Find Your
First AI Win
How to know if your business should be building AI tools
Pick one repeated job
Test a small version
Compare the result
You get evidence before starting a wider programme.
Where AI earns its keep
AI is exciting and we’re here to help you work out how to build useful stuff that saves money and has a clear return.
AI is more useful for mundane, repeated tasks that happen every day, not necessarily the biggest and flashiest use cases
It’s easier to use AI effectively if you have the right data
You can use AI in places where you already use software
We can help you work out which AI use cases to prioritise, how big a project each one is and how much they’ll cost to build, how much they’ll cost to run and how long until they payback
We can also tell you if AI isn’t appropriate and it’s better to look at cheaper options
Who we work well with
We work well with established businesses in the UK with a turnover in excess of £10m.
You’ve got people doing the same tasks on repeat
There’s a named person in charge of the area of operations and a metric you want to improve
You may already have tried building chatbots that people didn’t ask for or no-code tools that stopped after the demo. That’s OK. That will help us understand the appetite for AI.
What fails and what works
The typical AI project that doesn’t work is a website chatbot that no-one asked for or a licence for an AI software platform costing 6-figures, implemented by the IT department only and chosen by them, not the rest of the organisation, using faked data because the real data isn’t ready, with no owner who cares whether it gets used everyday and has a P&L, and with no metric that’s going to change as a result.
The typical AI project that works is a humble AI that saves money by helping people with a task they do everyday, implemented with the team that does the task, using data that’s already captured as part of doing the task, inside software that’s already used for the task, with one metric it’s trying to move and an owner who cares about moving it, knows how much development is required to make the AI and knows when it pays back.
How we choose the first one
- We score them on five criteria, which are: Can we see the effect on P&L, can we build a version to try quickly, is the data already there, can humans check it’s working, is there a champion of it. We hand you this so you can prioritise future ideas yourself.
- We look for use cases that can be tried quickly , have data that’s already there and a named owner in Operations. We prioritise ideas that Finance can see the bottom line impact from and IT can get data from. We love ideas where it’s important to be safe and accurate so a person is double-checking the AI’s suggestions.
- We pick one to build using realistic examples from Operations. We agree with Operations, Finance and IT who’s going to use it. We also agree what number to start with so we know whether it’s helped, how long to try it for, and what number would be good/bad enough to continue or stop. We build it using software Operations already have access to.
- We hand you the list of AI ideas, ranked, plus the baseline numbers and process to try more. We’re done. You’re good to go.
What it costs
We like to agree fees for the first phase, exploring ideas and agreeing a shortlist, separately to the second, building it to try. At the end of phase 1 we’ll tell you the development and running costs for each idea as well as the number we think it’ll improve and by how much. This should tell you when each idea paysback. We’ll show you how to calculate this for other ideas in future.
Book a call to find out more