AI integration for UK business software
- First call
- Free
- Builds from
- £1,999
- Review step
- Always
When you do not need AI
A lot of AI projects should be a filter, a template or a tidy database. Those are cheaper, and they are right every time.
Skip it when
- A rule gives the correct answer every time.
- There are only a few items a week. A person is quicker.
- A wrong answer would be costly and nobody will check the output.
- Your data is scattered across systems. Join those up first.
It earns its place when
- Staff spend hours reading, re-typing or summarising.
- The input is messy text or scans that rules cannot parse.
- A person stays in the loop to approve the result.
- A good draft now is worth more than a perfect one tomorrow.
What we build
Every item here is a feature inside software you already use or are having built. None of them is a standalone chatbot.
- Document extraction
- Invoices, delivery notes, application forms and scanned PDFs turned into records, with a screen where a person checks them before they are saved.
- Search over your own documents
- Staff ask a question in plain English and get an answer from your manuals, policies or past jobs, with a link to the source so it can be checked.
- Drafted replies and summaries
- A first draft of a quote, an email or a case summary written from your records. A person edits it and presses send.
- Sorting incoming work
- Emails, enquiries and tickets labelled by type and urgency, then sent to the right queue.
- Assistants inside your software
- A help box that can look up an order or a booking, limited to what that user is allowed to see.
How it works and how long it takes
The difference from ordinary software is that we measure before we build.
A free call
30 minutes
You describe the task and how much of it there is. We tell you if a plain rule would do it.
Proposal with your data
No charge
We run a sample of your real documents or messages through a model and report how often it is right. You also get a written proposal with the price and timeline, agreed before the build starts.
Build
4 to 10 weeks
The feature, the review screen, a fallback for when the model is unsure, and a log of what it did.
Tune
After launch
We look at real results with you and adjust prompts and rules on a support plan.
What we build it with
We do not train our own models, and these jobs do not need us to. The engineering is in everything around the model.
- A hosted language model
- Called through an API. We pick the provider per project and keep the code able to switch.
- Laravel
- Handles the data path, permissions, queues and logging, so the model only sees what it should.
- Custom or Filament admin dashboard
- For the review screens where staff approve, correct or reject what the model produced.
- MySQL and AWS
- Your records stay in your own database. Only the text needed for a task is sent to the model.
What it costs
Our fee is priced on your requirements, so the figures here are starting points. The running cost of the model is separate and depends on how much you use it.
One AI workflow, from
£1,999
A single feature with accounts, admin and a review screen, usually built in 4 to 5 weeks. Before you commit, we test a sample of your own material, so you know whether the model is accurate enough and what it would cost to run.
- AI inside a larger productRoles, an API and up to three integrations. 10 to 12 weeks.
- From £4,999
- EnterpriseMulti-tenant accounts and a security review. Quoted after a call.
- From £9,599
- Growth supportTen hours a month for tuning and changes.
- £599/mo
Model providers bill per use, and that bill goes to you directly. We estimate it in the written proposal so there are no surprises. Support hours can also be bought separately, from £50 per hour, as needed, without a monthly plan.
Working with us from anywhere in the UK or worldwide
AI work needs your data far more than it needs your presence. A haulage office in Carlisle or a surveying practice in Devon can send a sample export, agree with us in writing what may be shared, and see results on a Google Meet call the following week. Small offices that handle a lot of paperwork with few people are often the ones with most to gain, wherever they are.
- Calls
- Google Meet
- Team
- London, UK
- info@boolet.co.uk
- Phone
- +44 7413 564737
Questions buyers ask
Straight answers. If yours is not here, email us.
- Is our data used to train the AI model?
- We only use providers whose business terms state that they do not train on data sent through their API, and we check those terms for the provider we choose. Which provider is used, and exactly what is sent to it, is written into the proposal before the build starts.
- What does an AI feature cost to run each month?
- Providers charge by the amount of text processed, so it depends on volume. We measure a sample while writing the proposal and give you an estimate per document or per message. You pay the provider directly, and we can set spending limits.
- What happens when the AI gets something wrong?
- It will, sometimes. That is why every feature we build has a review step, a fallback for low-confidence results and a log. The sample test tells you the error rate on your own data, so you can decide whether it is acceptable before building.
- Can you add AI to software you did not build?
- Often, yes. We need access to the code or an API into the system. We read it first and tell you whether the feature can be added cleanly or the data needs work beforehand.
- Do we need a large amount of data?
- No. We are not training a model, so there is no need for thousands of examples. A sample of real documents or messages that shows the usual cases and the awkward ones is enough to test with.
Tell us the task you want off your desk
On a 30-minute call we will say whether AI is the right tool for it, or whether something plainer would do.