AI consultancy for businesses that want results, not a strategy document
We find the two or three places where AI would save real time in your business and prove each one on your own data. Then, as a software engineering company, we build it into the systems you already use.
What an engagement produces
If you run a small or mid-sized business, you have probably been told more than once that you should be using AI. The useful question is narrower: which tasks, and would it pay. A lot of AI consultancy answers with a strategy document. Ours answers with three things you can act on.
The first is a shortlist: the two or three tasks in your business worth automating, each with an estimate of the time it takes today and how much of that could be saved. The second is a prototype of the strongest candidate, tested on your own documents or cases, with the results written up and the mistakes included. The third is a costed plan: a fixed price to build the feature properly, an estimate of what it would cost to run, and our opinion on whether it is worth doing.
Sometimes that opinion is no. Finding that out from a small test costs far less than finding it out from a finished project.
Where AI tends to pay off
In ordinary businesses the same four kinds of work come up repeatedly:
- Reading documents and emails into systems. Orders, invoices, forms and enquiries arrive as PDFs or emails, and someone types them in.
- Answering questions from your own documents. The answer is in a policy, a manual, a contract or an old job file, and finding it takes too long.
- Drafting routine replies and reports. The same kind of reply, quote or summary is written from scratch each time.
- Sorting and routing incoming work. Someone reads everything that arrives in a shared inbox and decides who should deal with it.
These tasks are frequent, they involve text, and a person can check the result quickly.
Where it does not
AI is a poor fit for anything that must be right every time with nobody checking. Language models make mistakes, and they sound just as sure of themselves when they do. A calculation that feeds an invoice needs ordinary software with fixed rules.
It also does not mend a process that is broken for other reasons. If staff re-type data because two systems do not share it, the fix is to connect the systems. We will tell you when that is the case.
How we work
AI automation for a small business does not need a large programme. We start with one task, collect a sample of real examples, build a prototype and measure how often it gets them right.
Whatever is built afterwards keeps a person in control. Extracted data is confirmed before it is saved, drafts are edited before they go out, and anything the model is unsure about goes to someone to handle by hand. We then compare the new way with the old and decide with you whether to extend to a second task or stop.
Data protection
Where your data goes is decided and written down before any of it is used. Models can be used through services that keep data in the UK or EU and do not train on it. The written data flow sets out what is sent, which service processes it and what is kept, so it can be reviewed before work starts.
Consultancy and build
Our page on AI for existing systems is about the build: how a feature is fitted into software you already run. This page is about the step before, deciding what is worth building at all. If you already know which task you want automated, start there.
You will see similar work advertised as AI automation or AI integration services. Whatever the name, the test is whether a named task takes less time afterwards. We are a software company, founded in 2012, that maintains and builds business systems and uses AI tools in its own development work. The build that follows the consultancy is done at a fixed price.
Talk to us
Describe the system and what you need. You will hear back from someone who can answer technical questions.
Discuss where AI would pay 0800 433 7990What the engagement includes
- A look at how the work is done
- We talk to the people doing the work and find where their time goes on reading, re-typing, searching and drafting.
- A shortlist of tasks worth automating
- Two or three candidates, each with an estimate of the time it takes today and how much of that could be saved.
- A prototype tested on real examples
- The strongest candidate is built as a working prototype and run against a sample of your own documents or cases.
- Measured results
- You see how often the prototype was right, where it went wrong and what a person would still need to check.
- A written data flow
- What information is sent where, which service processes it and what is kept, agreed before any real data is used.
- A costed plan
- A fixed price to build the feature into your system, an estimate of the running cost and our view on whether it is worth doing.
From first conversation to a working feature
First conversation
You describe the business and where the time goes. We say plainly whether we think AI is likely to help.
Shortlist
We look at the work in detail and come back with two or three tasks, each with an estimate of the time it could save.
Prototype and test
One task is built as a prototype and measured on a sample of your real examples, with the failures shown alongside the successes.
Costed plan
If the results justify it, you get a fixed price to build the feature into the system your staff already use.
Build, then extend or stop
We build it with a person approving the output, compare it with the old way, and decide with you whether a second task follows.
A good fit when
- You keep being told the business should be using AI and want to know where it would pay.
- Staff spend hours each week reading documents, re-keying information or writing the same kind of reply.
- You want evidence from your own data before committing to a build.
- You want the advice to come from the people who would build the result.
Probably not for you if
- You want a strategy document or a maturity assessment for the board; we produce working software and a plan for the next piece.
- The task has to be right every time with nobody checking; current AI models are not reliable enough for that.
- The process is broken for reasons that have nothing to do with AI. That should be fixed first, and we will say so.
Questions we are asked
What does an AI consultant actually do?
A useful one looks at how work is done in your business, identifies the tasks where AI could do the first pass, and tests that on real examples before recommending anything. In our case the same people then build the feature into your systems, so the advice has to stand up in practice.
How much does AI consultancy cost?
It depends on the scope: how many areas of the business we look at and how involved the test is. The first stage is a small, fixed-price piece of work, with the price agreed before we start. A build only follows if the test results justify one, and it is priced separately.
Do we need to tidy up our data first?
Usually not. Most of the tasks where AI helps involve documents, emails and notes as they already are, and untidy text is something language models cope with well. The exception is documents that are out of date or contradict each other, because answers drawn from them will be wrong too. If the test shows a specific gap is holding results back, we will tell you what needs fixing.
Which AI models do you use?
We choose per task and are not tied to one vendor. Models differ in what they are good at, in speed and in cost, so the test stage compares the options on your own examples. We build features so that the model can be swapped later without rewriting the rest.
Is our data safe?
Where your data goes is decided and written down before any of it is used. Models can be used through services that keep data in the UK or EU and do not train on it. The test is run on a sample, and you choose what goes into that sample.
Tell us about your system
Say what it does, what it is built on and what is worrying you. We will reply with what we would look at first and whether we are the right people to help.