AI that takes the boring half of your week
Most of what AI is good at is dull. Reading invoices, sorting the inbox, writing the same reply for the hundredth time. That's the work we automate, and it pays for itself the fastest.
Where it earns its keep
The jobs below are the ones where the time saved turns up quickly, because the work is repetitive and there's a lot of it.
Enquiry assistants
An assistant on your site that answers the questions your team answers all day, asks a couple of sensible follow-ups, and passes a real summary to a person.
Document and data work
Invoices, forms, referrals, application packs. Pull the fields out and put them where they belong instead of retyping them.
Inbox and process triage
Sort what arrives, send it to the right person, draft the reply for someone to check. A human still presses send.
Search across your own material
Ask a plain question of your own documents, policies and history, and get an answer with the source attached to it.
First drafts
Reports, funding updates, knowledge base articles. Something to edit beats a blank page on a Friday afternoon.
Reporting nobody has to build
Weekly summaries and alerts written from your own data, in words your board or your trustees will read.
How an AI project runs here
It starts with the tedious job you want gone, and the technology is whatever turns out to fit it.
Find the job
A short look at how you work now. We come back with the two or three places this pays for itself, and the places where it won't.
Build one small thing
One contained piece of real work on your real data, with a result you can measure in weeks. Fixed scope, and no long contract to get out of.
Put it where the work happens
Wired into the tools your team already uses, with limits, logging, and someone who can turn it off.
Keep it working
Models change and so does your business. We keep checking it still does the thing it did on day one.
What AI will not do for you
Worth knowing before you spend anything.
Your data is not training material
Where we use Anthropic, its commercial terms don't permit your inputs to be used for training, and we'll tell you which provider your own project uses. If you handle records where that isn't enough, say so early and we'll design around it.
It does not replace your people
It takes the repetitive half of a job. The judgement, the awkward phone call and the decision stay with a person.
Sometimes the answer is not AI
A lot of what looks like an AI problem is two systems that don't talk to each other. We'll say so when that's what you've got, and quote for the plumbing instead.
Questions people ask before starting
Straight answers about scope, timing and what working together looks like.
What does an AI project cost?
The first piece of work is fixed scope and fixed price, quoted after we've looked at the job rather than before. Running costs are separate, and you'll see the estimate before anything gets built.
Will it make things up?
A language model can. That's why we limit what it's allowed to answer from and have it say it doesn't know rather than guess, which is how we build our own assistants. Where an answer comes out of your own documents we also make it show you the source, so you can check it.
Who is responsible if it gets something wrong?
You are, which is why we design these so the AI drafts and a person sends. Anything that leaves your organisation gets checked by a human first unless you specifically ask for something different.
Do you need our data to build it?
Usually a sample, under whatever agreement you need in place first. We can build against test data and swap in the real thing at the end if that suits you better.
What if AI is the wrong answer?
Then we quote for the thing that does work instead. Plumbing between two systems is cheaper than a model, and it breaks in fewer ways.
Where would this actually help you?
Tell us what eats the most time in your week, and we'll work out whether AI is the cheapest way to fix it.