FlowsBook a callSign inCreate free accountSign up
AI employees and automation

An AI agent for utility companies: crew notes, meter rounds and storm questions

Three AI employees a water, gas or power company can run: a morning note per depot on weather and roads, meter rounds in a sensible order, and a chat for storm days.

Part of the guide AI employees

By the xMap team · · 5 min read

An AI agent for utility companies earns its keep on the routine work that eats a supervisor's morning: reading the forecast and the road closures for every job site before crews leave the depot, turning the list of meters due this week into rounds a reader can finish, and answering customers who ask what to do when a storm hits. In xMap Flows each of these is an AI employee with its own timetable. None of them touches your network controls; they read public conditions and your own lists, and they report to people who decide.

This is for operations managers, field supervisors and customer service leads at water, gas, electricity and district heating companies, and at the contractors who do their field work. It describes three AI employees, what each one reads, a made-up morning and the mistakes worth avoiding.

One: the morning note for each depot

Every supervisor already checks the weather and the traffic before allocating jobs. The difference is doing it for every site on the list, not just the ones they remember. Keep your depots, and the day's job sites, in a table in My data or a spreadsheet your work order system updates, then build the note from three readings.

  • Weather forecast, hour by hour, at each job site: rain in millimetres, the strongest wind and the temperature. Rain matters for excavations and live electrical work, wind for anything at height, frost for water mains.
  • Traffic incidents around each site: road closures and roadworks within a radius you pick, with the street and how far the nearest one is.
  • Traffic Flow on the roads near the depot gates: live speeds and the jam factor at the moment the flow runs, so a 6 am run shows the 6 am roads. It does not count cars.

An xMap LLM block then writes one short note per depot: which jobs are fine as planned, which should swap to the morning before the rain, which crew should avoid a closed street, and anything that needs a person to decide. The note lands by email or in a team chat before the first van leaves.

Two: meter rounds in an order readers can finish

Meter reading and inspection visits fail when the list is long and the order is random. The flow takes the meters due, splits them fairly and puts each reader's day in order.

  1. New rows or My data brings in the meters due, each with an address or a position.
  2. Filter keeps only what is due this week, or only the visits that need access.
  3. Split by territory shares the stops between the readers on shift, in compact areas of similar size.
  4. Route planner orders each reader's stops from the depot, by car or on foot, with the distance and minutes for every leg and a directions link.
  5. The day's rounds go to each reader, and they check in at each stop on their phone, with a note or a photo for a meter they could not reach.

A Route planner route holds up to 50 stops, which suits a working day for most readers. For dense apartment blocks, walking routes give more honest times than driving ones.

Three: a chat for storm days

When a storm comes through, call centres fill with the same questions: is it safe, is the road to my street closed, what should I do about a fallen line or a flooded basement. An AI agent behind a Chat page can take the first wave. Give it the Weather forecast and Traffic incidents as tools, and your own written guidance in its instructions, and it answers each customer about their own address: the wind and rain expected there, closures nearby, and your safety advice in plain words.

It does not know about faults on your network unless you give it that list, and it should never guess. Keep its instructions strict: for anything dangerous, it gives your emergency number and stops.

A made-up morning near Lisbon

Invented, to show how the three fit together. A water utility with four depots around Lisbon runs the depot note at 05:45. The forecast shows heavy rain over the western depot's area from 11:00. Two mains repairs that need open trenches are flagged to move to the first slot, and a road closure near a third site sends that crew round by another street. The supervisor reads it on the way in and confirms one change through Ask a person.

At 06:30 the meter flow runs. Sixty-two visits are due across the eastern districts; it splits them between three readers and orders each round from the depot. One reader's round is mostly apartment blocks, so it is planned on foot. By the evening the check-ins show four meters with no access, which go into tomorrow's list.

In the afternoon the rain arrives. The storm chat answers questions about two flooded underpasses using the live closures and gives the emergency number to a customer who mentions water near an electrical box.

Mistakes to avoid

  • One forecast for the whole region. Rain arrives street by street. Forecast each job site.
  • Routing before checking the roads. A short route through a closed street is not short. Check closures first, then order the stops.
  • Letting the agent improvise on safety. Write your guidance into its instructions and tell it to hand dangerous questions to people.
  • Sending the note too late. A note that arrives after the vans leave changes nothing. Schedule it well before the shift.
  • Expecting network data it does not have. These employees read public conditions and your tables. If you want faults or outages in the note, connect your own database with Database query.

Templates to start from

Ready-made templates cover each piece. Closures near substations for field crews writes a route note every morning from incidents around your sites. Heavy rain hour by hour: stormwater crews on standby sends the duty manager a standby plan on rainy days. The storm and outage helpline chat is the customer agent described above, and the weekly utility operations brief rolls the week's weather, roadworks and new developments into one page for the director. More on putting AI employees to work in the field is in the field operations guide.

Questions

Can the AI agent see faults on our network?

Not on its own. It reads the weather, road incidents, live road speeds and your own tables. Connect your database with Database query if you want faults or outage lists in its answers.

How early can the depot note arrive?

Whenever you schedule it. Most teams run it 30 to 60 minutes before the first crews leave, so the supervisor can act on it.

How many stops can a meter round hold?

A Route planner route holds up to 50 stops, and the stops can be shared between 2 to 10 readers.

Will the storm chat give safety advice it makes up?

It should only repeat the guidance you write into its instructions. Tell it to give your emergency number and stop for anything dangerous.

Templates that do this

Read next

More on ai employees and automation · Every article · Templates

Put it to work

An AI employee in xMap Flows does this on live data, on its own timetable. Start from a template or describe it in plain words.