The short version
- The argument: strategy first, technology second. Most leaders are stuck in operational detail, and AI's real value is as a Thought Partner that shortens the time from data to decision.
- The practical core: the CRIT prompting framework, Context, Role, Interview, Task. The Interview step is the one nearly everybody skips, and it is the sharpest idea in the book.
- What is strong: immediately usable prompts, real advisory examples, and a consistent refusal to make the tool the hero.
- What is missing: the organisational half. Change management, governance, measurement, and the structured practice most people need before a framework becomes a habit.
- Verdict: four and a half stars. Read it, apply CRIT to one real decision this week, then decide how you scale the capability past yourself.
There is a particular kind of business book you buy in an airport or a holiday bookshop, read forty pages of, and never open again. I was braced for that. What I got instead was a book I have since recommended to clients, argued with in the margins, and lifted one framework out of that I now use most weeks.
So this is a genuine review rather than a summary, written from a specific vantage point. I co-run an Ofsted Good, Skills England approved AI apprenticeship and training provider. Every week I watch the gap between leaders who have read something like this and leaders whose organisations have actually changed. That gap is the interesting part, and it is where the book stops.
Who wrote it, and why that matters
Geoff Woods is the founder of AI Leadership. Before that he was Chief Growth Officer at Jindal Steel & Power, where, according to his publisher, the market cap grew from $750 million to more than $12 billion in four years. He also co-founded the training and consulting company behind The ONE Thing, advising businesses from $10 million to $60 billion in revenue.
That combination matters because it is unusual. Most AI books are written either by technologists who have never carried a P&L, or by consultants who have never shipped anything. Woods has run the operational job and the advisory one, and it shows in the examples. Naveen Jindal, chairman of Jindal Steel & Power, is quoted on the publisher's own page saying the market cap "soared from $750 million to $12 billion" under that guidance. Treat vendor testimonials with the usual scepticism, but the operator experience underneath them is real.
Check your edition. A new and updated edition was published on 7 July 2026 by Authors Equity, distributed by Simon & Schuster, 304 pages, ISBN 9798893313109. If your copy predates mid-2026 it will not have the updates. Given how fast this field moves, that is not a trivial difference.
The core argument: strategy first, technology second
Woods' central claim is straightforward and hard to argue with.
"The difference between growing your business and going out of business is your ability to think strategically."
The AI-Driven Leader — the book's own framingThe problem, he says, is that most leaders are trapped in the operational weeds and never get to that thinking. Used properly, AI becomes a strategic Thought Partner that collapses the time from data to decision, frees capacity, and multiplies the impact of people. Strategy first. Technology second.
That single framing puts it ahead of most of the breathless "AI will change everything, here are 47 tools" books. Chris Winton, former Chief People Officer at FedEx and Tesla, catches it well in his cover endorsement: the guidance aims at business goals "without making AI the goal itself".
It is the same argument we make to employers who ask which model they should standardise on. That is nearly always the wrong first question, and we have made the case at length both when an open Chinese model reached the frontier and when Opus 5 launched at half the price of the flagship. Your model will change three times before your capability does.
CRIT: the practical heart of the book
If you take one thing from the book, take the CRIT prompting framework. Instead of treating AI like a smarter search engine, you work through four steps.
The four steps
C
Context
Rich, specific detail about your situation. Sector, size, constraints, what you have already tried, what "good" looks like. Most people give it two lines and wonder why they get two lines back.
R
Role
A specific expert role. Not "you are a helpful assistant" but "you are an operations strategist who has run AI adoption in mid-sized UK logistics firms".
I
Interview
Make it interview you, one question at a time, before it answers. This surfaces the context you did not realise you needed to share. It is the step almost everyone skips.
T
Task
Only now the task, with the shape of the output you want. Options ranked by time to value, risks, and what would have to be true for each to work.
The same question, asked twice
The gap between a lazy prompt and a CRIT prompt is easier to feel than to describe, so here is the same underlying question both ways.
What most people type
How can we use AI in our customer service team?
What you get back: a tidy, generic list. Chatbots. Sentiment analysis. Ticket routing. Suggested replies. All true, none of it aware of your volumes, your systems, your headcount constraint or the fact you tried a chatbot in 2024 and it went badly. You could have got this from any blog post, including this one.
The same question, through CRIT
Context. I am operations director at a 240-person UK logistics business. Our service desk is 18 people handling around 900 emails a week, average first response six hours. The board wants headcount flat next year while volume grows 20%. We run Microsoft 365 and have no data team.
Role. Act as an operations strategist who has led AI adoption programmes in mid-sized logistics businesses.
Interview. Before advising, interview me one question at a time until you have enough to give a recommendation you would stake your reputation on. Ask one question, wait for my answer, then ask the next.
Task. Then give me three options for absorbing 20% volume growth without adding headcount, ranked by time to value, with the risks of each and what would have to be true for each to work.
Same question. Completely different conversation. The second one produces something you could take to a board meeting, because it has been forced to reckon with your actual constraints rather than the average company's.
Interview is the step everyone skips
Context and Role are now fairly widely understood. Task is obvious. Interview is the one that separates people who get real value from people who get demos, and it is the sharpest idea in the book.
The reason it works is that you do not know what you have failed to mention. You are too close to your own situation to see which details are load-bearing. When you instruct the model to question you one at a time before answering, it will ask you things you had not thought to volunteer: what happened when you tried this before, who has to approve it, what your actual constraint is as opposed to your stated one. Half the value arrives during the interview, before any answer is produced, because you end up thinking more clearly about your own problem.
It also costs you nothing to try. Take one real decision on your desk this week, add the Interview instruction, and see what it asks.
Why this looks familiar to our apprentices
I had been using variations of this with clients and learners for a while before I read the book, which is partly why it landed. Woods codifies it more cleanly than I had.
There is a closer parallel too. In month one of our AI & Automation Practitioner Level 4, before anyone touches an automation tool, apprentices learn RICF: Role, Instruction, Context, Format. Then they try it on a real problem from their own department. The overlap with CRIT is not a coincidence and neither of us invented it. Anyone who works with these models seriously converges on the same finding: structure beats cleverness. CRIT leans harder on the interview; RICF leans harder on format, which matters when the output has to drop straight into a report or a system. Either will do. Having one at all is the point.
Thought Leader, Thought Partner
Woods is equally clear on the mindset shift. You remain the Thought Leader. AI is the Thought Partner. You do not abdicate judgment. You use it to pressure-test assumptions, simulate stakeholders, surface blind spots and force clearer thinking.
The 80/20 discipline sits alongside it: focus people on the 20% of their role that drives 80% of results, and let AI handle or eliminate the rest. That is operational common sense rather than a revelation, but it is stated plainly and applied consistently. The Musk-inspired process questions land well too: question every requirement, delete, simplify, accelerate, automate, in that order. Most organisations automate first and wonder why they have made a bad process faster.
I also appreciated the consistent reminder to keep people at the centre and to lead with empathetic strength while driving change. In our world that is non-negotiable. AI fluency without the human system around it usually produces either quiet resistance or expensive shelf-ware.
What other reviewers make of it
A book review is not worth much if it only quotes the people who agree with you, so here is the range.
Enthusiastic · Jack Spirko, The Survival Podcast
Spirko was impressed enough to review it when he was only about a quarter of the way through, and says his endorsement got stronger on a second pass. His useful observation is that CRIT was designed for executives making high-level decisions but transfers almost anywhere: he has used it for supplier quotes, recipe development and building a research assistant. For newcomers, he calls CRIT "the perfect on-ramp".
Critical · Robert Carnes, Medium
Carnes gives Woods credit for the central idea, calling the Thought Leader and Thought Partner distinction "genuinely helpful advice", though he notes it is not new. His objection is not the substance but the craft: he is concerned by "the quality of the writing and the book's structure", and suspects the manuscript leaned heavily on an LLM. It is a fair criticism and one worth knowing before you buy.
The aggregate · Goodreads
It averages 3.86 out of 5 from more than 2,400 ratings. That is a solid business-book score rather than an adored one, and it fits what I found: a genuinely useful framework inside a book that some readers find thin around it.
My own view sits closer to Spirko than Carnes, but Carnes is not wrong. There are passages that read like they were generated rather than written. I found I did not much care, because I did not buy it for the prose. I bought it for a framework I could use on Monday, and I got one.
Where the book stops and the organisation starts
Woods is strongest on the individual leader's personal operating system: how you escape the weeds and think better. He is lighter on the harder organisational realities we see every week. That is not really a criticism, it is a difference in scope, but it is the difference that decides whether anything changes after you finish the last chapter.
What a book can do, and what it cannot
| What the book gives you | What it cannot give you | Where that gets built |
|---|---|---|
| A framework for better thinking (CRIT) | Repeated practice under supervision until it becomes habit | Coaching and work-based projects |
| The Thought Partner mindset | Consistency across 40 people of mixed confidence | A structured, assessed programme |
| Prompts you can use today | Governance that controls risk without killing velocity | the governance unit at Level 5 |
| Motivation to start | Evidence that it moved a business metric | Measured projects on your own priorities |
A book can light the fuse. An apprenticeship or a properly designed corporate programme has to keep the fire going for months while people deliver real outcomes in their day job. Different problems, different instruments.
Which route fits which problem
If the book landed and you are wondering what "next" looks like, it usually comes down to one of three situations. All are fundable in England through the Growth and Skills Levy, and none require a coding background.
-
One senior leader wants to think like this properly
AI for Senior Leaders at Level 4, which carries a free CMI Level 7, or a single Level 5 AI Leadership unit if you want the capability without a full apprenticeship. Start smaller still with a free taster day.
-
You want the whole leadership team fluent, together
The AI Leadership Pathway runs all three units, AU0009 strategy and opportunity, AU0010 adoption and governance and AU0011 delivery and transformation, as one closed cohort at £2,250 per leader from the levy. Closed cohorts work better here because the team argues about its own strategy rather than a case study.
-
You need the ideas turned into shipped work across functions
That is a practitioner job, not a leadership one. The AI & Automation Practitioner Level 4 builds people who deliver, with a Claude edition if you have standardised on Anthropic, and Coachy, our own AI tutor, for support at 11pm on a Tuesday rather than at the next review.
Tell us the problem and we will map it against your priorities and your levy position. 25 minutes, no obligation, and we will tell you plainly if a short course would serve you better than an apprenticeship.
Book a discovery call →The honest verdict
It is not a technical book and does not try to be. It will not teach you model architecture or the latest agent frameworks. It will, however, stop most executives asking the wrong first question, "how do we use AI?", and start them asking the right one: what is the highest-leverage problem here, and how might AI help us solve it while keeping our people and our strategy in the driving seat?
Clear, actionable, and unusually disciplined about keeping strategy above the tool. Four and a half stars for any leader serious about not being left behind, and a strong recommendation for the senior sponsors of any AI capability programme we run. Pair it with real practice and structured development and the combination is powerful.
So: read it. Apply CRIT this week to one real decision. Then decide how you are going to scale that capability beyond yourself, because that is where the actual results live.
Frequently asked questions.
What is The AI-Driven Leader by Geoff Woods about?
It argues that the difference between growing a business and going out of business is the ability to think strategically, and that most leaders are stuck in operational detail. Used properly, AI becomes a strategic Thought Partner that shortens the time from data to decision and multiplies the impact of people. The order matters: strategy first, technology second. It is a leadership book, not a technical one.
What is the CRIT prompting framework?
CRIT stands for Context, Role, Interview, Task. You give the model rich context about your situation, assign it a specific expert role, ask it to interview you one question at a time to surface context you had not thought to share, and only then give it the task. The Interview step is the one most people skip and the one that changes output quality most.
Is CRIT the same as RICF?
They are close cousins. RICF stands for Role, Instruction, Context, Format and is the structure we teach in month one of the AI & Automation Practitioner Level 4 apprenticeship. Both insist you specify who the model is being, what situation it is working in, and what shape the answer should take. CRIT adds an explicit Interview step, where the model questions you before answering. RICF puts more weight on Format, which matters when the output has to drop straight into a report or a system. Use either. The gain comes from having a structure at all.
Is there a new edition of The AI-Driven Leader?
Yes. A new and updated 2026 edition was published on 7 July 2026 by Authors Equity, distributed by Simon & Schuster, at 304 pages, ISBN 9798893313109. If you have a copy bought before mid-2026 it will not contain the updates. Worth checking the copyright page before you recommend it to your team.
Is The AI-Driven Leader worth reading?
For senior leaders who have not yet found a disciplined way to use AI, yes. We rate it four and a half stars out of five. It is clear, immediately actionable and refuses to make the technology the hero. Reviews elsewhere are more mixed: it averages 3.86 from more than 2,400 ratings on Goodreads, and at least one reviewer has criticised the quality of the prose rather than the substance.
What does the book not cover?
It is strongest on the individual leader's personal operating system. It is lighter on the organisational realities that decide whether AI adoption sticks: sustained change management across teams of mixed capability, governance that does not kill velocity, measuring whether the work is moving a business metric rather than generating activity, and the structured practice and accountability most people need before a framework becomes a habit.
How do UK employers turn these ideas into organisational capability?
Through structured, work-based development rather than a one-off workshop, funded in England through the Growth and Skills Levy. There are three usual routes. An individual senior leader takes AI for Senior Leaders at Level 4 with a free CMI Level 7, or a single Level 5 AI Leadership unit. A whole leadership team runs the AI Leadership Pathway, units AU0009, AU0010 and AU0011, as one closed cohort at £2,250 per leader from the levy. To turn the ideas into shipped work across functions, the AI & Automation Practitioner Level 4 apprenticeship builds practitioners. No coding background is required for any of them.
Sources and disclosure: book details, publication date, author biography and cover endorsements from the publisher's page. Reviewer perspectives from The Survival Podcast and Robert Carnes on Medium; aggregate rating from Goodreads, correct at the time of writing. This is an independent review. We have no commercial relationship with the author or publisher, and the copy was bought at full price in a bookshop.