The model

Three layers.
Always worked in order.

Saturn was named for the rings. At the centre sits the foundation a business actually runs on; around it, the operating model, and around that, the AI layer. Every engagement works from the inside out, because AI layered onto a broken foundation is theatre, and there's no interest in selling theatre.

Core

The foundation

What it covers

Data and the systems of record. The structure, cleanliness, ownership and integrity of the data estate, and the technology stack sitting on top of it: what exists, what's genuinely fit for purpose, and what should be kept, replaced or quietly retired.

The questions it answers

  • Can this business trust its own numbers?
  • What is actually running here, and what is all of it costing?
  • What has to be fixed before anything gets built on top?

What usually turns up

  • More systems than anyone can name in one sitting, several of them still being paid for
  • Two or three competing versions of the same customer or product list
  • A month-end close taking twice as long as it should, for reasons nobody has written down
Ring one

The operating model

What it covers

How the business actually runs on its systems, day to day. The ERP and core process platforms (finance, supply chain, commerce), the integration between them, and the workflows people execute every day, including the internal tools and trackers quietly holding the whole thing together.

The questions it answers

  • Do the systems fit the business, or has the business reshaped itself around the systems?
  • Where are the manual workarounds, and what are they costing in time and error?
  • Which single points of failure would actually stop trading if they disappeared tomorrow?

What usually turns up

  • A critical spreadsheet maintained by one person who hasn't taken a proper holiday in years
  • Processes designed around a system limitation that was fixed three versions ago
  • Two teams doing the same reconciliation twice, neither aware of the other
Ring two

The AI layer

What it covers

The tactical, reporting and actionable layer, where AI genuinely raises productivity on top of sound foundations. Agents, automation, AI-assisted reporting, and the practical business of making teams AI-native rather than AI-curious.

The questions it answers

  • Which AI use cases are real for this business, and which are noise?
  • What does the foundation need to look like before any of them will work?
  • How do these particular people actually change how they work: tools, practices, and the change management around both?

What usually turns up

  • Genuine enthusiasm, aimed at the wrong three use cases
  • Tools already bought and barely used, because nobody changed the workflow around them
  • One or two unglamorous, high-value automations nobody had thought to ask for
Where AI actually lands

Judgement first. Tools second. Always.

The evaluation lens

Every AI proposal, whether internal enthusiasm or vendor pitch, gets the same question: is this a genuine use case, or is it hype with a budget attached? Saying "not yet" is part of the job.

Working practices

Practical, workflow-level change: which tools fit which specific jobs, and the habits that make them stick. Deliberately unglamorous. A team drowning in meeting notes needs call capture and summarisation, not a strategy deck.

Purpose-built tools

Software can now be built quickly enough that the right answer to a broken internal workflow is often a small, purpose-made tool rather than another licence or a bigger spreadsheet. A tracker built around how a team actually works will beat a generic platform every time.

Adoption and change

Deciding what to adopt is the easy half. The harder question is whether people actually change how they work, and that is where most AI adoption quietly fails. Who is genuinely affected, what they lose as well as gain, who owns the change once the enthusiasm fades, and what has to be true for a new way of working to survive its first busy week.

Change capacity is finite. A business midway through fixing its foundations cannot absorb an adoption programme at the same time, and sequencing those honestly is part of the work.

The sequencing rule

Foundation before operating model.
Operating model before AI.

The order isn't a preference, it's a dependency. AI built on data nobody trusts produces confident answers that happen to be wrong. Automation layered over a broken process just makes the breakage faster. And a business that can't close its books in five days won't get value from an agent that reports on them.

Which means part of the work is being willing to say that AI isn't the next step yet, and to say it to someone who came looking for AI. That conversation is usually the most valuable one in the engagement.

Start with a conversation,
not a pitch.

Thirty minutes is usually enough for both sides to know whether there's a genuine fit.