Narrator: Three hundred AI licences. Fifty people using them. And every AI question in the building lands on one desk. An oven doesn’t make you a chef.

Narrator: Sound familiar? You’re not on your own. Maybe the board asked you to “do something with AI”, and it got filed under IT. Maybe you signed off the spend, and can’t say what’s changed. Maybe most people use it like a search box, and the pilot that demoed so well went quiet. Or maybe you haven’t bought anything yet, and people are using their own AI apps anyway. So why isn’t it moving, and what do you do next? In seven minutes: the steps organisations climb, what stalls each one, and how to tell where you stand. We’ll follow one job all the way up: winning a bid.

Narrator: First, the reassurance. About four in ten UK businesses with ten or more staff now use AI. But in June, the Office for National Statistics asked the ones using it how much. Only one in ten said extensively. Wide, but shallow. You’re not behind. You’re normal.

Narrator: Here’s the map: the AI Stairway™ to Success. It’s our own model, based on our own experience, though most published models share its shape. Think of it as a kitchen. People cook at home. The ovens arrive. People learn to cook, and write the recipes down. Then stations, a full service, and at the top, a restaurant designed around its guests. The lower steps are about one person getting good. The middle, the team sharing it. The top, changing the work itself. So which step are you on? Trick question. Hold that thought.

Narrator: Step zero. Siobhan, your bid manager, is pasting last year’s tender into an AI app on her own phone. Almost a third of UK workers who use AI at work say their employer doesn’t know. So your people can be further up the stairs than your organisation. Step one: the organisation catches up. Approved accounts, a policy, a lunch-and-learn. And then, not much. Two in three employees who use AI at work at least once a month say they’ve had no training in it. An oven doesn’t make you a chef. You learn to cook by cooking, with someone showing you.

Narrator: Step two. Siobhan gets good. Her bids get faster and sharper. So everyone sends her their bids, and their questions. Being good at it has only earned her more work. Meanwhile, someone writes a forty-page prompt pack, without asking a single team what they do. It’s been opened twice.

Narrator: In kitchen terms, Siobhan’s now the head chef, and the only one who can cook. In our experience, most organisations stall here, and it’s rarely the AI. Every step up is the same move: taking what works for one person, and making it work for everyone, without them. If it only works when she’s in, it isn’t a step you’re on. Hand the know-how to an AI working group, and the group becomes the queue. A good one does the opposite: it spreads the cooking. Stanford studied AI projects that succeeded. Most of the hardest problems were change, process and data. Not the model.

Narrator: Step three is the fix. Siobhan’s method gets written down: the best past bids, what good looks like, the checks she does in her head. Now it’s the house recipe, and anyone can cook it. Siobhan can go on holiday, and the bids still go out. Then the team improves it, a little, every time. That’s kaizen. It doesn’t need a big budget. It needs time, and permission to share what didn’t work.

Narrator: If you lead the organisation, that time and permission come from you. Approving AI isn’t backing it. In a 2026 survey of UK managers, only thirteen per cent strongly agreed their senior leaders actively use AI themselves. People copy what you do, not what you sign. Three things only you can do. One: use it in front of people, and say what went wrong. Two: give people time to learn, starting with their managers. Three: name an owner for every piece of work, and make sure it isn’t you.

Narrator: Step four: the work itself starts to change. A station: one job with its own recipe, its own kit, its own checks, and a chef who answers for it. That’s what a good AI agent is. For bids: an assistant that drafts from your real documents, tested on old bids first, with an owner. The agents we see fail skip this: a page of FAQs instead of the real documents, and no owner. They’re rarely bad AI. They’re unowned work.

Narrator: Step five: service. Every bid goes through the station, the tricky ones go to a person, and you measure turnaround, rework and wins against how it ran before. That means redesigning the work, not putting a new oven in the old corner. The first weeks often cost more time than they save. That’s when a leader’s backing counts. To be fair, from here the tools and data matter more. And unlike ovens, they get better every few months. Better ovens still need cooks.

Narrator: Step six: AI-first. It’s often heard as fewer staff. It means the first option you consider, not the first person you remove. One pantry feeds every station: the bid library also feeds case studies, reports and new starters. People own the work, and customers can still reach a person. Few are here: under one in ten large companies, in one recent consultancy study. And higher isn’t better for everyone. Climb as far as the work pays for.

Narrator: So why isn’t it moving? Often, because what works sits with one or two people, and nobody has the time, or the job, to spread it. What next? Not a big programme. Kaizen: small improvements by the people doing the work, written down and tried again. This week, find the best cook in one team, even if they’re using their own app, and ask what it would take to write the recipe down. And the trick question? Your organisation isn’t on a step. Its parts are: leadership, process, data, security and people. Our AI Readiness Score reports on each, and our course, The AI Stairway™, builds the skills. We make both, so weigh that. The bid that started on Siobhan’s phone now goes out without her. You can buy the oven. You have to grow the cooks.