Bootstrap your AI strategy from the documents you already have

The blank-page problem

Most leadership teams do not start their AI strategy from a blank page. They start from a folder: last year's strategy deck, a business plan the board has already signed off, a product roadmap, a market research report, a couple of investor updates. The problem is that all of that context lives in PDFs and slide decks that nobody wants to retype into yet another tool.

So it gets skipped. The AI strategy work restarts from scratch, forgets half of what the organisation already decided, and produces a document that leadership recognises as thinner than what they already had.

What context documents change

Florence's Context Documents step is designed for exactly this moment. You upload the artefacts you already have — strategy deck, business plan, board update, product brief, market research — and Florence does two things with them.

First, it stores a concise summary that the AI agents can quote in chat. So when Ruth asks "what does your current strategy say about international expansion?", the answer comes from your own document, not from a generic template.

Second, in extract mode, Florence proposes structured values for your Organisation profile, Business Model Canvas, AI Readiness assessment and AI Strategy sections. Every suggestion is reviewable: you accept the ones that match reality, edit the ones that need refining, and dismiss the rest. Nothing writes to your workspace without your click.

A worked example

Take a mid-sized logistics company: forty staff, a board that meets quarterly, and no prior AI work beyond a chatbot pilot that stalled. The founder has two documents to hand: a 2025 business plan (18 slides) and last quarter's board update (a two-page memo).

They upload both, set the business plan to extract mode and the board update to reference-only, since it is mostly commentary rather than structured fact. Florence processes the business plan first. Within a couple of minutes it proposes: company size and sector for the Organisation profile, three customer segments and two revenue streams for the Canvas, and a handful of AI Readiness flags — no data warehouse, one analyst on staff, prior chatbot attempt logged as a failed initiative rather than ignored.

The founder accepts the segments and revenue streams outright, because the plan describes them accurately. They edit the sector description, because the plan is a year stale and the company has since moved into freight brokerage as well as haulage. They skip a proposed AI Readiness suggestion about "data maturity" because it is vague boilerplate the source document itself never really substantiated.

Ten minutes later, Organisation and Canvas are two-thirds filled before the founder has typed a single field manually. The board update, still in reference-only mode, means that when they later ask Steph to draft AI strategy options, Steph can reference "the Q2 board update flagged margin pressure in haulage" without that memo ever being reduced to structured fields it wasn't suited to. That is the split working as intended: extract what is structured and current, reference what is contextual and prose-heavy.

A five-minute workflow

1. Open Context Documents and upload two or three of your most current artefacts. A strategy deck and a business plan is usually enough.
2. Choose 'Extract suggestions' as the mode. Reference-only is fine for supporting material like industry reports.
3. Wait a minute for Florence to process each file. Suggestions appear grouped by target screen with human-readable field names.
4. Accept the values that describe your business accurately. Skip anything vague.
5. Move to Organisation, Canvas, AI Readiness or Strategy — the accepted values are already pre-filled. Use the 'Generate' buttons to fill the remaining gaps, not to overwrite what you just seeded.

In practice this cuts the first pass of a strategy from a week of workshops to an afternoon of review.

What to upload, and what not to

Good candidates: current business plan, most recent strategy deck, latest board update, product roadmap, customer research summary, competitive analysis, any prior AI or data strategy.

Less useful: highly technical architecture documents (unless AI is already part of the estate), everything you have ever written on the topic, or documents older than 18 months that no longer reflect the business.

If in doubt, upload fewer, more current documents. Quality of extraction beats quantity every time. As the worked example shows, one well-chosen deck outperforms a folder dump — the founder got more usable output from two documents handled deliberately than they would have from ten uploaded indiscriminately.

Trust and control

Every uploaded file is scoped to your active strategy and inherits the same sharing rules — only people you have explicitly shared the strategy with can see them. Files are never used to train external models. You can delete a document at any time, and switching a document from 'Extract' back to 'Reference only' stops it proposing new suggestions without losing the summary.

Why this matters for AI strategy specifically

AI strategy fails more often from missing business context than from missing technical knowledge. The organisations that get the best answers from Florence are the ones that give it the same context a new Chief of Staff would need on day one.

If you have a strategy deck or business plan sitting in a shared drive right now, that is your starting point. Open Context Documents, upload it, and see how much of Organisation and Canvas fills itself in before you write a single field by hand.