AI automated the research and first drafts that used to teach juniors how to think. A creative director rebuilt junior roles around judgment instead of output.

AI made the team 40 per cent faster, so they said yes to everything and margins dropped eleven points. Faster is only more profitable if you take back the time.

Asked how she knew AI was working, a founder said she could feel it. Eight weeks of before-and-after tracking on three workflows turned that into evidence.

One founder rebuilt everything around AI: forty-two people to twenty-eight, same revenue, margins roughly doubled. AI as infrastructure rather than a tool.

A new Head of AI spent three months on training sessions nobody used. Sitting with teams and fixing what took too long that week got them coming to him instead.

A client sent an AI addendum and two weeks to sign it. A procurement exercise, not a confrontation, and it forced the governance work that was overdue anyway.

Eleven people, four AI tools, three teams, and nobody knew where client data went. Regulated-client data sat in a free-tier tool. Fixing it took two weeks.

She built a serious AI capability and pitched it as efficiency. Clients wanted to know what AI meant for their own business. The outward-facing answer wins.

Three agency AI rollouts quietly failed as people drifted back to old habits. The problem was ownership, not resistance. People implement what they design.

Founders say their agency is at Level 2 or 3 on AI maturity. Ask about a legally reviewed policy, contract cover and data flows, and the answer is not yet.

An eight-year financial services client was lost because a strategist used free ChatGPT under a 2019 NDA. Nobody had told her not to. No policy existed.

She thought her agency was at Governed on the AI maturity model. No written policy, no data map, 2022 contracts. Governed is where you can safely accelerate.

Her new business came from referrals and she had no idea which clients sent them or why. A referral practice is a habit: know your advocates and ask.

Day rates unchanged since 2020 meant the agency cut its price by RPI every year while clients raised theirs. Raise prices annually, without apologising.

Two days in a country house produced a fifty-six-page strategy nobody could recall by January. Agencies that execute commit to fewer things, not tidier ones.

Strategy days produced plans nobody executed. A small team of future leaders who research, design and pilot the changes themselves changed the execution rate.

A founder celebrated his sale, then reread the earnout: three years of targets under someone else's strategy. Build the agency so the buyer does not need you.

A founder admitted most clients would leave if he did. Two years building a layer of people with real authority between him and them got a far better deal.

Founders guess their biggest client is 20 per cent of revenue. It is often 55. Above 30, buyers price the risk aggressively. Know the real number first.

I once built a service line in a month and buried it in three. The method I use now is slow at the front: research, small pilot, then roll-out.