wild navigation
ConfidentialTM-2026-SAMPLE
The Stonepile Map
Where AI earns its place in a 28-person trades business, and where it must not go. The work of one Terrain Day and a week of written synthesis.
wild navigation
Where AI earns its place in a 28-person trades business, and where it must not go. The work of one Terrain Day and a week of written synthesis.
Part I
The question you walked in with, the one that surfaced once we started mapping, and the forces shaping both at once.
Should we put in one of those AI receptionists to catch the calls we’re missing, and what other AI should a shop like ours be using by now?
This was the question on the whiteboard at 8:30 in the morning. It is a fair question, and the missed calls are real money. It is also not the question that has been keeping you up, which surfaced about two hours later, when we asked what happens to Stonepile the day Marlene retires.
How does this business run without me, and without Marlene, so that in a few years I can step back or sell it, without it falling apart or losing the thing our name is built on?
You have built Stonepile over twenty-two years into the shop people in three counties call first. Marlene has been the dispatcher for nineteen of them, and she holds the schedule, the customer histories, the which-tech-for-which-job judgment, and the after-hours triage almost entirely in her head. She is planning to retire inside eighteen months. That clock, more than any AI question, is the real terrain.
The AI receptionist was the socially acceptable way to ask the succession question. Both are worth answering. Only one of them is load-bearing.
These are the structural forces we named together at the whiteboard. None is unique to Stonepile; every trades business at your stage is reading the same ground. Your work is to meet them on your own terms, not to accept a vendor’s read of them.
After-hours and in-field calls go to voicemail, and in a market where a homeowner with a failed furnace calls the next shop ninety seconds later, voicemail is lost revenue. This is your fastest, cleanest AI opportunity, and we will get to it. It is deliberately not first.
Marlene is the single point of failure and the single most valuable asset, at the same time. Nineteen years of judgment about customers, equipment, and crews is undocumented. This is the most time-sensitive variable on the whole map, and it gates almost everything else. You cannot automate a dispatch process that has never been written down.
Two of your younger techs are quietly using a chatbot on their phones to look up codes and draft customer texts. That is not a problem to shut down. It is unpaid product research telling you exactly where the official tools are failing the people doing the work.
Customer records live in three places: a paper day-book at the front desk, the accounting software, and an old dispatch tool nobody fully trusts. No single source of truth means no reliable ground for any AI to stand on. AI amplifies the data you have; it does not clean it for you.
Trades margins make the return case tight, which is a gift in disguise: it forces simplicity and rules out the expensive, complicated tools a city consultant would sell you. The right moves here are cheap and boring, and they pay back in weeks.
In a county this size, one bad public failure travels two towns over before the invoice clears. The upside is the mirror image: one thing that visibly works earns three referrals from a single conversation at the hardware store. Your reputation is the distribution channel, which sets a hard rule on what AI is allowed to touch.
Part II
Where Stonepile stands across the four things that decide whether AI holds: your leadership, your team, your processes, and your data. Each reads one layer of the WILD framework at the scale of a business.
Readiness is not a single number, because a single number lies. We read four tiers, and the lowest one sets your first move. The strengths here are real. The edges are where AI would break if you moved on it today.
You are unusually change-ready, and the succession clock gives you a real reason to move now rather than someday. When the owner is the one pushing, projects ship. That is the single largest predictor we look for, and you have it.
You have not used any of these tools yourself yet. A crew reads what the owner does, not what the owner buys. Until you are visibly using the system, adoption will stall at the truck level.
Your younger techs are already experimenting. That shadow-AI signal is appetite, and it is the raw material of adoption. You have early movers on the payroll and did not have to hire them.
Marlene reasonably fears that all of this is a plan to replace her. It is not, and the whole sequence depends on her knowing that. Her nineteen years are what we are trying to capture, not erase, and that has to be said to her directly, early, and by you.
Billing and payroll are clean and consistent. Where a process is already written down and stable, you are ready to build on it, and that gives us a place to prove the approach without risk.
Dispatch, the heart of the business, is undocumented and lives with Marlene. This is the lowest tier on your map and therefore the first thing to fix. Everything AI might do for scheduling and customer response sits on top of a process that does not yet exist on paper.
The accounting data is trustworthy and current. Nothing there needs fixing before it can be useful.
Customer records are split across paper, accounting, and the old dispatch tool, with no single source of truth. This does not block the first move, but it gates the second, and it is the quiet reason past attempts to "just add software" never took.
The four tiers do not act alone. In your case the chain is clear, and naming it is half the value of the Day.
Your Process edge, the undocumented dispatch, is holding back everything else. Because the process lives only in Marlene’s head, no tool can safely touch scheduling or customer response, so the missed-call fix you came in asking about cannot be built on solid ground yet. And because the data is scattered, even a documented process would have no clean record to run on.
This means the lever is Process: capture what Marlene knows, on paper, first. Do that, and the data cleanup gets a target, the missed-call fix gets a foundation, and the succession question gets its first real answer. Skip it, and you would be automating a business that still cannot run without one person, which is the exact opposite of what you are trying to buy.
Part III
What to do, in what order, and just as important, what not to do yet. The order is not optional; skipping steps is the most common way this fails.
Every opportunity we found, sorted into go, fix-first, or not-yet, with the reason. The not-yet list is written before the go list on purpose. Naming what AI should not touch is what earns the right to say where it should.
Capture Marlene’s dispatch knowledge into written SOPs. The foundation for everything. Start here, this month, while she is still in the chair and willing.
After-hours and missed-call coverage, once dispatch is documented. Your highest-return move, and a category of tool that pays for itself in captured jobs within weeks.
One source of truth for customer records. Consolidate the paper, accounting, and dispatch data into one place the whole team trusts. Enables everything downstream.
Give the techs a sanctioned tool for codes, drafting, and lookups, with a simple use policy. Turns shadow AI into something you can see and stand behind.
Automating the customer relationship calls. The trust conversation with a worried homeowner is your product. A machine does not take that call. This line does not move.
A full dispatch-automation platform. Right idea, wrong time. Document and stabilize the process first, or you are just paving a cow path at software prices.
Tell Marlene, first and in person, what this is. That the goal is to honor and capture her nineteen years so the shop can carry them, not to replace her. The whole sequence depends on her being a willing partner. This is your conversation to have, not ours.
Run the knowledge-capture sprint. Sit with Marlene, in short sessions, and get the dispatch judgment onto paper: how she triages a callout, how she matches a tech to a job, the customers who need a particular touch. A recorder and a patient interviewer are most of the tooling. AI helps turn the recordings into clean SOPs, but the value is in the conversation, and the conversation is human.
Sanction one tool for the techs, with a one-page use policy: what is fine, what stays off it, what never goes into it. The early movers become the people who help the rest of the crew, which is how adoption actually spreads in a shop.
Stand up after-hours call coverage. With the dispatch process documented, a call-answering setup can now capture the jobs you are losing to voicemail, hand off cleanly to your on-call tech, and log every call where it belongs. Start narrow, measure one number, the booked jobs from previously-missed calls, and let that number make the case.
Consolidate to one source of truth. Move the day-book and the old dispatch records into one system the team trusts, so the next layer has clean ground to stand on. This is unglamorous and it is the difference between a system that holds and one that quietly rots.
The orienting question is not "how much can we automate." It is "what would make Stonepile sellable, or runnable without you, in three years." Every move above is a down payment on that answer. A documented operation with captured knowledge, clean records, and a few reliable systems is worth materially more to a buyer, and materially less dependent on any one person, than the shop you walked in describing.
The AI receptionist you came in asking about is on the list. It is third, and it is better because it is third. By the time you turn it on, it will be answering with your process behind it, not instead of it.
Part IV
What to build, what to keep human, and how to keep this alive after we step back. Tool-agnostic on purpose; the category matters, the specific product changes by the quarter.
When the after-hours coverage goes live, it runs as a narrow 30-day pilot with one named owner and one number to move: booked jobs from calls that used to go to voicemail. At day thirty the question is never "is AI useful." It is "did that number move by what we said it would." If it did, you keep it. If it did not, you stop, and you have lost thirty days, not a year.
Every AI-facing step has a person at the decision point. Nothing books a job, quotes a price, or answers a distressed customer without a human in the loop. The machine drafts and routes; your people decide. This is not caution for its own sake. It is where your reputation lives.
Hand it: turning recordings into SOP drafts, first drafts of customer texts and quotes, after-hours triage that routes to a human, the tedious reconciliation of records. Keep for yourselves: the judgment calls on a hard job, the relationship with a worried homeowner, the decision to hire or let go, the trust conversations that are the whole business. The test is simple. If handing it off brings relief, delegate. If handing it off feels like the work disappearing, keep it.
The capability has to live in the shop, not with us. One person, likely one of your early-mover techs, owns the tools and the fifteen-minute monthly check that keeps them honest as the field moves. We build alongside that person, not around them, so that when we are gone the system is theirs.
One introduction we would make: a trades owner two valleys over who ran a nearly identical succession-and-AI question three years ahead of you and came out the other side with a business he could actually sell. He has offered to take a call from anyone we send. The best learning here is not a manual, it is a conversation with someone who has already crossed this ground.
Part V
What happens next, and how to keep this Map alive as the ground keeps shifting.
This Map is a snapshot, not a plan set in stone. Some of what we named will prove exactly right. Some will shift once you start moving, because that is how real ground behaves. The sequence is what holds; the specific tools inside it will change, and that is expected.
Included with your Terrain Day. A call, thirty days out, to see what moved, what stuck, and what surfaced that was not on the Map. If the knowledge-capture sprint is underway and Marlene is on board, you are already ahead of where most shops get in a quarter.
The Terrain Map is the diagnostic. The building is the Groundwork Sprint: six to eight weeks, working from this Map, standing up the priority systems with your team so the capability stays in the shop. Because you took the Terrain Day, the $5,000 credits in full toward the Sprint if you go within ninety days. Some owners take the Map and run with it themselves, and that is a real answer too. Either way, you leave today knowing the first move and the reason for it.
Chris & Andrea Greene
The WILD Navigation Company · Crested Butte, Colorado
A Final Word
Most of what we named, you already sensed. The Map is not new information so much as your own knowledge, written down and harder to defer now that it is. The missed calls are real, and we will fix them. But the clock on Marlene’s chair was always the real terrain, and now it has a plan.
You will get this business to where it can run without you. It is already starting. The Map just makes the path visible.
Chris & Andrea Greene
The WILD Navigation Company · Crested Butte, Colorado