Here is a sound AI planning framework that I can make impactful, doable and affordable for private B2B companies with 20 to 1,200 people and primarily knowledge-based workforces, plus the execution habits that turn the plan into results.
It builds on a strong 2027 AI transformation planning session from Section, an AI training company. Their thinking is excellent, but it’s built for large enterprises. Their own example budget for a 1,000-person company is about $6.6 million a year, with a full-time Head of AI, an AI governance council, and a large champions program. Most mid-sized companies can’t start there, and they don’t need to.
Here’s the same framework, sized for the companies I work with.
Six Steps, Right-Sized
Decide what you’ll measure, before anything else
Track three levels: are people using AI, which workflows did it change, and did the business move (revenue, margin, revenue per employee). Right-sized: pick three to five numbers you already track and write down where they stand today. Without a starting point, you can’t show anyone where the money went.
Sort the work before you automate it
Ask of every workflow: if we were starting this company today, would this even exist? Cut what isn’t needed rather than automating it. Right-sized: our Sort the Work First exercise does this with your leadership team in a single working session.
Start with one bottleneck, not a big program
Pick the first projects where a mistake does little damage and is easy to undo. Prove value there, then expand. Right-sized: a 90-day Lighthouse Project on one bottleneck, so the results can pay for the next step.
Budget in one place, and protect the people part
Keep one AI budget, not costs scattered across departments. Tools are the predictable part; training and day-to-day support are where most companies underspend, and it shows. Right-sized: instead of hiring a full-time Head of AI, use a fractional leader and a handful of internal champions who know the work.
Keep the rules light
Use business-grade AI tools that don’t train on your data, follow any rules your industry actually requires, and stop there. Right-sized: no council. The leadership team reviews AI once a month, and the CEO breaks ties.
Tell your people the truth, in writing
Explain why you’re using AI, what will change, and how everyone can take part. Don’t promise nothing will change; people can tell. Right-sized: a one-page note from the CEO, followed by an honest all-hands conversation.
Set your measures and starting numbers before you spend. Everything else, from budget to which workflows go first, depends on knowing what you’re trying to move.
From Plan to Results
Most AI plans look great on slides and die in implementation. The gap between “we have a plan” and “it’s working” is almost always an execution gap, not a technology gap. Four habits close it.
1. Start with a design sprint
A design sprint is a five-day, structured way to prototype a solution, test it with real users, and learn before you commit serious money: Map the problem, Sketch several options, Decide on the most promising one, Prototype a rough but realistic version, and Test it with the people who will use it. The prototype doesn’t have to be perfect; it has to be real enough to get honest reactions. Committees that design AI projects for months without showing users anything consistently build things that don’t fit how people actually work.
2. Run a steady Plan, Deploy, Review rhythm
- Plan: set objectives, measures, and an honest read of how much change your organization can absorb right now. Our Change Capacity Index helps here. If capacity is short, phase the rollout or build capacity first.
- Deploy: treat launch as a learning event. Watch early signals closely in the first weeks.
- Review: not a status update, a real learning conversation. What did we expect? What happened? What does the gap tell us? What do we change?
3. Hold an Achievement Assessment Conference after each milestone
Four questions, answered honestly by the whole team: What did we do? What did we achieve? What did we learn? What’s next? The third question matters most and gets the least time. One rule: this is not a performance review. No one is on trial. If it turns punitive, it stops being useful immediately.
4. Choose your path to scale
Adopt and scale when the sprint produced something that works reliably, the change is manageable, and the business case is clear. Move fast; hesitating after you’ve proven it is waste. Choose a Lighthouse Project when the solution is complex or readiness varies across teams. Prove it in one visible, well-supported unit, and let other teams see it working and ask for it. That pull builds momentum a mandate rarely can.
“The companies compounding their AI advantage aren’t the ones with the best strategies. They’re the ones with the best execution discipline.”
Why This Works for Mid-Sized Companies
Smaller companies have an advantage the enterprises don’t: leaders are close to the work, decisions are faster, and one well-chosen project can change the economics of a whole team. You don’t need a big program. You need a clear plan, one proven win, and the discipline to build on it.
If you’re setting your 2027 plan now, a 30-minute conversation is usually enough to sketch your version of this framework.
Source: Section, “How to Set Your 2027 AI Transformation Plan” (live session, October 21, 2026, recording available) and “Everything you need to know when building your 2027 AI plan.” Adaptations for mid-sized companies are ours.