Everyone has an AI strategy. Few have a plan for the people.

AI is being adopted much faster than it is being put to work.

Companies announce strategies and marketers use the tools every week, yet the value is slow to show up. Somewhere between the strategy and the working week the people considerations get lost. Who uses the AI, who controls it, and what happens to their time and their skills are rarely in the plan. We believe the biggest value is likely to be realised when AI is adopted purposefully and at scale. Instead we are seeing adoption by individuals and little by way of collective connection. As a result the real value that powers marketing as an engine for growth is slow to materialise, if at all.

We built Crew & Engine on the opposite principle. Humans use AI. Humans keep control. Humans are freed to spend their time on the work only they can do.

Adopted, not implemented

Start with marketing. McKinsey’s Global Marketer Survey found nearly 60% of marketers use AI multiple times a week. Only 28% say their company is pursuing a fundamental rewiring of marketing teams and workflows around it. The tools arrived, but the habits didn’t change. We set that out in The competency trap.

Leaders see the same gap from the top. In EY-Parthenon’s CEO Outlook survey of 100 UK CEOs, 34% named legacy technology and operating models as the biggest obstacle to turning AI’s productivity gains into measurable business value. EY-Parthenon’s CEO Outlook also found almost half of CEOs (47%) say their organisation isn’t developing skills quickly enough to keep pace with AI.

Where are the people in the strategy?

Nobody has measured how many AI strategies are built around people. The nearest evidence points one way.

The Thomson Reuters Foundation and UNESCO studied 2,972 companies across 11 sectors. 43.7% publicly communicate having an AI strategy. Only 12.4% report having a policy to ensure a human oversees AI systems. 31% claim AI training programmes, and 12% offer structured training with comprehensive coverage. The authors’ conclusion is that the challenge is “no longer awareness but operationalisation.”

Marketing leaders show the same leaning. In Gartner’s survey of 402 senior marketing leaders in North America and Europe, 65% of CMOs said AI will dramatically change the CMO’s role within two years. Only 32% said significant changes are needed to the CMO’s profile and skill set.

We should read the first set of data with care. The figures answer different questions, so we shouldn’t subtract one from another. They come from public disclosures and survey responses, and the authors say a lack of disclosure doesn’t prove a lack of practice. Sectors differ widely but the direction holds all the same. The plan gets announced and the people part gets left until later.

Humans use it. Humans keep control.

A policy can’t oversee anything. A person can.

The control that the Thomson Reuters Foundation data measures is a person, and a person needs the skill, the authority and the time to do it. A person asked to supervise an agent they were never taught to question isn’t a control. They are just a signature.

In a deployment, that means every control is a person with a decision to make.

• Two named owners. One is accountable for the machinery and one for the outcomes. Both are named people, not job titles.
• A person approves first. The agent proposes and a person decides. Autonomy comes later.
• No playbook improvises to your customer. Three of our fifteen can send a message, and every word they send was written or approved by a named person before it could go out. The machinery selects and sends, and it never composes. Price, contract and complaint always go to a person.
• A measured agreement rate before go-live. That means how often the agent matched your people’s judgement over a real sample of your own work. A demo doesn’t count.
• A signed change whenever a gate widens. It records that a named person has agreed to give the agent more room.
• A rollback switch a person has tested. One action returns the whole flow to manual, and someone has pressed it.

None of this is clever, but all of it is checkable, and it makes automation faster rather than slower. A measured agreement rate is the evidence that lets you widen a gate. Autonomy you have earned can be extended with confidence, and autonomy you haven’t earned gets switched off after the first mistake.

Humans are freed up

Look at a typical marketing week. In our experience, a third of it at most is judgement and experience: the positioning call, the hard conversation with sales, the thing in the numbers nobody else spots. The rest is overhead, such as routing leads, de-duplicating records, chasing approvals and rebuilding Monday’s numbers.

Playbooks can do that work continuously, at a scale no team matches. We want more automation, not less. Handing the work over is how the people get their time back, and time is what oversight, judgement and creativity run on. A policy won’t give anyone time, but an operating model will.

Ours rests on a simple division of labour. Playbooks take the industrial work, supervised. The Crew do the work that needs a person: judgement, taste, skill, creativity and the calls that matter. We believe that’s what keeps marketing value-led, supports growth and builds real connection between a company and the people it serves. Both ends of that connection are human.

Our Capability Map reflects it. It lists 36 capabilities a marketing function needs, and 14 are frontier: new, or newly scarce, since AI arrived. Of those 14, 13 are led by people or directed by them with machines executing. Seven are Crew-led outright, including research judgement, creative direction and taste, applied AI fluency and agent supervision. Six are Hybrid. One, building and running the agentic workflows, is led by the Engine. People set the direction and the standard, and machines carry the volume.

Why Crew & Engine

Both halves are in the name, and Crew comes first because the judgement is theirs. We’re a practical company. We make human skill and AI playbooks work together in a way that respects and preserves that skill, to deliver growth and business value.

We place people as well as build playbooks, and we supply four evidence-based answers: hire it, borrow it, automate it, or stop doing it. Where the evidence says automate, we automate. The aim is a team with more time for the work only people do, and playbooks doing the rest.

Seven questions to ask

Take these into your next conversation about AI, with us or with anyone else.

1. Who are the two named owners? Ask for names. A role title isn’t an answer.
2. What was the agreement rate before go-live? And over how many real decisions.
3. Where is the rollback switch? And when was it last tested, rather than last documented.
4. Can I see the decision log? Every action the agent took, and why.
5. Which gates have been relaxed, and who signed?
6. Who are your sub-processors? Ask for the register and the data processing agreement before any access is granted.
7. Who is responsible for realising the benefits? The original business case only works if someone is ensuring value is being achieved at scale.

Our own answers are on the page that sets out how we use AI. Check them.

The Engine delivers the volume. People supply the judgement and the control. The value comes from joining them.