The readiness action guide
Your readiness profile, turned into action.
Six dimensions decide whether your organization is ready for AI — or about to spend heavily learning that it is not. This guide gives you the concrete moves for each one. Start with your lowest score, because that is where the next bit of readiness is cheapest to buy.
How to use this guide: take the readiness diagnostic if you have not already. Then come here with your dimension scores and work your lowest dimension first. Each section tells you what a low score actually means, the moves that raise it, and the single first move to make this week. Do not try to fix all six at once — one dimension, moved deliberately, pulls the others with it.
Strategy & sponsorship
A low strategy score means AI is happening without a mandate — no written ambition, no accountable owner, no protected budget. Effort scatters, nothing compounds, and the first hard tradeoff quietly ends the momentum.
Write the one-paragraph AI ambition
Tie AI to two or three business outcomes that matter this year, in language a board would accept. If you cannot say it in a paragraph, you do not have a strategy yet — you have activity.
Name a single accountable owner
One executive who owns AI transformation outcomes — not a committee, not "everyone." Ownership without a name is how initiatives stall in the gaps between departments.
Protect a real budget line
Move AI out of "things we fund if there is slack" into a committed line with a number on it. Pilots die the moment they have to compete with operational budgets for scraps.
Force agreement on the first problems
Get the top team to rank, on one page, which problems AI tackles first. Disagreement surfaced now is cheap; disagreement discovered mid-build is not.
Draft the one-paragraph ambition and send it to your leadership team with a single question: "Would you fund this?" The answers tell you whether you have alignment or just enthusiasm.
Data & technology
A low data score means your ambitions are ahead of your foundations — the data your best use cases need is scattered, distrusted, or locked behind access nobody will grant quickly. AI amplifies whatever data reality you actually have.
Inventory the data your top use cases need
For each priority opportunity, list the data required, where it lives, who owns it, and whether you can legally use it. Do this before vendor calls, not after.
Fix trust before scale
If people do not trust the numbers today, AI built on them will be trusted even less. Clean the few datasets your priorities depend on — narrow and deep beats broad and shallow.
Pressure-test integration early
Confirm AI tools can actually reach your existing systems and workflows. A brilliant model that cannot connect to where work happens is a demo, not a capability.
Make security a design input, not a final gate
Decide now how AI will handle sensitive data, access, and retention. Bolting security on after a pilot succeeds is how promising projects get frozen.
Take your single highest-priority use case and answer four questions about its data: does it exist, who owns it, can you use it, do you trust it. Each "no" is a task, not a surprise waiting for month three.
Skills & capacity
A low skills score means the plan assumes people and time you have not actually secured. AI transformation rarely fails for lack of tools — it fails for lack of anyone with the capacity to make them real.
Separate "can build" from "can use"
You need both — people who can configure and deploy AI, and teams who use it well in daily work. Map which you have, which you can hire, and which a partner covers.
Protect capacity, not just intent
Name who will actually run AI initiatives and free the time to do it. "Everyone will find a few hours" is how initiatives quietly starve.
Appoint a translator
One person who can move between business needs and technical delivery. Most AI projects fail in translation, not in technology.
Write the capability plan
For each gap, decide hire, train, or partner — with dates attached. A skills gap without a closing plan is just a risk you have chosen to ignore.
List the three roles your first initiative actually needs, and mark each as have, hire, or partner. The blanks are your real constraint — better found now than in month three.
Process & operating model
A low process score means you are aiming AI at work nobody has mapped, inside an operating model that cannot absorb the change. Automating an unclear process just makes the confusion faster and harder to unwind.
Map one priority workflow end-to-end
Half a day with the people who do the work — owners, handoffs, systems, exceptions. The conversation is the deliverable; a whiteboard beats a tool.
Baseline before you build
Capture current cycle time, cost, error rate, and volume. No baseline means no honest "after" — and no way to prove value later.
Clear the decision rights
Know who can approve a change to how work is done before you propose one. Ambiguous authority is where redesigns stall for months.
Plan the path from pilot to operations
Decide up front how a successful pilot becomes the normal way of working. Pilots with no route into operations stay pilots forever.
Book the mapping session for the workflow you most want to change, and invite the people who actually do it — not just the people who own it.
Governance & risk
A low governance score means AI is being used faster than anyone is managing its risk — no clear rules, no named accountability, no way to catch a model going wrong. That is not caution you are saving; it is exposure you are accumulating.
Publish a one-page acceptable-use note
What data and uses are allowed, restricted, and prohibited — plain language, real examples. If it needs a lawyer to read, employees will not.
Give risk a named owner
One person accountable for AI risk, privacy, and compliance. Not a committee — a name who can answer "is this still safe?"
Set criteria for vendors and tools
Decide what you require on security, data handling, and reliability before you evaluate, so choices are made on standards rather than sales decks.
Monitor after deployment
Decide what gets checked, how often, and what triggers escalation or shutdown — including drift, bias, and error. Governance that stops at launch is theater.
List every AI use happening today — including the unofficial ones — and mark which have an owner. The blanks are your actual risk register.
Culture & adoption
A low culture score means the tools may be bought but behavior has not changed — and it will not, unless people feel curious rather than threatened and safe rather than watched. Licenses are a cost; changed behavior is the value.
Have leaders use it visibly
Adoption follows what leaders do, not what they mandate. If executives are not using AI in their own work, no memo will move the organization.
Make it safe to experiment out loud
People adopt what they can try without fear. Create room to test, report what does not work, and be thanked rather than punished for it.
Train on real work, not demos
Each team practices on their own actual tasks with support in the room. Generic prompting workshops produce generic nothing.
Measure adoption and value, not tools deployed
Track usage, output quality, confidence, and business outcome together. A hundred logins that produce work people redo is not adoption — it is theater.
Pick one team and one workflow, and ask a leader to use it visibly for two weeks. Culture moves on example far faster than on encouragement.
Keep going
One dimension at a time.
Work your lowest dimension first, then retake the diagnostic in ninety days and watch the profile move. The field notes cover the same ground between now and then — practical guidance on leading AI transformation, sent when there is something worth reading.