Operations · 6 min read
Where to start: sizing an AI opportunity before you buy anything
A simple method to find the process worth automating first, and to estimate what it is worth before any software is bought.
Why do most AI projects start in the wrong place?
Most businesses meet AI through a product demo. The demo is impressive, a licence is bought, and only then does anyone ask which part of the business it should fix. The result is software that is used lightly and a cost that is hard to justify.
The better order is the reverse: find the work that costs you most, then decide whether AI, ordinary automation or nothing at all is the right answer.
How do you find the right process?
Look for work with these features. The more boxes it ticks, the stronger the case.
- —It happens often: daily or weekly, not once a quarter.
- —It follows roughly the same steps each time.
- —It moves information between systems, inboxes or documents.
- —Delays cost you money: lost leads, late invoices, unhappy customers.
- —Skilled people are doing it when they could be doing something better.
How do you put a number on it?
Use three figures: volume (how many times a month), time (minutes each time) and the cost of an hour of the people doing it. Then add the cost of what goes wrong: enquiries that go cold, errors that need rework.
Worked example: 300 enquiries a month, 20 minutes of handling each, at £35 an hour, is £3,500 a month in staff time alone — £42,000 a year. If one in ten enquiries is answered too late to win, the lost revenue is usually larger than the time cost.
When is AI the wrong answer?
If the steps are fully predictable, ordinary automation is cheaper and more reliable. AI earns its place where judgement is needed on messy input: reading an email, qualifying a lead, drafting a reply. Many of our recommendations are a mix of both, with a person approving the decisions that matter.
Key takeaways
- ✓Choose the process first and the tool second.
- ✓Size the opportunity with volume, time, hourly cost and the cost of delay.
- ✓Use plain automation for predictable steps and AI only where judgement is needed.
- ✓If the yearly value is small, move on to the next process.
Frequently asked questions
How long does it take to size an AI opportunity?
For one process, an hour with the people who do the work is usually enough to get a reliable estimate. A full audit across a business takes two to four weeks.
Do I need clean data before starting?
No. Most first projects work with the emails, forms and systems you already have. Data problems tend to surface during the work and are fixed as part of it.
What is a good first AI project?
One that is frequent, well understood, and costs you money when it is slow, such as handling new enquiries, chasing invoices or preparing quotes.
Next step
See what this could be worth in your business.
Read about our AI audits work, map a process with the Blueprint tool, or talk to us directly.
