Prism House home

AI for small and mid-size businesses

Your data, your advantage.

Prism House builds AI that adds to what your company owns. I work with businesses in the UK and Australia, using whichever models fit the job.

Book a callSee what I do

  • Fixed fee for the first step
  • You own what I build

What I do

Three ways in

Start with a conversation about your team, a product built for your company, or a free Mac app.

Where AI earns its keep

Typical first jobs

Most businesses start with work that is repetitive and easy to check. These are examples of the kind of job a map tends to surface, not promises about your business.

A morning brief

What changed overnight in your inbox and order system, summarised before the team starts.

A weekly owners' summary

The numbers and decisions of the week in one page, drawn from the systems you already use.

Finding things in your own records

Plain-English search across documents and catalogues that are too sensitive for a public AI tool.

First drafts of routine documents

Quotes and routine replies drafted by AI and checked by the person who owns them.

How an engagement runs

Small steps, each one proven

We agree what success looks like before any results come in, and I only scale what works. You can stop after any step and keep everything produced so far.

The three steps in detail

  1. 1

    About two weeks

    AI Opportunity Map

    Where your team's time goes, and which jobs AI could take on, ranked by value.

  2. 2

    Three to four weeks

    Gated pilot

    One question and one pass mark, agreed in writing before any results exist.

  3. 3

    Monthly

    Operator retainer

    AI built into one department at a time, under a written permission ladder.

The data promise

Your data stays yours

Light passes through. Your data doesn't.

I choose the model that fits the data. Sensitive work runs on private open-source models on your own infrastructure, and the rest can use enterprise cloud services that don't train on your data. Often one solution combines several.

You own what gets built, and your data never becomes someone else's training material.

Inside your systems Your dataRecords and documents Private modelOpen source, runs on your side Cloud modelEnterprise terms,no training
Runway menu bar card, sample data
Runway in the menu bar, shown with sample data.

Built in-house, shared for free

See how I build before you hire me

Runway is a free Mac app for people who work in Claude Code. It sits in your menu bar and shows how far this week's Claude allowance will go at your current pace.

I built it for my own work first. It follows the same rules I bring to client work: it reads only what it needs, and every number says whether it is measured or estimated.

Get Runway for Mac

Clients

Who I work with

Clients so far include a music-tech start-up in the UK and an educational toys company. Owner-led businesses where people's time is the constraint and the data is too valuable for public AI tools.

Run by a founder with nine years in finance and operations

Hi, I'm Olly

I spent nine years running finance and operations, including a seed raise. I now build AI for owner-led businesses, and I start where finance taught me to start, with the cost of a process and the controls around it.

I run my own business on the same kind of system I build for clients. An AI agent can propose a change, but only a person can approve it.

Book a call with meAbout Olly

Common questions

Which AI models do you use?

Whichever fits the job and the data. I am provider-agnostic, and one solution often combines a private open-source model with a cloud service.

Who owns what you build?

You do, along with your data.

What if the pilot doesn't work?

Then you have a clear answer. The pass mark is agreed before any results exist, a no costs a reduced final balance, and you keep the findings.

More questions

Got a job your team shouldn't be doing by hand?

Thirty minutes, no preparation needed. I'll tell you honestly whether AI can help.

Book a 30 minute call