· 1 min read · Abdul Aziz Arifi
AI agents for SMEs: where they actually help — and where they don't
AI agents are sold as a fix for everything. In practice they're very good at a well-defined set of jobs and a poor fit elsewhere. Here's how we judge it when a client asks whether an AI agent makes sense for them.
Where AI agents work well
The common thread: high volume, repetitive, rule-based communication where speed matters.
- Answering calls and WhatsApp messages outside office hours
- Qualifying enquiries with the same few questions every time
- Booking, rescheduling and reminding appointments
- Following up on quotes and leads that would otherwise go cold
- Answering recurring questions about services, prices or opening hours
Where they don't belong
AI agents shouldn't make decisions that need professional judgement or carry liability: medical diagnoses, legal or financial advice, complaints that need tact and discretion. And where enquiry volume is low they rarely pay off — if your team handles a few calls a day comfortably, an agent adds more complexity than value.
Always design the handover to a human
A good agent knows its limits. Decide in advance which situations go straight to a person — and make that handover seamless, with the context already captured, so the customer never has to repeat themselves.
How to start
- Map the workflow: where are enquiries lost or delayed today?
- Pick one use case with clear volume and a clear outcome
- Connect it to the tools you already use — calendar, CRM, email
- Run it with a human in the loop first, then widen its scope
- Review real conversations regularly and adjust
The most expensive mistake
The most expensive mistake isn't the software — it's automating a broken process. If the workflow behind the agent is unclear, the agent just makes the confusion faster. That's why we start with the process, not the technology.