The map arrives in January. Two hundred accounts, a number attached to them, and no explanation of how the lines were drawn. You can accept it, or spend an afternoon working out whether that patch can produce that number.
Every page ranking for prompts on this subject is a template that assumes a clean account list and no politics. Neither assumption survives contact with a real patch. This page is the prompt layer for the carve-up itself and the capacity maths underneath it; the plan you write afterwards is covered in how to use AI to build a 90-day sales territory plan, and this page does not re-tell it.
Start from the data you have, not the data a model assumes
Most territory prompts fail at the first step because they ask for a plan before establishing what is known. Begin by forcing an inventory.
Here is my account list with industry, employee count, region, current ARR, close date of the last won deal, and date of last activity. Before planning anything, list which fields are missing or stale for more than a quarter of these accounts, and say which planning decisions I cannot make reliably until they are filled.
What comes back is the real starting position. A territory plan built on a list where half the activity dates are blank is a plan about your CRM, not about your market. Prompts for filling those gaps properly sit in Account Research & Buyer Intelligence.
Prompts that segment by something other than geography
Geography is the laziest cut and usually the worst one. The useful question is which accounts behave alike.
Group these accounts into no more than five segments, using buying behaviour rather than location: deal size, sales cycle length, who signs, and whether they bought after inbound or outbound contact. Name each segment in plain language and say how many accounts and how much pipeline sit in it.
For each segment, state the one working motion it responds to and the one it does not. Base this only on the outcomes in the data I gave you, and say where the sample is too small to support a conclusion.
That second constraint matters. Ask a model which segment is best and it will answer confidently from eleven data points. Ask it to flag thin samples and it will tell you that three of your five segments are guesses.
The capacity question: how many accounts can one rep genuinely work
This is the arithmetic that most territory plans skip, and it is the one that decides whether the patch is coverable.
Assume I can run [n] meaningful touches a week and that this segment needs [n] touches before a first meeting. Work out how many accounts I can keep in active cadence at once, how many stay dormant, and what that implies for the number I have been given. Show the calculation.
Based on my last four closed-won deals and their cycle length, work backwards: how many active opportunities do I need in each month to hit [number], and does my account count support creating them?
If the answer is that it does not, you have a conversation to have rather than a quarter to lose quietly. The prospecting cadence that sits under these numbers is in Prospecting & Pipeline Creation, and the daily version of the same work is in AI prompts for SDRs.
Finding the accounts nobody has touched in a year
Every inherited territory contains accounts that were worked once, by someone who has left, and then forgotten.
Rank these accounts by time since any buyer-side activity. For anything over 12 months, summarise what the notes say happened, what the stated reason for no decision was, and whether that reason is likely to have changed since.
From this list of dormant accounts, pick the ten where the recorded objection was about timing or budget cycle rather than fit, and draft the first line of a re-approach for each that references what actually happened.
Dormant accounts with a timing objection are the cheapest pipeline in a new territory, because somebody already did the qualification work and wrote it down.
Building the case for the territory you want
Territories are negotiated, not assigned, by anyone who brings evidence to the conversation.
Here is my current account list and the accounts I am asking for. Build the argument for the change using only coverage and capacity evidence: accounts I cannot reach at my touch rate, segments where I have no comparable win, and accounts adjacent to my strongest segment. Then write the counter-argument my manager will make.
Run the counter-argument before the meeting, not during it. The reporting view your manager will be looking at is covered in RevOps, Data & Analytics Support.
For managers: splitting a territory without losing a quarter
Splitting a patch is a different job from planning one, and the cost always lands in the handover.
Here are [n] reps and [n] accounts with ARR, stage and last activity. Propose two balanced splits using different balancing criteria, state which criterion each one optimises, and list for each split the three accounts most likely to be damaged by changing owner mid-cycle.
For every account changing hands, write the handover note: what is true today, what the open question is, and what the new owner must not re-ask the customer.
The second prompt is the one that protects the quarter. The coaching conversations that follow a reorg are in Leadership, Coaching & People Management, and the manager workflow around them is in AI prompts for sales managers.
Take one prompt free, with no account, at /free-prompt. Pro is $12 per month or $89 per year, and lifetime access is $149 — See pricing.
What a model cannot see, and why its balanced split usually is not
A model splitting your accounts is working from a spreadsheet. It cannot see that one rep has a four-year relationship with the account you are about to move, that another is three months from leaving, or that the logo everyone wants is run by a buyer who will not take a new contact. It will still produce a split that looks balanced, because ARR and account count were the only things it was given to balance.
So use it for the parts that are arithmetic and surface the rest for a human to weigh. Three habits keep this honest: give it only recorded data and make it name what is missing, require the calculation rather than the conclusion, and always ask for the second-best option and the reason it lost. A territory plan that survives February is one where somebody argued with the map in January.
The rest of the catalogue, by role and by stage, is mapped in AI prompts for sales, and the account-executive view of the same patch is in AI prompts for account executives. Browse the library → or read the full taxonomy.
Frequently asked questions
Can ChatGPT build a sales territory plan?
It can do the arithmetic and the segmentation, which is most of the work nobody enjoys. It cannot see rep capacity beyond what you tell it, existing relationships, or internal politics, so treat its output as a first cut to argue with rather than a plan to adopt. The useful sequence is inventory the data, segment by behaviour, then test coverage against your real touch rate.
What are good AI prompts for territory planning?
Four earn their place: a data-inventory prompt that names what is missing before planning starts, a segmentation prompt that groups accounts by buying behaviour rather than geography, a capacity prompt that shows the calculation from touches per week to opportunities per month, and a dormant-account prompt that ranks by time since buyer-side activity.
How do I split sales territories fairly with AI?
Ask for two splits optimised on different criteria rather than one balanced answer, because fairness depends on which thing you are balancing — ARR, account count, or coverable workload. Then have it list the accounts most likely to be damaged by changing owner mid-cycle, and write a handover note for each one. That list is where a reorg actually costs money.
What data does AI need to plan a territory?
At minimum: account name, industry, size, region, current revenue, stage of any open opportunity, date of the last buyer-side activity, and the recorded reason for any past no-decision. Missing activity dates are the most common blocker, so ask the model to flag stale and empty fields before it plans anything.
How often should a territory plan be revisited?
Monthly, against two questions: which accounts have moved out of active cadence, and which assumption in the capacity calculation has turned out to be wrong. A plan reviewed once a year is a January artefact. A plan reviewed monthly is a working document that tells you when to ask for a different patch.