Real-world use cases
Concrete examples of how small businesses are using AI agents day to day. Paraphrased patterns from the OpenClaw user community — your own setup can look like any of these, or nothing like them.
Family personal assistant
A household runs a personal-assistant agent on WhatsApp that researches flights and hotels, compares grocery prices across Coles, Woolworths, Aldi and Costco, and shortlists property listings against criteria like quiet streets and scenic views — judging the view straight from the listing photos. The same agent quietly tracks art-prize call-outs, university open days, and product discounts in the background.
Payroll PDFs to import files
An accounting firm receives payroll PDFs from clients, has an agent extract the figures, update the firm's payroll template, and generate the import file ready to load straight into the payroll system. Cuts what used to be hours of manual data entry per client into minutes.
AI merchandiser for an online shop
A hobby shop runs an AI merchandiser that scouts product opportunities from overseas suppliers, pulls in specs and pricing, and helps the owner stand up listings for the storefront. Turns trade-show scouting into a continuous background job rather than a once-a-quarter trip.
After-hours lead prep
A solo agent's assistant works the enquiry backlog overnight: for each new lead it drafts a qualifying reply, researches comps and neighbourhood context, logs budget and timeline to a shared sheet, and pencils a viewing onto the calendar. The agent skims and sends the batch first thing — every client message gets a human eye before it leaves.
Candidate screening & shortlist
A small recruiter facing 60+ CVs a week has the agent screen and rank applicants on job-relevant criteria only, draft a tailored interview invite or polite decline for each, and book the qualified ones onto the calendar. Messages go out from the recruiter's own inbox after a quick approval — an overflowing inbox becomes a ranked shortlist with replies ready to send.
WhatsApp support for an online shop
A direct-to-consumer brand routes WhatsApp customer messages to a support agent that answers the common questions — shipping times, returns, sizing — from the shop's own FAQ, captures order-help and return requests cleanly for the team, and hands anything billing-related to a human. Quick replies day or night, and it never invents a refund or a delivery date.
Overnight maintenance intake
A property manager runs an agent that takes maintenance requests on WhatsApp around the clock, sorts each by urgency — flagging anything like a burst pipe or no heat for an immediate human handoff — and logs the rest as structured, timestamped entries in a shared sheet. The morning starts with a triaged list instead of 30 unread messages.
Discovery-call screener
A coach puts a chat widget on their site that greets visitors, answers questions about programs and pricing from the coach's own FAQ, captures fit details from anyone interested, and shares the booking link for a paid discovery call. The coach spends their time on qualified prospects instead of fielding every tyre-kicker.
After-hours quoting
A trades business runs a Telegram bot that catches the after-hours enquiries the owner would otherwise miss on a job site: it captures the details (location, scope, urgency), gives a rough ballpark from the owner's own price list, and notes a preferred site-visit time for the owner to confirm in the morning. Fewer missed jobs, no phone tag.
Daily recipe planner for the family kitchen
A household sets the agent up once with their cooking preferences — cuisines, weekday cooking-time budget, number of dishes per meal, and a handful of trusted sources (favourite YouTube channels, blog sites). Every morning the agent suggests that day's recipes on WhatsApp, with links and a grocery list ready to copy. Removes the daily "what are we cooking tonight?" decision fatigue without locking the family into a rigid meal plan.
Examples are paraphrased composites from public OpenClaw community discussions, not specific customers. Your results will depend on your setup, prompts, and the channels you connect.