Not a chatbot in the corner.
We help you work out where AI is worth using in your organisation and where it is not, whatever your sector. We review what you already have or are thinking of buying. When something is worth building, we build it with your team, hand it over and explain how it works, and a person signs off anything that matters.
A line where people stay in charge
Click or tap IN to add an awkward oneREAD pulls out what matters, SORT decides where each document goes and ANSWER drafts the reply. Anything they are not sure about goes to the tray for a person, with the reason on it, and the person decides. Click or tap IN to drop in an awkward one, or a robot to see what it is doing.
We trained a small model for one job.
Our free website check finds the problems. This model reads the report and writes the fix plan: what to do, why it matters, who does it and how long it takes. It is a small open model with an adapter we trained ourselves on 345 real small business websites, and it runs on Cloudflare.
0 made up problems
Across 59 sites it never told anyone to fix something the check had not found. The same base model, given careful instructions and an example, made up 71.
98% of the right fixes
It picked the fixes that matter most nearly every time. The instructed model managed 48%.
81% in the right order
Most important first, the way we would do them. The instructed model got the order right on 3% of sites.
Scored against 59 plans checked by hand. Training taught it our rules for what to fix first. It only knows what the check found, which is the point. How we built it.
What the check found
The model reads these and nothing else.
See the full report to check it against what was found.
Advice first, then building alongside you.
- Working out where AI helps and where it does not
- Reviewing an AI tool before you buy it
- Reviewing an AI project you already have
- Agents built with your team, with a person signing off what they do
- Agents that use your own systems, within limits you set
- MCP servers, so an assistant can work with your software
- Picking the right model, and swapping it later
- Answering from your own documents, with the sources shown
- Reading documents and pulling the data out
- Checking invoices against purchase orders, with exceptions sent to a person
- Sorting an inbox, enquiries, referrals or applications
- Drafting replies in your own tone, for a person to send
- Summarising long threads and reports
- Evals, so you can see whether it is any good
- Confidence scores and review points for your team
- Spend caps, logging and cost reporting
- Help answering the AI questions from a client, funder or regulator
- An honest list of what not to automate
In your cloud, or on a box in your office.
In your own cloud
The usual answer. It runs on your infrastructure, with the keys and the spend caps in your account rather than ours, so you can see and change everything it does.
On a machine in your building
A small always on machine, a Mac Mini or similar, sitting on your own network. It can reach systems that were never meant to face the internet, and the data stays in the building. We help your team set it up and understand it.
Somewhere in between
Most setups end up split. The heavy model calls go out and the sensitive parts stay in. We go through which is which with you before anything is built.
Something your team understands and controls.
People make the decisions
You agree the rules for what it may do on its own. Anything outside them, or anything it is not sure about, goes to a person with the reason attached.
You can check the working
Every answer shows where it came from, so somebody can check it in seconds rather than take it on trust.
It is yours to run
We hand over the code, the prompts and the notes, and walk your team through them. We also measure what it saves against what it costs to run, and say so when a part of it is not worth keeping.
Want a second opinion?
Tell us what your team does by hand every day, or which AI tool you are weighing up. We will look at it with you and give you a straight view of where AI would help, where it would not, and what building it together would involve.
Or email us at
hello@moonlathetechnologies.com