Because my clients want to understand what they get. I run an AI agency from Dubai, and in more than 90 percent of my projects coaching and consulting are part of the delivery. Nobody wants a black box, and 10 automations handed over at once never get used.
I come from marketing, and there the deal used to be simple. A client gave the agency a project, trusted it to deliver, and the agency did the work as a service provider. Over the last two years that changed completely with AI. People whose companies make eight figures a year now tell me they want to learn how to use it themselves. Right now I have exactly one project where the client does not want to understand what I build. Working from Dubai for business owners as far away as Switzerland, that shift decides what I sell, how I cut a project into pieces and what a client keeps once I am done. The sections below are how I build projects around it.
Clients stopped buying a black box
The old agency model assumed that the client did not need to know how the result was made. With AI that assumption is gone. When I described my clients in an interview, the pattern was the same across all of them: everybody wants to understand it, and no one wants a black box.
That is why coaching and consulting are included in more than 90 percent of what I do. A project always starts with consulting, where I map out the strategy and improve it. Then we apply it, and after every milestone I teach them, I coach them and they ask me their questions. By the end they have their setup and their own productivity has improved. The delivery is a whole mix, and the part where the client learns is not an extra on top. It is the reason most of them hired me in the first place.
Building got cheaper, so understanding became the product
I am not a developer, and I still build very good systems, because AI does the development with me. That is exactly why I expect development itself to matter less and less for AI agencies, while the strategy, the business side, the consulting and the coaching matter more.
My own working day shows how far this has gone. As an AI agency owner in Dubai I have my Claude Code setup open 16 hours a day, and one week I had a record of 101 agents working for me at the same time. I could not say how long that would have taken me alone. When building is that fast, the bottleneck moves to the client: whether they understand what was built, whether their team uses it, and whether they can decide what comes next. An agency that only delivers the build is selling the part that keeps getting cheaper.
Ten automations at once means none of them gets used
The second reason for coaching is what happens when a client receives too much at once. Business owners are busy, and every one of my clients has many other projects running beside mine. If I delivered 10 automations to a client tomorrow, they would be overwhelmed and could not apply a single one of them.
I have seen the same thing from my years in marketing. When you run too many projects with one client at the same time and each one is only a few percent done, the client loses track of all of them. Everybody wants a quick win instead, a result they can use quickly rather than a large system they have to take in all at once. So the question for me is never only what I can build. It is how much a team can take in at once, and how to deliver in an order where every step is used before the next one lands.
Four monthly milestones with coaching in between
My answer to that is scoping by milestones. Instead of selling one big project, I break it down into four milestones over four months. The client sees what they get in the first, second, third and fourth month, with the deliverables for each one. After every month they have something they can get used to before the next part arrives.
Between the milestones are the coaching sessions, for the owner and for the team. While they learn to work with the first part, I build the next one, and then the one after that. This is the rhythm that turns an AI project from a handover into a change in how the company works. It also keeps the client in control, because each month ends with something they already use and understand before they agree to the next part.
The first month decides whether we continue
The milestone model also solves the trust problem, and without trust there is no sale at all. Many clients do not know me well enough yet to commit to the whole roadmap. So the first month is a closed project of its own, and after that month the client decides whether to go ahead with me or not.
Everything is built on one foundation, so if we continue, each milestone builds on the last one instead of starting over. The client sees the full roadmap from the first day and still gets a quick sense of success, because the first project is done very quickly. This is the part clients tell me they really love. They are never asked to trust a large plan blindly. They are asked to judge one month of real work, with coaching included, and then to decide.
Coaching is part of the delivery because understanding is what clients buy
So why do more than 90 percent of my AI projects include coaching? Because the build is no longer the scarce part. Clients want to understand their systems, they can only absorb one step at a time, and they decide on the next step once they have seen the first one work. Every milestone ends with coaching, so the team grows with the system it uses.
That holds for the business owners I work with from Dubai, and I expect consulting and coaching to matter more for every AI agency from here on. If you are planning an AI project, ask your provider for the first milestone and the session where your team learns to run it, before you ask for the full scope.