An AI agent for field teams: a route each morning, answers on the road, notes at every stop
Reps and merchandisers need help at three moments of the day. How AI employees plan the route, answer questions from the road and turn check-ins into tomorrow's brief.
Part of the guide Workflow automation
An AI agent for field teams is most useful at three moments of a rep's day: before they leave, when the day's stops need putting in order and the roads checked; on the road, when a question comes up that the office would normally answer; and at each stop, when what happened needs recording without paperwork. In xMap Flows these are three small AI employees rather than one large one: a morning route planner, a chat agent the team can message, and check-ins on the phone that feed a daily brief.
This article is for sales and merchandising managers, distributors and service companies with people out in a city every day. The field operations guide covers the wider picture; here we focus on how the three pieces fit.
Before they leave: the route
A morning flow takes the day's places, from a table of accounts in My data or from new outlets a search found overnight, and gives each rep their stops in order:
- Split by territory shares the places between reps by area, in compact territories of about the same size.
- Route planner puts each rep's stops in the best order from a start you pick, by car, on foot or by bike, with the distance and time of each leg and a directions link.
- Traffic incidents around each stop flags closures and roadworks before anyone sets off.
Each rep finds their route in Results, Visits on their phone, their own route first.
On the road: an agent they can message
Questions in the field are usually small and urgent: is there a new grocery store near my next stop that is not our customer?, is the road to the depot clear?, what did we agree with this shop last month?. A chat agent answers them. Start a flow with the message event of the team's Telegram bot or WhatsApp Business number, or with a Chat block shared as a link, add an AI agent, link its tools, and end with Reply.
| The rep asks about | Tool to link |
|---|---|
| Outlets around them | Nearby places, its point left to the agent |
| Whether a place is already an account | Your customer table, through My data or Database lookup |
| Closures and roadworks | Traffic incidents |
| Tomorrow's weather | Weather forecast |
| Prices, promotions, policies | Look things up in, pointed at your files |
Give the agent only the tools your reps really need.
With Memory on Per conversation, each rep's thread is kept separately, so a follow-up such as and the one after that? makes sense.
At each stop: check-ins
In Visits, each stop has Navigate, Call and Check in. A check-in records what happened (Visited, Not interested, Closed for good, Nobody there), a note and a photo, made smaller on the phone and kept with the result for your team only. The visit goes into the result's history, so the next person who visits that shop sees how the last call went.
A worked example (made up)
A snack distributor in Lima has six merchandisers covering corner shops and small supermarkets. Before, a supervisor sent routes by message each morning and collected notes by phone each evening. Now:
- 06:30, the route employee: My data (the shop list), Only new outlets added from a weekly Nearby places search, Split by territory between six, Route planner from the warehouse, Traffic incidents around each stop. Each merchandiser opens Visits and sees an ordered list.
- During the day, a merchandiser writes to the team's Telegram bot: new bodegas near Avenida Arequipa 2400 that we do not serve?. The agent searches around that address, checks the customer table, and answers with three shops and their distances.
- At each stop, a check-in: Visited with a photo of the shelf, or Closed for good.
- 18:00, a briefing employee reads the day's visit notes and posts a summary to the supervisor: shops closed, shelves to restock, new shops worth a call.
Mistakes to avoid
- Routes that cannot be done. Route planner's Most stops in a route keeps each day to what fits; set it from what a rep really manages.
- Alerts for the whole city. Search closures around the day's stops, not everywhere, or the team learns to ignore them.
- Prices from memory. Put price lists in files the agent searches; memory is for the conversation, not for facts.
- Replies to the wrong place. Reply answers on the channel that started the run. A morning flow started by a schedule needs Email or an app block instead.
- Shared memory. Shared by the whole flow mixes reps' threads; keep Per conversation.
What lands where
The route lands in Visits on each phone, with closures noted on the stops near them. Answers to questions land in the chat the rep wrote from. Check-ins land in each result's history, and the evening summary lands in the supervisor's email or Slack. Ready-made starting points: new outlets, graded and routed to each rep, road and weather chat for drivers and field visit notes into a daily briefing.
Questions
Do reps need an xMap Flows account?
To see their routes and check in from Results, Visits, yes: they are members of your organization. To ask the chat agent through a Telegram bot or a shared chat page, they only need the link or the bot.
Can the route take traffic into account?
Traffic incidents flags closures and roadworks near the stops. Route planner orders stops on roads by car, foot or bike; when it cannot reach its routing service it gives an estimate and says so.
What does a check-in record?
What happened at the stop, a note and an optional photo, kept in the result's history for your team only.
Can one agent do all three jobs?
It is simpler as three small AI employees: each runs at its own time, and a problem in one does not stop the others.