Where to Start with AI in Your SME
A practical 3-step playbook for picking your first AI use case, choosing a tool, and proving value in 30 days.
TL;DR
Pick one repetitive workflow that costs you measurable time each week. Try an off-the-shelf AI assistant on it for 30 days. Track hours saved. That's your first AI project.
Quick answer
Start with one repetitive workflow — customer FAQs, sales follow-ups, or weekly reports — that costs your team measurable time. Use an off-the-shelf AI assistant for 30 days, measure hours saved, and only then consider custom integration. This keeps risk near zero and gives leadership a concrete ROI number to build on.
Key takeaways
- Pick a workflow, not a technology — let the use case choose the tool.
- Measure baseline hours before you start, otherwise you can't prove value.
- Start with an off-the-shelf assistant before building anything custom.
- 30 days is enough to either prove or kill a use case.
- Train two people first, then expand once they're confident.
Most SMEs hear about AI and immediately reach for the wrong question: 'Which AI tool should we use?' The right first question is 'Which workflow do we already lose hours to every week?' Pick that, then choose a tool to fit it.
Step 1 — Find your costliest repeating workflow
Ask each team lead: 'What do you do every week that you wish you didn't have to?' Common answers: answering the same customer questions, writing weekly reports, summarizing long emails, qualifying inbound leads.
Step 2 — Try an off-the-shelf assistant for 30 days
Don't build. Start with ChatGPT, Claude, or a vertical SaaS that already exists. The point of the first 30 days is to learn whether AI can do this job at all — not to build infrastructure.
Step 3 — Measure and decide
At day 30, compare hours spent before vs after. If you saved >20%, expand. If you saved <10%, kill it and try a different workflow. Either outcome is a win — you bought certainty cheaply.
Frequently asked questions
- How long should the first AI pilot run?
- 30 days. Long enough to learn whether AI fits the workflow, short enough to abandon cheaply if it doesn't.
- Should we build custom AI or use existing tools first?
- Existing tools first. Always. Custom only makes sense once an off-the-shelf assistant has proven the use case and you've hit a clear limit on what it can do.