The Drover Estate

Help me build an estate map of my business. Interview me first, then draw it. Start with Round 1 straight away, after one line telling me I can say "quick version" at any point.

## What an estate map is

A living, one-page view of everything my business runs or is building: systems, apps, portals and websites, plus the ordinary assets people rely on, like proposal templates, costing spreadsheets, price lists and registers. It shows what exists, what's being built, what's planned, what's being retired and what used to be there. For each piece it says who is responsible, who it serves, where it lives and what it connects to. It also shows where the business is on its AI journey, and what AI is likely to change for a business like mine.

It's an entry point, not an encyclopaedia. Detail stays where it lives and the map points to it. Think of a contractor arriving at an old farm estate: before running power to a new shed, they need to know what's already there, what's reserved, what's buried and who to call.

## How to interview me

- Ask in short rounds, one topic at a time, five questions at most per round.
- If I upload or paste documents (system lists, org charts, spreadsheets, old proposals), read them first and only ask about the gaps.
- If you can see connected tools in this chat (for example Microsoft 365, Google Drive, HubSpot or Slack), offer to look there before asking me, and tell me what you found. Treat anything you find as a lead for me to confirm, not a fact.
- "I don't know" is a good answer. Record it as an open question and ask who would know. Never fill a gap with a guess.
- Use my words, and keep the detail at name-and-address level.
- After Round 3, draw a rough first version of the map so I can see it taking shape. Then keep going and update it as we learn more.
- If I say "quick version", stop after Round 4, ask the decision velocity question from Round 9, then draw the map and list the later rounds as open questions. The AI journey ladder still appears, marked provisional.

Round 1, the title deeds: legal entities and trading names, and only the ownership context that explains who decides. Our brand directory: where the logo files, fonts, colour codes and icons live, and who looks after them. Our sites and what happens at each. What we sell, who we sell it to, and roughly how big we are. What customers would say they choose us for.

Round 2, the buildings: every system and tool (accounts or ERP, CRM, production or MRP, timekeeping, websites, portals, file storage, design or engineering, marketing automation, AI tools) and every ordinary asset people rely on (templates, costing and pricing spreadsheets, registers, checklists). For each: what it is, where it lives, and its state.

Round 3, business areas and their names: how do our people group the work, and what do they call each area? What does each name mean? Do we organise around departments, sites, products, customer journeys or something else? Which components serve more than one area? Use our established terms. If we have no established names, suggest plain descriptions based on what you've learned, mark them as provisional and ask me to confirm or correct them. Names that mean something to us are always better than borrowed ones. Only use names from an example map, or a themed set like farm names, if they genuinely fit how we work. Revisit the groupings after learning who uses each component and why.

Round 4, people and reasons: for each, one named, accountable person to call for a decision or when something goes wrong (a name, not a team), who built it if that's someone else, who uses it, and why it exists.

Round 5, what's moving: work underway or planned, who is building it, who it's for, and its definition of done. Anything being retired: what still depends on it, and what knowledge is inside it.

Round 6, outside the fence: customers, dealers, suppliers, vendors and contractors, and which of my systems each one touches.

Round 7, below ground: retired systems, old databases and workbooks that still hold knowledge, earlier attempts that were abandoned, and who built them if anyone remembers.

Round 8, the common questions, for the five or six components that matter most. Give me these as a table to fill in rather than one question at a time: sign-in and MFA; where the data lives, including the country; backups, and whether a restore has ever been tested; who holds keys and tokens and how they're rotated; what could be revoked or contained after a breach; whether AI is used and how its spend is managed; what API it offers, and whether an MCP server is connected or planned (which one, how many tools, whether they can write or delete, and who connected it); which third parties sit behind it. Start the API and MCP part from what you can see: list the connectors available to you in this chat, how many tools each brings and whether any can write, send or delete. Then ask me which belong to the business and who connected them.

Round 9, the AI journey and decisions: who uses AI today and for what, whether that's chat tools, AI features inside systems we already pay for, or anything connected to our own data. What a business like ours could realistically do with it. The decisions we make most often, and the ones that matter most: who makes them, how long they take, what information they wait on, and what slows them down. Then ask me the decision velocity question described in the NB at the end.

## What to build

When you have enough for a first draft, build one self-contained HTML page (or another format, if I ask for one). Make it readable on a phone; the map can scroll sideways. Include:

1. A header: business name, legal entity and trading names, and "statuses as at [date]".
2. A legend of states, plus a key for the AI journey ladder.
3. The map, grouped into the business areas and names I confirmed during the interview. Let the number of areas and their layout follow the work; don't impose four zones or an org chart, and prefer names that are relevant to us over themed ones. Add a short plain-language description beside any unfamiliar area name. Legal entity, trading names, brand directory and ownership at the top. The AI journey ladder for the whole business in the top-left or top-right corner (see the NB). Outside parties beyond the boundary fence. Retired assets below ground. Each component shows its name, its state as a word, and its person to call. Tag any component with an MCP server connected, showing how many tools it brings. Draw only the connections someone needs to find their way: shared records, data flows, replacements. Lay it out on a grid so lines run between boxes and never through them, and keep labels short enough to fit inside their boxes.
4. An entry for each component that opens when it's clicked: what it is; who to call; built by and maintained by, if different; who it serves; why it exists; where it lives; what it connects to; definition of done, if it's underway or planned; its rung on the AI journey ladder, if it has one, with a sentence on why and what the next rung would look like; open questions; and who confirmed the entry, and when. For the brand directory, also list the colours with their hex codes, the fonts and the icon style.
5. A shared-questions table: the main components across, the common questions down, so I can read a row and see how many times the same problem has been solved.
6. An APIs and MCP table: for each component, the API it provides, whether an MCP server is connected or planned, how many tools it brings and when you counted them, what those tools can do, and who it's connected for.
7. Where we are on the AI journey: the overall rung, why you placed us there, and one or two concrete steps that would move a business like ours up a rung. See the NB.
8. What AI changes for a business like ours: a short written reflection, described in the NB.
9. Decisions: the decisions I named in Round 9, who makes them, how often, how long they take, which components on the map feed them, and my answer to the decision velocity question.
10. Open questions, each with who can answer it, the next step, and which part of the map it would change.
11. Where a document and reality disagree, if anything I've told you conflicts with something I've shared.
12. Sources: what the map was drawn from, and when.
13. A change log: dated updates, newest first.

## States (always a word as well as a colour)

- built: solid green outline
- underway: dotted orange outline
- planned: dotted light-grey outline
- retiring: still holding something up while it's replaced; give it its own style
- retired: history kept for context, shown below ground
- theirs: outside parties, not mine to build
- unknown: a recognised gap, styled so nobody mistakes it for resolved

## Rules

- The map records; it doesn't decide. Note each decision with who made it and when.
- Grade state by what a thing does today, not by whether its code or file exists.
- Never show a plan as built.
- Every component gets one named, accountable person to call. If nobody owns it, that's an open question.
- Every component gets a beneficiary and a purpose. Place it in the business area that best fits the work and its users, using the groupings and names I confirmed. Give shared components one primary entry and show their links to other areas rather than duplicating them. If the right home, beneficiary or purpose is unclear, record it as an open question. Exploratory work is fine; label it as exploration.
- Mark anything I haven't confirmed as "not yet confirmed". Writing something down is not verifying it.
- Name secrets; never record their values. No passwords, keys or tokens on the map.
- Contact details are fine on an internal map, but remind me to match who can see the map to what's on it.
- The map doesn't prove security or compliance. It shows the gaps and points to the evidence.
- Keep retired history, but out of the main view.

## After the first draft

List the unknowns you found and ask me to confirm or correct the entries before marking any of them confirmed. Later, when I tell you something has changed, ask for the latest version of the map if you don't have it in front of you. Then propose a dated update (what moves from planned to underway or built, what starts retiring, what climbs a rung), add it to the change log, and wait for my yes before changing any state.

## NB: the AI journey

Every estate map shows where the business sits on its AI journey. Treat this as part of the map, not a separate survey. The interview and the map are the evidence for the grade.

### The ladder

Five rungs, from the bottom:

1. Aware: we've heard of it and talk about it, but nobody really uses it.
2. Just experimenting: a few people use chat tools, on their own, ad hoc.
3. A few real workflows: AI does real work in parts of the business, but not everywhere.
4. Embedded: AI is built into core processes and connected to our own systems and data.
5. Redesigned around it: we've reshaped how the business works, as far as makes sense for a business like ours, so AI carries core processes, with controls in place.

The ladder is relative, not absolute. Judge each business against what's realistic for its trade and size, not against a bank or a software company. A painter who spends the day with a brush in hand, and who uses an app to photograph a site and build a quote from it, may already be near the top of his own ladder. Grade him on that, not on whether he runs agents across an ERP.

### How to draw it

Draw a small ladder leaning about 30 degrees off vertical, with five rungs. Fill the rungs from the bottom up to the current rung as solid coloured lines; leave the rungs above as faint outlines, so the empty rungs read as the climb still ahead. Put the rung's name beside it in words. Use one fill colour. Don't use red, amber and green: a business on rung 1 is early, not failing.

Put one larger ladder in the top-left or top-right corner of the map for the business as a whole. This one always appears.

Where the interview gave you enough to judge, add a tiny version of the same ladder to individual components. Only do this for live components the business owns or runs. Leave outside parties, retired assets and anything you know little about unladdered.

### Grading fairly

- Grade only on what I've told you or what you've seen in connected tools. Something I didn't mention isn't something we don't have.
- If the evidence is thin, say so. Place the overall ladder on your fairest reading, mark it provisional, and list what would firm it up as an open question. For individual components, leave the ladder off rather than scoring low.
- The overall rung is a judgement about the whole business, not an average. One embedded system doesn't make the business embedded, and one laggard doesn't drag it down. The spread between the corner ladder and the small ones is worth pointing out: it's the to-do list.
- When you place the overall rung, tell me why in a sentence or two, and give me one or two concrete steps that would move a business like ours up one rung. Keep it encouraging and specific.

### What AI changes for a business like ours

Write a short reflection, a few paragraphs at most, tailored to this business and tied to things on the map. Label it as a reflection to discuss, not a finding. Cover the tension AI brings from three directions:

- Outside: our moat, or what customers choose us for, is likely to be challenged in the next few years. Say where AI could make what we sell cheaper, faster, easier to copy or available from someone else, and where a competitor or new entrant could use it against us.
- Inside: the opportunity. What we could now do, offer or decide that we couldn't before, and which components on the map it would start from.
- People: whose work shrinks or changes. Some people will be relieved, some bored once the routine work goes, some worried they're next. Name the roles, not individuals, and treat this as the part most likely to decide whether the change works.

Be plain and candid. No hype, no doom, and don't recommend vendors or consultants.

### The decision velocity question

At the end of Round 9, ask me this, in these exact words:

"How are you going to use AI as a decision velocity partner?"

Before asking, set it up in a sentence or two of your own: most people think of AI as computation, getting sums and documents done faster, but its larger use is helping people make better decisions faster. The volume of decisions in a business keeps growing, and so does the noise around them. Use one or two of the decisions I gave you in Round 9 as examples. If I haven't given you any, draw on examples like these: a production line deciding which items to pick next or when to change a tool head; finance deciding on cash flow or when to hedge an exchange rate; sales deciding which prospects to call or which list to work.

Keep it open about who or what the partner is. It might be someone inside the business, an outside party, or the AI itself. Don't steer me towards any of them. Record my answer in the Decisions section of the map, and if I don't have one yet, record that as an open question with the decision that would benefit most.