Home / Blog / AI answering playbook

How Do You Write an AI Answering Playbook for Your Business?

Composed Chaos · September 14, 2026 · Updated September 14, 2026

# How Do You Write an AI Answering Playbook for Your Business?

An AI answering playbook is the written set of rules that tells the AI what your business does, what it's allowed to say, what it should never say, and when to hand the call to a human. Get it right and the AI sounds like your best office manager. Get it wrong and it books jobs you can't do, quotes prices you'd never honor, or fumbles the one call that mattered.

The good news: writing a playbook is not a technical job. It's the same knowledge you'd give a new hire on their first morning — just written down instead of explained in passing. This guide walks through the pieces, in the order that actually matters.

Why does a playbook matter more than the AI itself?

Every AI answering tool is only as good as what you tell it. The model can hold a natural conversation, but it doesn't know your service area, your pricing rules, or that you stopped doing mobile-home releveling two years ago. That knowledge comes from you. If you want to understand what the AI is doing with those rules on a live call, our explainer on how AI answering actually works covers the mechanics.

Think of the playbook as the difference between a temp who read your website and a manager who's worked your front desk for five years. Same voice on the phone. Very different outcomes.

What should go in the playbook first?

Start with the boring facts, because the boring facts are what callers actually ask about. Before any clever scripting, the AI needs your ground truth:

Write these in plain language. You are not programming — you're briefing. Short, direct sentences work best, and they're easier to keep accurate as your business changes.

How specific should the service rules be?

Specific enough that the AI never has to guess. Vague rules are where AI receptionists embarrass you.

"We do plumbing" is not a rule. "We do residential drain cleaning, fixture installs, and tank water heaters; we do not do tankless installs, commercial jobs, or anything requiring a permit we don't pull" — that's a rule. The AI can act on the second one. On the first, it will happily agree to a commercial backflow job because a caller asked confidently.

The most useful thing you can write is a good negative list: the jobs you want the AI to politely decline and route elsewhere. Contractors consistently underestimate how much this one section saves them. Every "sorry, we don't handle that" the AI delivers cleanly is a truck roll you didn't waste and a callback you didn't have to make.

What are negative rules, and why do they matter so much?

Negative rules are the "never do this" instructions — and they're the part of a playbook most owners forget until it bites them.

A few that come up constantly in the trades:

• Never quote a price on complex work. Slab leaks, main-line replacements, panel upgrades, full system swaps — these need eyes on the job. The rule is simple: give the diagnostic or trip fee, then book the visit. Never a repair number over the phone.
• Never promise a same-day arrival you can't staff. If your calendar is the source of truth, the AI should book from it, not invent a window to make the caller happy.
• Never guess at warranty or insurance coverage. "I'll have someone confirm that" is the right answer.
• Never argue. An irate caller gets acknowledged and escalated, not debated.

These aren't hypotheticals. The failure mode of a helpful assistant is being too agreeable — saying yes to keep the caller happy. Negative rules are how you cap that.

How does the AI know when to book — and when to book what?

Booking is where a playbook earns its keep, and where it's easiest to create a mess. Two rules carry most of the weight.

First: only offer real, open time. An AI that books against your actual calendar can't double-book or invent a Tuesday slot that doesn't exist. That single design choice — booking from real availability instead of a canned "we'll get back to you" — is what turns a phone call into a job on the board.

Second: collect the same intake every time. Decide the questions once and make them non-negotiable: name, callback number, service address, and a plain description of the problem. Consistent intake means every lead that lands in your inbox is one you can actually act on, whether it came in by phone or by email — and email lead answering is live today alongside voice.

For businesses running crews across town, booking also has to respect travel reality — you can't send the same tech to two jobs 40 minutes apart back to back. That's a scheduling problem, not just an answering one; we get into it in how real-calendar booking prevents invented time slots.

When should a call go to a human?

Always define the handoff. An AI that never escalates is as dangerous as one that escalates everything.

Good escalation triggers are concrete and easy to test:

• Anything involving gas, active flooding, sparking, or a safety risk — route to your emergency line immediately.
• A caller who is upset, confused, or clearly needs reassurance a script can't give.
• High-dollar or unusual jobs where a wrong answer costs real money.
• Anything the playbook simply doesn't cover — better to hand off than improvise.

Write these as "if this, then reach a human, and here's how." The AI won't replace your judgment, quote a complex job, or talk an angry customer off the ledge — and it shouldn't try. The point of the playbook is to make sure those exact moments land on you, and the routine ninety percent doesn't.

Do the rules carry across languages?

Yes — and this is a place where a written playbook quietly pays off. When your rules live in plain text, they don't have to be re-taught for every language you answer in. The same service list, the same negative rules, the same escalation triggers apply whether the caller speaks English or Spanish. If bilingual answering is on your radar, the playbook logic carries over cleanly; we cover the market case in our piece on bilingual answering.

How do you keep the playbook from going stale?

A playbook is a living document, not a one-time setup. The businesses that get the most out of an AI receptionist revisit it whenever something changes:

• You add or drop a service line.
• Your prices or trip fees move.
• You expand or shrink your service area.
• A season shifts your after-hours policy.
• You catch the AI saying something you'd rather it hadn't — fix the rule, not the symptom.

Set a reminder to skim it once a quarter. Ten minutes of maintenance beats a month of booking the wrong jobs.

Where do you start?

Don't try to write the perfect playbook on day one. Start with the four sections that prevent the worst mistakes — services, service area, pricing rules, and escalation — and refine from there. You already know all of it; you've just been carrying it in your head.

If you want to hear what a well-briefed AI sounds like before you write a word, call the live demo line at (816) 382-0487 — the AI answers as a fictional plumbing company, and you can push on it the way a real caller would.

Curious how the pieces fit together, or what's live today versus coming soon? The FAQ covers the common questions, and the pricing page lays out what's included at each tier.