Farm Succession Planning: Use AI to Prepare the Family Conversation

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for Farm Succession Planning: Use AI to Prepare the Family Conversation.
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for Farm Succession Planning: Use AI to Prepare the Family Conversation.

AI can prepare a farm succession conversation by turning your records into a one-page fact sheet, writing a private question list for each family member, and summarising two or three scenarios in neutral words. It cannot decide who farms, who is paid and when, and it cannot give legal or tax advice. A lawyer and an accountant do that.

The value is in preparation. Succession talks stall because nobody has the same facts, because the first meeting turns into an argument about fairness, or because the person who owns the land finds it hard to start. A tidy fact sheet, questions people can answer alone, and scenarios that read like options rather than verdicts remove several of those obstacles. The work you give AI is organising and drafting. The work you keep is listening.

Follow me on Instagram@sagnikteaches

What AI is good for here, and what it must never be asked

Succession planning combines money, land, law, tax and family feeling. AI helps with only the first and last in the sense of structure, never in the sense of judgement. A sensible split:

Connect on LinkedInSagnik Bhattacharya
JobFair to give to AIKeep with people
Collecting factsMerging your notes into one tidy sheet; spotting gaps ("no figure for machinery debt")Supplying and verifying every figure
QuestionsDrafting neutral, open questions for each personDeciding who is asked what, and in what setting
ScenariosLaying out options side by side, with what each would need to be trueChoosing, and testing with advisers
Meeting prepAgenda, timings, a neutral summary after the meetingChairing, listening, handling emotion
Legal and taxPlain-English explanations of terms you were givenAdvice on wills, trusts, partnerships, transfers and tax: a lawyer and an accountant

The fourth row is where AI most often overreaches. A chatbot will happily say that a family partnership "usually" saves tax or that a will "should" leave the farm to one child. Land, tax and inheritance rules differ between places and change, and the right answer depends on details the chatbot does not have. Treat any such statement as a question to put to your adviser, never as information.

Subscribe on YouTube@sagnikteaches

Step 1: Build the fact sheet from your records

A fact sheet is one or two pages that everyone in the conversation can read before the first meeting. It should be dull, specific and agreed. Start from what you already have: the latest accounts, a list of land and who owns what, machinery and livestock lists, a loan schedule, and any existing wills or partnership papers (you do not need to paste the documents themselves; a short description is enough).

Write rough notes using roles instead of names, then ask the model to turn them into a standard structure.

Turn these rough notes into a one-page fact sheet for a family farm
succession conversation. Use headings: The business today | Land and
buildings | Machinery and livestock | Debts | Who works on the farm |
What is not yet known. Do not add figures or facts that are not in the
notes. Where a number is missing, write "Not provided" and list it
under "What is not yet known".
Notes (roles only, figures rounded):
- Mixed farm, 220 acres: 140 owned by Parent A, 80 rented on a 3-year
  tenancy ending in 2 years.
- 120 suckler cows, 300 ewes, 150 acres of cereals.
- Machinery worth about 180k, finance outstanding about 45k.
- Mortgage on owned land about 250k, repayments 28k a year.
- Parent A (62) and Parent B (60) work full time. Child 1 (31) works
  full time on the farm. Child 2 (28) works off the farm, helps at peak times.
- Annual profit before drawings roughly 85k over the last three years.

Illustrative output, shortened:

The business today. A mixed beef, sheep and cereal farm of 220 acres, 140 owned and 80 rented. Profit before drawings has averaged about 85,000 a year over the last three years.
Debts. Mortgage about 250,000 (repayments 28,000 a year); machinery finance about 45,000.
Who works on the farm. Parent A and Parent B full time; Child 1 full time; Child 2 helps at peak times and works elsewhere.
What is not yet known. Current land value; whether the tenancy will be renewed; existing wills and any partnership agreement; each person's drawings.

Two things to notice. First, the model has done what you asked and turned a gap into a list. "What is not yet known" is the most valuable section, because it becomes the agenda for the adviser meetings. Second, check every number against the original. Models can transpose figures, round them without saying so, or merge two items. Read the sheet beside your notes with a pencil. (For why a confident number is not the same as a correct one, see AI hallucinations explained for business owners.)

The Complete Excel Course with AI Integration thumbnail

The Complete Excel Course with AI Integration

From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.

By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner

Buy now

Step 2: Write separate question lists, answered in private

The most useful move before a family meeting is to ask each person, separately, what they want and what worries them. In a group, people answer in the shape of the room: the older generation speaks first, the quiet child nods, and the real worry comes out two years later. Written, private answers surface things early.

Ask AI for a list for each role, written in neutral wording, with no leading questions.

Write a private question list of 10 questions for the child who works
off the farm and helps at peak times. Purpose: to understand what this
person wants from the farm's future before a family meeting. Rules:
open questions only; no question that implies they ought to want or not
want to farm; include one question about what would feel fair to them
and one about what would worry them most. Plain, warm wording.

Typical good questions in the output are "If the farm carried on as it is now, what part, if any, would you like to have in it in ten years?" and "What would feel unfair, and what would feel fair, to you in how things were shared?" Cut anything that sounds like a test. Add questions only you can ask, such as what happened last harvest that people are still thinking about.

Lists for each role

  • The owner generation: what they want to keep (income, a house, a say), how long they plan to work, what they fear (losing purpose, running out of money, causing a rift).
  • The child working on the farm: what they need to commit (security, a role, pay and a stake), what they do not feel heard about, what would make them leave.
  • Children off the farm: whether they want any involvement, what they would want in place of it, how they feel about the amounts involved.
  • In-laws and partners: often forgotten, and often affected by decisions on housing and income. Ask whether they want to be included and how.

Offer each person the choice of answering in writing, on a call with a neutral adviser, or not at all. Replies come back to one person (the owner, or the adviser) who combines them into a summary that does not attribute lines to people unless the person agrees. Ask the model to do that merge: paste the answers with names removed and ask for "themes that appear in two or more answers, and points where answers differ, in neutral words". Read the output yourself before anyone else sees it.

Step 3: Summarise scenarios in neutral words

Scenarios are not forecasts. They are short descriptions of what would need to be true for an option to work. Three or four are plenty. Describe each in your own notes first (one paragraph each), then ask AI to set them side by side and ask what each assumes.

Here are three scenarios for a family farm. For each, write a plain
one-paragraph description and then a list titled "This only works if",
containing the assumptions that must hold. Add a final list per scenario
titled "Questions for our lawyer and accountant". Do not recommend
one. Do not give legal or tax advice. If a point depends on local law or
tax rules, say so and put it under the adviser questions.
A) Child 1 takes over the farm business over 10 years, buying out the
other family members' shares by instalments.
B) Child 1 and Parent A farm as partners for 5 years, then review.
C) The owned land stays in family ownership and is let to Child 1; the
parents keep the house.

The "This only works if" list is where the value lies. For scenario A, the output might read: "The farm earns enough to cover living costs, loan repayments and instalments at the same time; the instalment total is agreed in writing; the tenancy on the 80 rented acres is renewed; the parents have another source of income for later life." Each assumption becomes a question for a meeting. If you find an assumption nobody has checked, you have found work to do.

A quick sum to test a buy-out scenario

Say Child 1 would pay Child 2 a share in instalments. Suppose the family and an adviser agree that a fair sum for Child 2's share is 150,000 (illustrative figure only). Spread over 10 years, that is 15,000 a year before any interest. Set that beside the farm's illustrative profit of 85,000 and the existing mortgage repayments of 28,000: after repayments 57,000 remains, and 15,000 of that would go to Child 2 each year, leaving 42,000 for living costs and reinvestment for all those who live from the farm. Ask the model to make this table, then change the numbers and watch what changes. The sum is not an answer, but it makes a vague worry ("can we afford this?") into a specific one ("we can afford 15,000 a year only if profit stays above about 70,000, because below that the family has less than 27,000 a year to live on after the mortgage"). Have your accountant redo it with real figures and the right treatment of tax.

For a broader method with best, worst and likely cases, scenario planning with AI sets out the approach this tutorial adapts for a farming family.

Plan the meeting itself

With the fact sheet and the summary in hand, the first meeting needs to be small. Its job is to agree the facts, hear each person, and decide who will gather what for the advisers. It is not a meeting to decide who gets what.

Draft an agenda for a 90-minute first family meeting about the future
of the farm. Attendees: Parent A, Parent B, Child 1, Child 2. Goals: agree
the fact sheet is accurate, hear each person's hopes and worries in turn
without interruption (5 minutes each), agree the next three actions.
Rules: no decisions today; no one has to commit to anything; a neutral
person chairs. Include timings, a ground rule on interruptions and a
line on how disagreements will be parked.

A typical agenda from this prompt runs: 10 minutes for purpose and ground rules, 15 for the fact sheet, 20 for hopes and worries (five minutes each, no replies), 20 for common themes from the private answers, 15 for what we still need to know, 10 for actions and a date for the next meeting. Print it. Appoint a chair who is not one of the main parties if you can: an adviser, a trusted neighbour, or a facilitator from a farming body. After the meeting, type up rough notes (again without names) and ask the model for a neutral summary of agreed points, open points and actions. Send it round for corrections before it becomes the record.

When a loaded sentence derails the talk

Wording matters. AI is good at rewriting a sentence that would start a fight. Take a line from your own notes and ask for three neutral alternatives.

Before: "Child 2 has never really pulled their weight, so it wouldn't be fair for them to get the same share."

After (neutral): "Child 1 has worked full time on the farm for several years, and Child 2 has helped at busy times while working elsewhere. We want to talk about how to recognise those different contributions."

The rewrite keeps the point (contributions differ) and removes the verdict. Use the model as a first filter, but remember that a polished sentence can still land badly if the underlying feeling is unresolved. A rewrite is a way into the conversation, not a way around it.

The adviser meetings: what to bring and what to ask

A lawyer and an accountant are needed, and an agricultural adviser or farm business consultant often helps with the figures. Use AI to make those paid hours go further by arriving prepared. Ask the model to turn your "What is not yet known" list and the scenario summaries into a one-page brief for each adviser: what you want to understand, the documents you will bring, and the questions you want answered. Then close the loop: ask what each adviser recommends, and write it down yourself.

It is worth asking each adviser about how they want to work with you, including whether they are comfortable with you using AI to prepare. Some will welcome a tidy brief; others will want to start from original documents. Either is reasonable. Do not paste an adviser's written advice into a chatbot to "get a second opinion"; take your questions back to them instead. For an idea of how legal offices use AI (useful when briefing yours), see an AI setup for a two-lawyer estate planning firm.

Privacy: what to leave out of the prompt

Succession papers are some of the most private a family holds. Keep the following out of every prompt: names, addresses, field names that identify the farm, bank and account numbers, medical details, and the content of wills. Use roles and round the figures. If you want to work with real documents, use a plan where your content is not used for training by default (business plans of ChatGPT and Claude work that way, per the vendors' own pages) and check what you have switched on. If you use a free consumer plan, turn off use of chats for training. How to stop AI tools training on your business data walks through it.

A family conversation that went sideways, and why

An illustrative case. A family sends the AI-written scenario summaries to everyone the night before the meeting, with no explanation. One child reads "Scenario A: Child 1 takes over the farm" and takes it as a decision. The meeting begins with an argument about a plan nobody had made. The summaries were accurate and neutral, but the context was missing.

What would have prevented it: a short covering note ("these are options for discussion, none has been chosen; we will go through them together"), delivery in a setting where questions could be asked, and a ground rule that nobody says "you always" or "you never". The lesson is that the paperwork is not the conversation. Prepare people for the document as well as the document.

Documents to gather before the adviser meetings

Advisers charge by the hour, and the first meeting is often spent asking for papers. A checklist saves most of that. Ask AI to build one from your notes, then edit it. A realistic list for a family farm includes:

  • Ownership papers: who owns each parcel of land and buildings, in whose name, and any existing charges or loans secured on them.
  • Tenancy and agreements: the tenancy documents for rented land, with dates and terms, and any grazing or contracting agreements.
  • Accounts: the last three years of business accounts and the latest management figures, plus a list of regular payments and drawings.
  • Existing legal papers: wills, any papers that let someone act for you if you lose capacity, partnership or company documents. Describe their existence to the model; do not paste their content.
  • Assets and liabilities: machinery and livestock lists, loan schedules, leases and insurance policies.
  • People: who lives in which house, who works on the farm and for how long, and any off-farm income or pensions the owners rely on.

Ask the model to mark each item as "have", "need to find" or "unsure who holds it". The result is a to-do list with an owner, and an adviser who receives a complete bundle starts on substance in the first hour rather than the third.

A neutral summary after the first meeting

Within a day of the meeting, write brief notes with no names and ask for a neutral summary. This is a sample of what comes back from a prompt asking for "agreed points, open points, actions, and anything that was said that people may want to be reminded of, written without taking sides":

Agreed: The fact sheet is accurate except for the rented acreage, which Parent A will confirm. Everyone wants the farm to remain in the family if it can support those who depend on it.
Open: What each person would want in ten years; how to value different kinds of contribution; what the parents need to live on.
Actions: Parent A to find the tenancy paperwork; Child 1 to list the investment the farm needs in the next five years; all to answer the private questions within two weeks; the owner to book the first adviser meeting.

Read it for tone before sending. If a line reads as a verdict ("Child 1 will take over"), change it. The summary should be something every person in the room could sign as accurate. It also gives the next meeting a place to start, which keeps the process moving between meetings that are often months apart.

A short list of checks before anything is shared

  • Every figure matches your records, and you have initialled the fact sheet.
  • Names and identifying details are in your own copy only, not in any chatbot history you share.
  • No sentence recommends an option or states a legal or tax conclusion.
  • Every scenario has an "only works if" list and a list of adviser questions.
  • Each person has been offered a private way to answer.
  • The meeting has a chair, an agenda, a time limit and a rule that no decisions are made at the first one.
  • Advisers know what is coming and have been told what the family wants from them.

If these seven hold, AI has done what it is good at, which is saving hours of drafting so the family can spend its energy on the conversation that matters. If you would rather plan that preparation with someone, that is what my AI implementation consultation covers: finding the jobs in a family business that AI can draft safely, and leaving the decisions with the people and advisers they belong to.

Questions families ask before using AI for succession

Can AI decide who should take over the farm?

No. Who takes over, on what terms and when are decisions about people, money and law, and they belong to the family with the advice of a lawyer and an accountant. AI can organise the facts and set out the options in plain words so the talk is better prepared, but it has no way of knowing what each person wants or what the law allows for your land.

Is it safe to put our family finances into ChatGPT or Claude?

Keep names, addresses, account numbers and anything that identifies a person out of the prompt. Use roles such as 'parent', 'farming child' and 'child working off the farm', and round the figures. A business plan that does not train on your content by default is a safer place than a free consumer plan, but anonymising is the real protection.

When should the first succession conversation happen?

Earlier than feels necessary. Many advisers suggest starting years before anyone plans to step back, because the legal and tax work, land valuation and any training or off-farm careers all take time. A first meeting to share information and listen, with no decisions asked for, is a sensible place to begin.

What if a family member refuses to take part?

Do not use AI to pressure them. Offer a private conversation with a neutral adviser, a shorter meeting, or a written list of questions they can answer in their own time. The question lists in this tutorial are designed to be answered separately, which often helps someone who finds family meetings difficult.

Further reads

Sources: general succession-planning practice as described by farm advisers and agricultural extension services; vendor privacy pages for ChatGPT and Claude plans. No legal or tax rules are stated here, because they differ by place.

Want family paperwork made clearer, not riskier?

On a 1:1 call we'll look at the preparation jobs in your family business that AI can draft safely, and set up a routine your adviser can review before anything is decided.

Book a 1:1 call with me