Do we really need a business model canvas first?

Yes, usually. Without it, AI strategy optimises the wrong things. Here's what the canvas exposes.

Short answer

Yes, in most cases. You cannot build a useful AI strategy without understanding your value proposition, key activities, resources, customer relationships, and revenue streams. The canvas exposes where AI genuinely matters, and where it would just be technology theatre. Skip it and you end up choosing AI projects based on what's fashionable rather than what's connected to how your organisation actually makes money.

What the canvas actually is

The Business Model Canvas is a one-page way of describing how an organisation creates and captures value. It has nine blocks: customer segments (who you serve), value proposition (what problem you solve for them), channels (how you reach them), customer relationships (how you interact with them over time), revenue streams (how you get paid), key resources (what you depend on to operate), key activities (what you must do well), key partners (who does the rest for you), and cost structure (where the money goes). Filled in honestly, it takes an afternoon. Filled in badly, it takes ten minutes and tells you nothing.

The point isn't the diagram. It's the argument you have while filling it in. Most leadership teams discover they don't agree on the value proposition until they try to write it in one sentence.

A worked example

Take a mid-sized logistics firm. Its canvas might show: customer segments split between large retail contracts and smaller local businesses; value proposition of reliability and speed, not price; key activities being route planning and driver scheduling; key resources being driver relationships and a fleet management system; revenue concentrated in the retail contracts, with the local business segment barely covering its own cost to serve.

Once that's on the page, the AI conversation changes. Route optimisation touches a key activity and a real cost line — that's a strategic candidate. A chatbot for the local business segment might look appealing because it's cheap to build, but it's touching the smaller, lower-margin part of the business. It might still be worth doing, but it's a nice-to-have, not the centre of the strategy. Without the canvas, both ideas arrive with equal weight in a workshop, and the louder voice in the room wins.

What the canvas surfaces

Where value actually comes from. AI applied to a low-margin, low-volume channel rarely moves the needle, however good the demo looks.
Which activities are core vs contextual. Automating a core activity is strategic. Automating a contextual one is a nice-to-have, and should be budgeted and prioritised as one.
Which resources you depend on. Data, people, partners. If AI needs a resource you don't have — clean data, a particular skill set, a partner's API — that becomes a foundation project, not a delivery project. Skipping this step is how organisations end up six months into a build before discovering the data doesn't exist in usable form.
How you make money. Revenue-adjacent AI (pricing, retention, upsell) carries different governance and risk requirements to back-office AI (finance, HR). Conflating the two in a single project list means the wrong things get signed off with the wrong level of scrutiny.

Pitfalls to watch for

Writing it alone. A canvas one person fills in from their desk reflects their view of the business, not the business. Get commercial, operational, and delivery perspectives in the room.
Describing the business you wish you had. Canvases drift towards aspiration — the value proposition you're working towards rather than the one customers actually experience today. Be honest about the gap; the strategy needs to close it, not assume it's already closed.
Treating it as a one-off. A canvas from three years ago, before a pivot or an acquisition, will send your AI strategy in the wrong direction just as confidently as no canvas at all.
Skipping straight to the assessment. It's tempting to jump to readiness scoring because it feels more concrete. But readiness to do what? Without the canvas, the assessment has nothing to be relevant to.

Next steps

Block two hours with the people who actually run the business — not just the people who talk about it.
Fill in all nine blocks before debating any one of them. Half-finished canvases get argued over prematurely.
Circulate it and ask each function to challenge the parts that don't match their experience.
Only then start naming AI use cases, and test each one against the canvas: which block does it touch, and how much does that block matter to revenue?

Where Florence helps

Florence's Business Model Canvas step is built for exactly this. It gives you the structure so you're not starting from a blank page, and it carries what you capture straight into the AI Readiness Assessment and Insights steps that follow — so the AI ideas Florence later surfaces are anchored to your actual value proposition and revenue streams, not to a generic list of AI use cases that happen to be popular this year.