How to Write a Door Policy for Your Sales Calendar

TL;DR: A door policy is a short list of firm rules that decide who gets a meeting with your team. In Aijent's SureGate, each rule is one of three kinds: a number, an amount of money, or a category. Each has a line where Go begins and a line where Review begins, and everything past that is a warm no. This is the practical guide: how one rule is built, the four protections that apply to every rule whatever you write, a seven-step method for writing your own, six worked policies for six different kinds of business, and how to write the decline so that nobody leaves feeling turned away. SureGate is available on Growth and above.

We have written about why your website needs a door policy, in GO, REVIEW, NO-GO. This post is the other half: how to actually write one.

It is a worksheet more than an essay. By the end you should be able to sit down with a coffee, look at the last few meetings you wish you had not taken, and turn them into a handful of rules your website agent will apply the same way, every time, to every visitor.

If you are new to the idea of an agent that qualifies visitors on your website at all, start with what an AI sales agent is.

Judgement first, rules second

Every Aijent qualifies with judgement before it ever applies a rule. That judgement is Visitor Screening: you answer two questions, who your ideal customer is and who the agent should politely screen out, and it labels each conversation good fit, poor fit or not a prospect. It is on every plan, and our post on ideal buyer fit scoring explains exactly how it decides.

Judgement handles most of the job. It is good at the fuzzy, many-sided calls: a real business that is not quite your kind of buyer, a student writing an assignment, a supplier pitching.

What judgement does not do is promise the same answer to the same facts every time. Some decisions should not be judgement calls. If you never take projects below a certain size, or you only work in three states, that is a rule. You want it applied identically to the first visitor of the day and the four hundredth.

That is what a door policy is for, and it is what SureGate does. It sits on top of Visitor Screening. Screening is step one, and SureGate does not run without it.

The anatomy of one rule

A SureGate rule, which the dashboard calls a criterion, has a handful of parts. Once you know them, every policy is just a few of these written down.

What it is about. A plain label: "Funds under management", "Number of staff", "State".

What kind of rule it is. One of three:

• A number. Staff, sites, seats, users, rooms.
• An amount of money. Budget, project value, annual spend, funds under management. You choose the unit, so a rule can be written in thousands or millions.
• A category. State, country, industry, type of project.

Where Go begins and where Review begins. For numbers and money, you set two thresholds. With "higher is better", anything at or above the Go line is Go, anything at or above the Review line but under Go is Review, and anything below the Review line is No-go. You can flip the direction to "lower is better" for things like maximum team size, where too big is the problem.

For categories, you write two lists: the values that are Go and the values that are Review. Then you choose what happens to anything on neither list: No-go, or Review if you would rather a person decided.

How the agent learns it. Either it asks, or it only uses the fact if the visitor volunteers it. Choose "ask" for the facts that really decide a meeting. Choose "volunteered" for facts that matter only when they come up.

The question. For a rule the agent asks, you get the exact wording it will use. It is drafted for you when you create the rule, and you can rewrite it. Once you edit it, your words stay; nothing regenerates over them unless you ask it to.

That is the whole rule. Nothing in it is a prompt you have to engineer.

The four protections every policy gets

Whatever you write, four rules sit underneath it. You cannot switch them off, and you would not want to.

1. The strictest outcome wins. If one rule says Go and another says No-go, the visitor is No-go. If the worst known answer is Review, it is Review.
2. An unknown fact is never held against anyone. If the visitor has not said how many staff they have, the staff rule is silent. No block, no downgrade.
3. Declining to answer is never a no. A visitor who would rather not share a figure goes to Review, never to No-go. A person decides.
4. Review follows your screening setting. You already chose, in Visitor Screening, what happens when the agent is unsure: hand the visitor to a person, or let the conversation continue tagged for review. Review in SureGate does the same. There is one borderline setting, not two.

There is a fifth, quieter protection. A fact only counts if the visitor actually said it. The value the agent works from has to be traceable to the visitor's own words in the conversation, not something the agent assumed or said itself.

Seven steps to your first policy

1. Start with the meetings you regret. Open your calendar and look back over the meetings you wish you had not taken. Write one line for each: why was it wrong? Too small. Wrong state. Wanted something you do not do. Already locked into a contract for two years.

2. Keep only the firm reasons. Cross out anything that is a feeling rather than a fact. "Seemed unserious" is a judgement, and screening already handles judgement. "Budget under our minimum" is a fact you can write a line for.

3. Pick no more than three rules to start. Every rule the agent asks about is a question in the conversation. Three good questions feel like a thoughtful salesperson. Eight feel like a form with a personality. You can add more later.

4. Draw the lines. For each number or amount, decide two figures: the point where you would happily take the meeting (Go), and the point below which you would never take it (the bottom of Review). The gap between them is your maybe zone. Make it wide enough to matter. A Review band that is one dollar wide is a rule pretending to be a safety net.

5. Choose ask or volunteered. Ask for the one or two facts that decide most meetings. Leave the rest as volunteered, so they only count when the visitor brings them up.

6. Read every question out loud. If it sounds like a form field, rewrite it the way your best salesperson would say it. "What size team would be using this?" beats "Number of employees?".

7. Test it before it goes live. The SureGate builder has a simulator: type in a hypothetical visitor's answers and run the test. It uses exactly the same logic the live agent uses, so what you see is what a visitor will get. Try the edge cases: a figure just under each line, a missing answer, an answer on neither category list.

Six worked policies

Each of these is illustrative, not a real customer. The figures are examples to show the shape, not recommendations. Your lines are yours.

1. A wealth advice practice

The practice only takes on clients above a certain level of investable assets, and is happy to talk to people a little under it if they are growing.

• Investable assets (amount, in millions, higher is better, asked). Go at 2. Review from 1.
• Location (category, volunteered). Go: Australia. Anything else: Review.

Somebody with 3 million is Go. Somebody with 1.5 million is Review, so an adviser decides. Somebody with half a million gets a warm no. A visitor in Singapore with 5 million is Review, because the strictest outcome wins and location says Review.

2. A managed IT provider

The business serves offices from about twenty staff and only in two states, because that is where its technicians are.

• Staff (number, higher is better, asked). Go at 20. Review from 10.
• State (category, asked). Go: New South Wales, Victoria. Review: ACT. Anything else: No-go.

You write the states in your own vocabulary. If a visitor says "we're in Parramatta", the agent maps that onto your list rather than looking for the word "Parramatta" in it.

3. An architecture practice

Commercial work only, above a minimum construction budget.

• Construction budget (amount, in millions, higher is better, asked). Go at 1. Review from 0.5.
• Project type (category, asked). Go: office, retail, hospitality. Review: mixed use. Anything else: No-go.

A homeowner asking about a kitchen renovation is No-go on project type, and never has to share a budget to find that out. The decline arrives warmly, and the agent keeps answering their questions from the practice's website.

4. A recruitment agency

The agency places permanent technology roles and does not handle temporary staffing.

• Role type (category, asked). Go: software engineering, data, product, cyber security. Review: IT support. Anything else: No-go.
• Number of hires this year (number, higher is better, volunteered). Go at 3. Review from 1.

The hires rule is volunteered because asking a hiring manager to commit to a number in the first conversation is off-putting. If they mention it, it counts. If they do not, it is silent.

5. HR software built for small teams

This is the "lower is better" case. The product is designed for small businesses, and larger companies churn because they outgrow it.

• Number of employees (number, lower is better, asked). Go up to 150. Review up to 300. Above that: No-go.

A company with 80 staff is Go. One with 220 is Review, so someone can judge whether the product will stretch. One with 2,000 gets a warm no, which is kinder than a demo that ends in "we're not built for you".

6. A commercial cleaning company

The company services multi-site clients and has crews in the eastern states.

• Number of sites (number, higher is better, asked). Go at 3. Review from 2.
• Region (category, volunteered). Go: Queensland, New South Wales, Victoria. Anything else: Review.

Region is volunteered and anything off the list goes to Review rather than No-go, because the owner would rather hear about a promising lead in South Australia than lose it.

Notice what none of these policies does. None of them asks about timing, authority, or whether the visitor is "just looking". Those are judgement calls, and screening already makes them, carefully. A door policy is for the handful of facts that end the conversation no matter how nice everyone is being.

Writing the warm no

A badly handled no is brand damage, so this part matters as much as the rules.

Here is what happens when a visitor reaches No-go. The agent does not announce anything. It keeps helping, answering questions from your website in the same voice, but it no longer has any way to book a meeting or share your link. When the visitor reaches for that next step, your decline message is posted once, exactly as written. If they push back, they are offered a callback form, once, and your team decides. The full transcript is saved.

You write the decline message, and three rules make it work.

Never name the line that fired. Nobody should read "your budget is below our minimum". The decline is the same message whichever rule decided it, by design.

Point somewhere useful. A resource, a partner, a different service, a reason to come back later. A no with somewhere to go feels like advice.

Sound like you. The message is drafted for you from your rules, and it updates as the rules change, until you edit it. From that moment your words are kept.

Three examples, to show the register:

"Thanks for the detail. From what you've described, we're probably not the right fit for this one, and we'd rather tell you that now than waste your time. The guides on our website are yours to use, and if things change, we'd love to hear from you."

"We focus on a fairly specific kind of project, and this one sits outside it. We'd rather point you in a better direction than take a meeting that doesn't help you. If your plans change, come back any time."

"Honest answer: we're built for smaller teams than yours, and you'd outgrow us quickly. You deserve a product that fits from day one."

None of them names a number. All of them would be fine to read aloud to the visitor's face.

Mistakes to avoid

Too many rules. Every asked rule is a question. Start with three at most.

Using a rule for a feeling. "Serious buyers only" is not a rule. It is screening's job, and screening is already doing it.

A Review band with no width. If Go and Review are almost the same figure, nobody ever lands in Review, and you lose the safety net that catches near misses.

Forgetting the direction. If smaller is better for your product, switch the rule to "lower is better". Otherwise your best-fit customers will read as too small.

Making everything No-go by default. For a category, anything on neither list goes wherever you send it. If you are not sure about the rest of the world, send it to Review. You can always tighten it later.

Never reading the transcripts. Every conversation's transcript is saved. Read the Reviews and the No-gos for the first few weeks, and put any answers that surprised you through the simulator, which tells you which rule decided them. If a line is in the wrong place, move it.

Where to start

Turn on Visitor Screening first and answer its two questions. It is on every plan, and SureGate builds on it.

Then write your door policy in SureGate, which is available on Growth and above, at one flat monthly price with conversations included. Growth is $199 a month with 200 conversations, and there is no setup fee and no contract. Every plan is laid out on the pricing page.

A door policy is not about turning people away. It is about deciding, once and calmly, who your calendar is for, and then letting every visitor hear an honest answer in a voice you chose.