Portolan™ · the AI roadmap session

One process. Measured in six weeks.

Most AI programmes end as a slide deck. Portolan ends with before-and-after numbers on one process you run every day, and a named owner who keeps them.

Portolan session: half a day, up to 15 people, on site or remote, in English or French.

The problem

The pilot worked. Nothing changed.

You have run the workshops. You have a strategy deck, maybe a pilot that impressed the steering committee. Ask what moved in operations and nobody can give you a number.

The gap is rarely the model. It is the missing baseline, the missing owner, and the absence of a date on which anyone checks.

Portolan fixes those three things on one process before it touches a second.

The fix

Measure first. Change one thing. Measure again.

Portolan puts back the three things that are missing, on one process, before anyone talks about scaling.

A baseline

Numbers before any change

Week 1 measures the process as it runs today: hours per case, error rate, end-to-end delay. Nothing is redesigned until those numbers exist.

An owner

A name from the business

Not IT, not a vendor, not us. One manager who runs the process, signs off the numbers and keeps it running after we leave.

A date

Week 6, in the calendar

The same measurement, the same method, on a date fixed in week 1. The result goes to your executives whether it flatters us or not.

How it works

Three steps. Each one built on your processes.

Nothing here is a template. Each step produces something that only fits your organisation, and you can stop after any of them.

01

Portolan session

Where you actually are

Half a day with your team. We map where AI already runs in your organisation, pick the one process worth measuring first, and hand you a written roadmap with the steps and the tools.

You get: your roadmap and the first process to measureHalf-day · up to 15 people · on site or remote
02

One process, measured

Baseline, redesign, remeasure

Week 1 baselines hours, errors and delay, and names a business owner. Weeks 2 to 5 redesign the process with your people. Week 6 measures it again with the same method.

You get: before-and-after numbers on your process, and its owner6 to 8 weeks
03

Scale it, or test the next bet

Retainer or evidence first

A few advisory days a month to extend what worked. Or a small internal prototyping unit, with kill rules and one cleared sandbox, to test the next idea before the budget is committed.

You get: your next processes improved, or your next ideas tested with a go or stop decisionMonthly, or about 14 weeks

What you hold at week 6

Numbers and a name. Not a report.

If the numbers did not move, you will know that too, and you will know it in six weeks rather than twelve months.

On your process
Week 1
Week 6
Hours per case
Baseline
Remeasured
Error rate
Baseline
Remeasured
End-to-end delay
Baseline
Remeasured
Signed off by a named business owner, who runs the process after we leave.

Who it is for

Built for institutions where change has to survive audit.

Banks, insurers and regulated institutions in Mauritius and across Africa. Places where a new process has to clear compliance, risk and the board before it counts.

Sessions run in English or French, on site or remote.

Who runs it

Marc Israel

Founder, Aetheis

  • 17 years at Microsoft, latterly regional CTO across 49 African countries
  • Sits on boards as an Independent Director, so he knows what a board needs to see before it approves
  • Vice-chair in the drafting of the Mauritius National AI Strategy
  • Trains boards and executive teams on AI in Mauritius and East Africa

Why “Portolan”

Charts drawn by people who made the trip.

From the late 13th century, Mediterranean pilots sailed with portolan charts. The word comes from the Italian portolano, “relating to ports”.

A portolan ignored what a sailor did not need. No inland kingdoms, no decoration for its own sake. Coastlines, ports, hazards, and lines of compass bearing built up from voyages actually sailed. To reach the next port, a pilot drew a line on the chart and held the heading.

The session works the same way. Where you actually are, the next port worth reaching, and the heading to get there, drawn from evidence on your own processes rather than from someone’s vision of 2030.

How it compares

Built to decide, and to leave you independent.

A Portolan session is small and narrow on purpose: one process, chosen from your own with your own indicators, with an owner and a date on which someone measures it. The goal is a process your people run without us.

Two other kinds of help do different jobs. Card-based AI workshops are good at getting a room, sometimes two hundred people, to understand AI. Firms that build AI agents are good at building. Their model, though, is recurring: the more of your operations run on their platform, the better their year. Ours is the opposite.

Independence also limits what we can recommend. With no platform to sell, we can only put forward solutions that survive a change of vendor: your data stays yours, components can be swapped, and no single supplier holds the process hostage.

Portolan
Card-based AI workshop
Agent-building firm
Starting point
Your processes, your volumes, your error costs
Card decks, the same for every organisation
Their platform and the agents they already build
Who does the work
12 to 15 people who run the process choose it, then redesign it
Up to 200 people build a mural in subgroups
Their engineers build; your teams receive
You leave with
One process, one named owner, numbers before and after
Shared understanding, then a synthesis weeks later
A working agent and a support contract
Proof
Hours, errors and delay, measured on your process
Use cases explored, benchmarked against others
A demo that works
Choice of technology
Only what survives a change of vendor
Outside the scope
Their stack, or the vendor they partner with
Success for them means
You run it without us
A room that understands AI
You stay on their platform

Questions

Before you book

The session

What happens during the half-day?

We map where AI already runs in your organisation, officially or not. We list the processes that could be measured and compare them on volume, cost of errors and the data you already hold. The room chooses one, and names who will own it. After the session you receive the written roadmap: the process, the baseline to take in week 1, the recommended steps and the tools.

What do you need from us beforehand?

A sponsor at executive level, a date, and three to five processes your teams complain about. We use the preparation time to tailor the session to your sector and your systems.

Who should be in the room?

The people who run the work, not only the people who sponsor it. Include someone from risk or compliance: a choice they did not see being made tends to stall later. The session works best with 12 to 15 participants.

Is this a training course?

No. Participants learn a lot, but the output is a decision, not a certificate. If your teams need AI skills first, we run that separately and say so in the roadmap.

How is this different from a card-based AI workshop?

Those workshops build shared understanding across a large group, using the same cards for every organisation, and send a synthesis of use cases a few weeks later. A Portolan session works on your own processes with the 12 to 15 people who run them, and ends on the day with one process chosen, an owner named and a date to measure. If your organisation first needs broad awareness, do that first; the Portolan session is the step where you commit to one measurable change.

How is this different from a firm that builds AI agents?

Agent-building firms are good at building, and some of your processes will need one. Their business is recurring: success means you stay on their platform. Ours is the reverse: success means your people run the process without us. That is also why we only recommend solutions that survive a change of vendor. If the roadmap says you need a builder, it says what to ask for, so you choose one on your terms.

What if no process is worth measuring?

Then the roadmap says so, and why. You will have spent half a day finding that out instead of half a year.

Logistics

Why is the first session paid?

Because it has to stand on its own. You leave with the roadmap and the process worth measuring first, whether or not you continue with us. A free workshop is a sales pitch, and you would be right to treat it as one.

Remote or on site?

Remote anywhere, on site in Mauritius. On-site sessions elsewhere in Africa are quoted with travel. The room works the same way in both cases.

Which languages?

English or French, including the written roadmap.

Data, tools and what comes next

Do you need access to our systems or customer data?

Not for the session. If you continue to the measured engagement, access is agreed with your DPO and CISO before week 1 starts.

Which AI tools do you recommend?

The ones that fit the process, your security rules and what you already pay for, and that you could replace without rebuilding everything. We have no platform to sell, so the roadmap names the options and the reason for each, including how you would switch.

Are we committed to the next step?

No. Each step stands on its own and you can stop after any of them. The measured engagement is scoped after the session, once the process is known.

Start here

Pick one process. Measure it.

Tell us who you are. You receive the session brief by email straight away, and a formal quote within one business day, ready for your procurement team.

Six weeks from now you will have a number, or a reason.

Get the session brief → Book 20 min

Portolan™ session

Get the session brief

The session brief arrives by email straight away. Your formal quote follows within one business day.

Prefer to talk first? Book 20 minutes with Marc →

Portolan™ session

Book 20 minutes with Marc

A short call to check whether a Portolan session fits your organisation. Pick a slot that suits you.

Loading the calendar…