Dispatch October 3, 2026

What the Robot Needs to Know Before It Answers for You

Think about the last time you handed your phone to someone to cover it for an hour. A cousin, a new hire, the person at the front desk on their second day. They can take a name and a number just fine. But the second a caller asks "do you do tankless installs out in Westfield, and roughly what's that run me," they're stuck. They don't know, so they either guess or promise you'll call back.

The machine answering for you is that person, every hour of every day. It's polite, it's quick, and it will answer anything you let it, with total confidence, even when it's dead wrong. The gap between a robot that helps and one that embarrasses you comes down to one thing: what you told it before you let it talk.

We call that bundle the knowledge pack: your services and prices, the towns you cover, the jobs you turn down, and who to grab when something smells off. Write it once, keep it honest, and the guessing stops.

Start with what you actually sell

Most owners can rattle off their services from the driver's seat but have never written them down in a way a stranger could follow. The robot needs the stranger's version.

List each service in plain words a customer would use, not your trade shorthand. "Drain cleaning," not "hydro-jetting line clearance." Next to each one, put a rough price range and a note on what moves it up or down.

  • The service, named the way customers say it. If they call it a "tune-up" and you call it "preventative maintenance," write both.
  • A price range, not a single number. Something like "most drain jobs run $150 to $400 depending on how deep the clog is."
  • What changes the price. Access, the age of the system, how far you're driving, whether it's a weekend.
  • What you don't do. Just as important, and we'll get to it below.

Price ranges without painting yourself into a corner

Owners freeze right here. Give a number and you're scared of being held to it; stay vague and the customer calls someone who'll just answer. A range with a reason attached solves both.

Teach the robot to always give the range and always name the catch. "Standard service calls start around $89, and we'll confirm the full price once we've seen the job." That's honest, it keeps the conversation moving, and it leaves you room.

One rule worth burning in: the robot quotes ranges, never firm bids. A firm number on a job nobody has looked at is how you end up eating the difference. Pricing the actual job is judgement, and judgement stays with you.

The towns you serve and the jobs you turn down

Your service area lives in your head as a feeling ("eh, I'll go that far for the right job"). A machine can't feel. Give it a list.

  • Towns or zip codes you cover without a second thought.
  • The edge cases, where you'll go but charge a trip fee, and roughly how much.
  • The hard no: too far, not worth the drive.

Same goes for the work itself. Every trade has jobs it quietly doesn't take. The commercial job when you're residential only. The repair on a brand you won't stand behind. The rush you can't staff this week. Write those down as clear refusals, so the robot can give a fast, friendly no instead of booking something you'll have to cancel.

Saying no fast is a gift to everybody. The customer finds the right person sooner, and you don't burn a slot you needed.

The questions that always go to a human

This is the heart of it. A knowledge pack isn't about teaching the robot to answer everything. It's about drawing a bright line around what it must never answer on its own.

Some questions carry too much judgement, risk, or money to hand to a machine. The robot's job with those is to catch the question, hold it, and get it to you fast. Put these on the always-hand-to-a-human list:

  1. Anything that smells like an emergency. Water everywhere, no heat in January, a gas smell, a hurt animal. That gets a human now, not a booking link.
  2. Firm prices on anything custom or big. Ranges are fine; a real bid is yours.
  3. Upset or angry customers. A complaint handled by a cheerful bot only gets worse. Route it to a person.
  4. Anything legal, safety, or liability. "Is this up to code," "will my insurance cover it." Don't let the machine have an opinion.
  5. Anything it doesn't actually know. Better to say "let me get the owner to confirm" than to invent an answer.

That final point matters most. A good robot knows the edge of what it knows and stops there. Tell it plainly: when you're not sure, don't guess, hand it to a person.

Keeping it current without it turning into a chore

A knowledge pack that's a year stale is almost worse than none, because now the robot is confidently quoting last spring's prices. The fix is a small habit, not a big project.

Tie updates to things that already happen. Raised your rates? Update the pack that same hour. Dropped a service, added a town, stopped working with a brand? Same deal. Keep the whole thing in one plain document you can open from your phone, not buried in a system you have to log into.

Once a quarter, read it top to bottom. Fifteen minutes, maybe less. You'll catch the price that crept up and the town you quietly stopped driving to. That's the entire maintenance plan.

Worth doing this week

You don't need the perfect pack. You need a first one, today, that beats guessing.

  • Open a blank note and list your top five services the way customers name them, each with a rough price range.
  • Write your service area as actual towns or zips, plus the trip-fee edge and the hard no.
  • Name three jobs you refuse, so the robot can turn them down fast.
  • Write the five questions that always go to a human, and name exactly who that human is.
  • Put it all in one document on your phone, and set a quarterly reminder to reread it.

Do that, and whatever's answering for you, a missed-call text-back, a chatbot, a setup we help put together at TellRobots, starts sounding like someone who actually works there. The machine runs the relay. The judgement, the real number, the call that needs a gut, stays with you, which is exactly where it belongs.

More dispatches →