In short
Fabio Armellini, who is building his role as a Fractional AI Orchestrator, has finished the Decision Dashboard 2.0 programme and now builds dashboards and apps in minutes. The new obstacle is different: he still spends too many hours at the PC steering the AI. An ABC analysis on about one million euros of purchases, which would once have taken six weeks, was closed in two weeks, with about one hundred hours of work with AI. The next step is not a better tool. It is moving from user to orchestrator of agents that work while he does strategy.
Key points
- The 12-session programme removed repetitive manual work: dashboards and apps in minutes.
- The next bottleneck is the time spent at the PC steering the AI, not manual work itself.
- An ABC analysis on about 1 million euros of purchases went from about 6 weeks to 2 weeks, with about 100 hours of work with AI.
- The declared 12-month goal: one day less out of five and +20% revenue, which means about +50% revenue per day worked (1.2 / 0.8 = 1.5).
- The next step is a small team of persistent agents with one person deciding, not one more tool.
Fabio Armellini is a fractional manager, and he is building the role of Fractional AI Orchestrator. A few months ago his problem was clear: half his time went on collecting data. We solved it with the Decision Dashboard 2.0. His case study, which you can find here, tells how.
This article tells what happened next.
The programme is over, the problem is not
Twelve sessions in about four and a half months. First the AI Mirror, then the Dashboard. Along the way Fabio moved from Claude Chat to Cowork and then to Claude Code.
The most visible result is this: today he builds dashboards and apps in minutes. Things that once took days, or never got started.
In the recap call on 2 October, Fabio put it very clearly: the bottleneck is no longer manual work. It is manual work with AI.
What that means, in numbers
An ABC analysis on about one million euros of purchases. Before the programme it would have taken about six weeks. With his tools he closed it in two weeks.
Two weeks is a good result. But it took about one hundred hours of work with AI: prompt, corrections, checks, prompt again. He did the work himself, by hand, with the AI in front of him.
The problem has moved. It is no longer “how long does it take me to produce”. It is “how many hours do I spend steering the tool”.
From user to orchestrator
What Fabio wants now is a change of role. Not a user who steers the AI, but an orchestrator: a few persistent agents working on his behalf. They access his files and his knowledge base and send him a periodic brief, like a status page to read instead of rereading chats.
The three roles he has in mind
- Planner: manages the agenda and communication with clients and stakeholders.
- Analyst: works on the same data through several lenses.
- Coordinator: turns the analyses into actions that can be carried out in the field.
He stays on strategy and on removing constraints.
The twelve-month goal
One day less out of five and 20% more revenue. The sum, as Fabio worked it out: working 4 days out of 5 is worth 0.8, revenue is worth 1.2, and 1.2 / 0.8 = 1.5. In practice, about +50% revenue per day worked. He puts it in a line of his own:
I work one day a week on the client, but it is as if I worked five.
These are goals he declared. They are not results yet.
What Fabio says on 22 October
Twenty days after the recap call, Fabio thought about the next step. Three lines of his own, from the call of 22 October 2026.
On trusting the AI’s results:
“I trust the result, but then you have to explain it to whoever sees the result” — Fabio Armellini, call of 22 October 2026 (translated from Italian)
To be an orchestrator, receiving a brief is not enough. You need to be able to explain the reasoning behind the result, not just the result.
On the size of the system:
“I wouldn’t want a system that is oversized compared to what I actually use” — Fabio Armellini, same call
The next step has to be sized on real use, not on what would be possible.
On how to get there:
“learning to drive the machine could also be of value” — Fabio Armellini, same call
He does not only want to receive a ready-made system. He wants to learn to drive it.
What I take from it
When a professional closes their first big bottleneck, they find another right after. That is normal, and it means the first one is really solved.
For anyone working with several clients, the second one is usually the same: context. As long as it lives in your head, you are the system. Every time you have to explain again:
- who the client is;
- what has already been decided;
- what the AI must not do.
That is why the work we do now starts from memory on files, keyword skills and rules around the tools. If you recognise yourself in this step, you can see how we build it on the Second Brain for Fractional Managers page.
The figures in this article are those declared by Fabio Armellini, Fractional AI Orchestrator, in the call of 2 October 2026. The +50% calculation is Fabio’s own. He repeated it aloud in the call of 22 October and read and approved the original text before publication. The lines in the 22 October section are his, reported from the call.
Frequently asked questions
What did Fabio do in the programme with AIEH?
Twelve sessions over about four and a half months: first the AI Mirror, then the Decision Dashboard 2.0. Along the way he moved from Claude Chat to Cowork and then to Claude Code.
What is the bottleneck after manual work?
Manual work with AI. Fabio no longer loses hours collecting data, but he still spends many steering the tools step by step. That is why the goal is to hand recurring tasks to agents that work on their own and report back with a brief.
Are these figures verified?
They are the figures Fabio declared in the recap call of 2 October 2026. They are results declared by the client, not an independent measurement.