoon with the right people in the room, not a quarter with a consultancy. It also makes every downstream conversation sharper, because the argument for or against a project stops being about the technology and starts being about the business.

Pitfalls to watch for

The most common failure is doing the canvas alone at a desk instead of with the people who actually run the segments. A canvas built by one strategist reflects one person's model of the business, which is rarely the real one. Get sales, operations, and finance in the room, even for an hour.

The second failure is treating the canvas as a one-off compliance step before "the real AI work" starts. If it is not referenced again when initiatives are prioritised, it did not do its job.

The third is vagueness in the value proposition block. "We deliver quality service" tells you nothing about where AI helps. Push for the sentence a customer would actually say about why they choose you over the alternative.

A note on shared language

The canvas also gives non-technical stakeholders a vocabulary they can use with confidence. "This project changes our cost structure by automating tier-one support" travels through a boardroom far better than "we are deploying an LLM-powered agent," because it tells the board what changes for the business, not what changes in the stack.

Where Florence helps

Florence's Organisation and Canvas steps capture this model before you touch the Assessment or Insights steps, so every AI initiative that later reaches the Strategy and Projects stages already carries a tag back to the block it affects. If a project can't be tagged, Florence surfaces that gap rather than letting it quietly through to a brief.