somai is the context engine that connects the intelligence already
distributed across your business, giving people and AI the context to
understand, decide and act.
You control what AI can reach, what it is allowed to do, and you can
see why it answered the way it did.
Your data stays yours.
somai works with your existing information without taking ownership of it.
Your systems stay in place.
Connect the systems and tools you already use. No rip-and-replace.
Your rules govern AI.
Set the permissions, policies and boundaries that determine what AI and agents can see and do.
Every insight has context.
Understand where an answer came from, what it is based on and what might be missing.
We do not publish client names, logos or quotes. The engagements
themselves are set out in full, with what was installed and what the
team was left running. Where it’s proven →
§ 05How it lands
How it lands
01 - Blueprint.
We start with a discovery call. Next comes the question audit. We collect about two hundred actual questions from your business, and we tell you which ones a company brain answers well, which need a different tool, and which nothing can answer until the outstanding work has been completed. You get that in writing before you commit to anything.
02 - Build.
The brain is installed: your knowledge organised, your systems connected, and relevant, genuine answers being provided. Engagements are bounded, with their duration measured in weeks. Senior people, working with your team.
03 - Handover.
Your team is in charge of it: the files, the connections, and the knowledge to run it without us. Ownership is the design, not a provision.
04 - Grow.
Add blocks. Each block causes the brain to carry out one task - the accounts that you serve, the cash that you are owed, the obligations that you have. Every block reads the same brain, so each one you add makes the others sharper.
“We have a shared drive and an AI copilot. Isn’t that a company brain?”
Not at all - and this is evident when the first difficult question arises. A copilot on a shared drive searches whatever is stored: drafts, duplicates, the version that was withdrawn in March. A company brain holds what someone has decided is true, with an owner and a source on every entry. One is a search box over your mess. The other is a single source of truth.
“Who can see what?”
You decide, and the brain enforces it. Who may see what is built into the system, not into a polite instruction the AI is asked to follow. And some things should never go in at all; we agree that list before anything is captured, not after.
“Are we too small for this?”
The honest test is not headcount. It is whether the same questions get answered from scratch every week, whether knowledge leaves when people do, and whether AI tools you’ve paid for have quietly disappointed. If none of that is true yet, you don’t need us - and we’ll say so.
“What happens in week one?”
The twenty answers your business needs most, written down properly and connected. Not an import, not a crawl. It is deliberately unglamorous, and it is the step that decides whether everything after it works.
“Which AI tools does it work with?”
The ones you already use, and the ones you’ll switch to next year. The connection layer speaks the open standard AI assistants understand, so the brain outlives any one tool choice.
“Where does it run?”
In the cloud, with a provider, or on your own servers. Either way it is yours: plain files, an index that rebuilds from them, and nothing trapped in anyone’s platform - ours included.
§ 07Studio
A studio of senior operators. No juniors, no offshore, no associates in
the room.
somai is a team who have run the kinds of companies we work in:
heads of operations, engineering leaders, CFOs, chiefs of staff,
and founders - most of the bench with scars earned under regulators,
auditors and risk committees.
We take the principle that the studio installs and the client
operates seriously. We do not sell retainers, managed services, or
ongoing seats. Our job is to make sure your team runs it without us.
Automation makes work faster. It does not ask whether the work should happen - and in most organisations the larger cost is not work done slowly, but work that should never have started.
A language model is brilliant at turning many facts into one clean answer, and unreliable at turning too little into the truth. That asymmetry - not the model's cleverness - is the real constraint, and it is why grounding matters more than scale.
The market defines ai-native by birth certificate - the firms born with the model. The truer definition is structural, not chronological, and it is fully available to the company you already run.
How much of your company’s intelligence can you actually use?
Your company contains more intelligence than you can currently see.
The AI-Native Diagnostic measures one thing: how much of how your
company runs a model can actually reach and act on today. Not how
much AI you already use.
the diagnostic, in outline25 questions · 10 minutes
What your company knows
Your market
How the work gets done
How you decide
How you stay in control
Your result is your own. Nothing is filled in until you take it.
An audit, not a quiz.
Twenty-five plain-English questions about your business. No technical
knowledge needed, one question at a time, and your progress saves as
you go. It reads the five systems every company runs on and names the
one holding the most value out of a model’s reach - your biggest
block.
Then it puts a number on the prize: what an AI-native rebuild could
be worth for a firm your size, modelled from public benchmarks. That
is a population estimate, not a forecast for your business. You end
with a one-page brief, emailed to you and yours to keep. No licence,
no retainer hook.
The diagnostic is itself AI-native: a model reads your situation in
your own words and writes your brief, while a calibrated instrument
keeps the score honest. Private by design - banded inputs only, never
your raw financials.
