AI Without
the Hype
Internal playbook for the training lead. One rule underneath all of it: don't believe it until you've seen them do it.
Why this isn't optional
The gap, then the prize
Two sets of numbers. Read the bad one first.
Almost everyone has deployed AI somewhere. Almost nobody is seeing value from it. The teams that redesign the work around it see a lot.
The gap
of companies report no significant value from AI, despite nearly nine in ten having deployed it somewhere by the end of 2025.
McKinsey, State of AI 2025is the entire share of companies that count as AI high performers, seeing more than 5% EBIT impact from it.
McKinsey, 2025have fundamentally redesigned even some workflows around generative AI. The rest mostly bolt it onto the old process.
McKinsey, 2025are scaling AI agents in any function at all.
McKinsey, 2025in product development are not using agents at all.
McKinsey, 2025report at least one negative consequence from AI in the past year, most commonly inaccuracy.
McKinsey, 2025follow most of the recognised practices for adopting and scaling generative AI. This is where most programmes stall.
McKinsey, 2025The tool is nearly everywhere. The value, almost nowhere. Most teams never change the shape of the work.
The prize
Done properly, the same tool produces this instead.
labour productivity gains at the companies most exposed to AI, far outpacing other businesses.
PwC, 2026 Global AI Jobs Barometerhigher growth in revenue per employee at the industries most able to use AI, versus those least able.
PwC, 2025higher productivity growth at the most AI-exposed companies than at the least exposed.
PwC, 2026productivity growth in AI-exposed industries since 2018, nearly four times the 7% rate they ran before generative AI.
PwC, 2025average wage premium for workers with AI skills, reaching as high as 118% in consumer markets.
PwC, 2026of companies using AI agents say those agents already deliver measurable value through higher productivity.
PwC, 2025faster growth in jobs requiring AI skills than the overall jobs market, roughly 69% against 9%.
PwC, 2026headcount growth at the most AI-exposed companies, versus 36% at the least. AI tends to expand strong teams, not shrink them.
PwC, 2026The gains are real, and large. They go to the teams that learn the tools and redesign the work around them.
Read this first
Don't believe it until you've seen it
This is for you, not the client.
AI is a steam engine moment. Not a faster typewriter. It changes what's possible, it doesn't just speed up what's already there.
If you don't believe that yet, sit with it before you run a session. The room will smell it on you in ten minutes.
The practical version for clients: AI is leverage. More output for less input.
Three places leverage hides: Money, Operations, Happiness. You'll lean on this on page 4, and constantly after that.
One warning, for you first. AI amplifies whatever's already there.
A chaotic firm with no clarity on who owns what gets faster chaos, not less of it. Spot that early. Flag it internally, don't surprise the client with it in week six.
Unless I've seen you do it, I don't believe you. People will tell you they're "using AI now." Usually that means a holiday itinerary, six weeks ago. Don't take their word for it. Watch them work, on real tasks, in the room.
Before you teach anything
Exploratory sessions, not assumptions
Before you build a single workflow, find out what's actually true.
What people use. What they know. Where they're already quietly drowning.
This is diagnostic work. Give it real time, not a five-minute warm-up before the "real" session starts.
What do they use
Audit the whole toolkit
Not just Claude, everything. Email, calendar, whatever spreadsheet someone built in 2019 that the team is too scared to touch. You're mapping the actual day, not the job description.
What do they know
Skip the assumption
Don't assume everyone's used ChatGPT by now. Some built real habits, some tried it once and got spooked, some never touched it. Find out which, without making anyone feel behind.
Exercises, not demos
The rule, in practice
Give people a real task from their week. Have them do it live, while you watch. Uncomfortable for about four minutes. Then the most useful four minutes of the day.
Bring me your chats
Review the real conversations
Ask people to share an actual chat with Claude, the messy kind, not a highlight reel. Look at it together. People also tidy up their habits once they know a chat might get reviewed.
Somewhere in this audit, you'll find what's actually drowning them.
Sometimes it matches the leverage workshop on page 4 exactly. Sometimes it's an inbox with four thousand unread emails, a calendar that's back to back with no thinking time, a sales team retyping the same email from memory.
Write it all down. Some of it becomes the first win. Some of it is just useful context.
The real shift isn't tool literacy. It's the default question changing from "is this worth automating" to "could AI help with this at all," even if the first pass is slower than just doing it the old way.
The first time pays for every time after.
Whatever else changes, these don't
The foundations every programme needs
Regardless of what the exploratory sessions found, every programme needs these underneath it.
Treat them as the floor, not optional extras.
Money
Where does the budget actually leak? Ask for specific numbers. Vague answers mean you push further.
Operations
What's the headache nobody's fixed because it's never quite bad enough to stop and address? Usually the best first win.
Happiness
What work genuinely drains people? Don't skip it. It turns "we got faster" into "we got our evenings back."
The least exciting session. The most important one.
Your job isn't to write it for them. It's to extract what's already in their heads, get it into a Claude Project, and keep it alive.
A Master Document written once and never opened again isn't context, it's an artifact. Every workflow you build afterward should add to it, not sit beside it.
Skip this, or let it go stale, and every later session produces weaker output. You'll feel it without knowing why.
You'll meet three types in every engagement.
Sponsors
Light up the second you mention AI. Use them. Get them talking in front of the room.
Neutral Blockers
The majority, roughly 80% of any room. Not resistant, just unconvinced. Give them an undeniable win.
Blockers
Usually afraid, not stubborn. Have that conversation privately, never in front of the room.
Ask the room lightly: "who here has already tried Claude or ChatGPT for something at home?" The hands that go up are your early Sponsors.
Yours to design
Foundations, building, scaling
Here's the shape, not the script.
Every programme moves through three phases. You can't scale before people trust the basics. You can't stay in foundations forever, or it turns into a seminar nobody applies.
What fills each phase is yours to write, based on what page 3 actually found, not a generic template.
Use the questions below to check your plan. Not to copy it.
- Has everyone done something real, live, that actually saved them time?
- Has the firm started writing down what's currently just in people's heads?
- Do you know your Sponsors, Neutral Blockers, and Blockers yet?
- Have you found at least one genuinely enthusiastic person to build around?
- Are the workflows theirs, built around what the audit found, not a template?
- Is someone besides you starting to champion this internally?
- Have you resisted rolling everything out everywhere at once?
- Did a win elsewhere travel to your biggest sceptic, instead of starting there?
- Can they keep going without you in the room, team or just the one enthusiast?
- Is there a rhythm that doesn't depend on you showing up to remind them?
- Could they explain this programme in their own words?
- Are people still asking "could this be easier too," unprompted?
Not a firm that's "used AI"
What you're actually trying to leave behind
- ✓Master Documents that are actually used, not filed away
- ✓At least one genuinely enthusiastic person, team or solo, still testing things
- ✓At least 3 to 5 real, working, repeatable workflows
- ✓A written playbook in their own voice, not a copy of this one
- ✓A rhythm that doesn't depend on you turning up to remind them
- ✓People who are still curious, still asking what else could be easier
Closing note to yourself
The deliverable isn't the workflows. It's curiosity that survives without you in the room. A firm still asking "could this part be easier too" in month nine has succeeded, even if half the original workflows have already been replaced by better ones they found themselves. You're not selling magic, you're building leverage and the habit of going looking for more of it. Run it well, keep it honest, and don't be afraid to tell a client something isn't working halfway through month two. That's worth more than another polished slide.