Silenos AIBook a Discover call, opens in a new tab

What we do

How we work

Three steps, in this order, and a straight answer at each one about whether you need the next.

Decide, build, run

Decide

We work out whether AI is the right tool before anyone writes code. Sometimes the answer is that a spreadsheet rule would do it, and we will say so.

Build

We build the thing, with the data controls in from the first day rather than added before the audit.

One supplier contract, as the model receives it

  • Contract reference
  • Renewal date
  • Notice period
  • Signatory namenot sent
  • Bank detailsnot sent
Five fields from one supplier contract. Three go to the model. The signatory name and the bank details are removed before the question leaves your systems, which is why they are struck through and marked not sent.

Run

We keep it running and we tell you when it stops being right. Models drift. Yours will.

The log of what the AI was asked and what it answered

  • can read Your compliance officer
  • can read Your IT team
  • can read Us, while we are running it
  • cannot read The company that made the model
Four people who might want the log, and whether they get it. The words say which is which, not a colour or a tick.

What we believe

A diagnosis should come before a build.

Nobody should be sold a solution before someone has honestly checked whether AI is even the right fix.

Where we go deep

The two problems we are actually deep on

Documents that have to move through a process

Unstructured content that has to be processed, routed, validated, redacted or audited at volume. The work is exception heavy, and somebody with a compliance responsibility is usually in the room.

Records that should agree and do not

Two systems that are supposed to describe the same customer, product or account, and drift apart. Or outside information that never gets joined to what you already hold.

Both look the same in retail, banking, insurance, legal and healthcare. That is what makes fitting a problem rather than an industry a claim you can actually check.

We do not say we can solve any industry's problem. That claim cannot be tested, sounds like every competitor's claim, and a buyer tends to hear it as an admission that we have not yet worked in their world.

What we are building

Four lines, each with a page of its own: Pulanfor documents, Symblon for product records, Mai for schools, andS Labs for proofs of concept. They are at four different stages, and each page says which. One of them, S Labs, is something you can buy today.

Every one of those pages carries a list of what that line does not do. Some of this has been used outside Silenos and some of it has not, and the page says so either way. We would rather write the gap down than let a buyer find it in week two of a pilot.

No lock-in

What you own, and what we keep

This is not a smaller version of the no lock-in promise. It is that promise stated exactly. A client does not actually want to own our method, they want not to be trapped, and this gives them the whole of what they were afraid of losing.

What you own

  • Your data and everything the system produces
  • Your own configuration and your own instance
  • The source, the documentation, the models and the credentials
  • A perpetual right to run all of it without us

What we keep

  • The method and the gate framework
  • The resolver logic and the matching approaches we use
  • Reusable tooling and code components
  • Whatever we already knew before we met you

What it costs

We are looking for three founding clients at a fixed fee. You would be working with us directly. Talk to us and we will tell you the number on the call.