What an AI agent can realistically do for a UK business
An AI agent is software that can read a request, decide what to do next and use your systems to get it done. That is useful for a narrow set of jobs and unhelpful for others, so the value comes almost entirely from choosing the right job first.
By Saleem Ahmed · · 6 min read
Last updated:
Jobs agents handle well
The pattern is repetitive work with clear inputs, where a person still signs off on anything consequential.
- Answering the same product or service questions from documented information
- Sorting and summarising enquiries so a human replies faster
- Drafting responses that a person edits and sends
- Pulling information out of documents into a consistent structure
- Triaging support tickets to the right queue with a suggested next step
Where agents go wrong
Almost every disappointing project fails for the same three reasons: the agent was given no reliable source of truth, it was allowed to act without review, or nobody defined what a good answer looks like.
Fixing those is design work rather than model work. Ground the agent in your own content, keep it inside a narrow remit, log everything it does, and make the handover to a person obvious rather than hidden.
How to start small
Pick one process, measure how long it takes today, and set a target. Run the agent alongside the current process for a fortnight before it replaces anything. If it cannot beat the manual route on that one task, a bigger rollout will not fix it.
About the author
Saleem Ahmed, Director of CodeCue Technologies
With an MSc in Project Management and degrees in Computer Science, Saleem has led a 40+ person in-house team delivering software, mobile and AI projects for international clients since 2018.
View profile