The Money plane · second engine
Turn workflows into money
Everyone is trying to bolt AI onto a mess and call it a strategy. The people actually making money do it backwards: they find the real process first — the one in the log, not the one in the manual — and only then decide what gets automated.
Here's the secret the consultants charge for: the waiting is the process. The work is minutes. A bill that takes 24 days to pay involves maybe 20 minutes of actual work. The other 23 days, 23 hours and 40 minutes are chasing, waiting, re-asking and re-doing. That's where the money is hiding — and it doesn't take a genius to find it. It takes someone who knows how the work actually runs.
Spoiler: that's you.
The method, in order
Six steps. AI is step four.
Find the real process
The engagement letter says six steps. The practice log says fourteen. Interviews give you the why — why a rule exists, who really decides, which step is theatre. The records give you the what — every case, every hand-off, every wait. You need both before you touch anything.
Measure the waiting, not the working
Time the whole journey, start to finish — then time the actual work inside it. The gap is the opportunity. In one accounts-payable process, the cycle ran 24 days while the touch time was measured in minutes. Nobody had ever written both numbers down side by side.
One process, one owner
Somebody who can sign the change across every department it touches. Without an owner, every improvement dies in a hallway conversation. Nothing about this step is AI. It's the step most people skip — and the reason most AI projects quietly fail.
Sort every step — in this order
Delete it. Then plain code. Then an agent. Then a human. In that order, always. In the accounting-firm case, five of fourteen steps weren't automated — they were deleted. Most people start at the model. The pros start at delete.
Baseline before you build
Write the number on the wall before you change a thing: steps, days, cost per unit. Same policy, same approvers, full audit trail. If you can't measure it before, you'll never prove it after — you'll just have an opinion and an invoice.
Build once. Deploy everywhere. Measure again.
One fix, built properly, rolled out to every place that process runs — that's how a private-equity firm put the same month-end close into five companies that were each doing it a different way (12, 9, 15, 18 and 13 steps became one 6-step definition). Twelve months later, the re-measured number is the only proof it worked.
What it looks like when it works
17 steps became 7. $31 became $6.
One accounts-payable process, mapped from its own transaction log: 17 steps became 7, six exception loops became one, the cycle dropped from 24 days to 6, straight-through processing rose from 18% to 87%, and the cost per invoice fell from $31 to $6. Same policy. Same approvers. Full audit trail. Nothing exotic — just the waiting, removed.
And it's already an industry. Investment funds are buying accounting firms, IT shops and law practices, putting engineers inside, and re-engineering the work this way. The arithmetic is simple: a services firm is valued on its profit, so double the margin at the same multiple and the firm is worth about twice as much. The workflow is the asset.
Honest footnote: the figures above come from Varick Agents, a firm that sells this work — treat them as direction, not gospel. The method stands on its own; you can run it on any process and measure your own numbers.
Why this is a retiree's game
Three people in one. You're the rare one.
The job title of the moment is the forward deployed engineer — and it's really three people in one:
The operator — understands how the work actually runs. Finance, sales, operations. Knows which step is theatre and who really decides. The engineer — ships working software into the systems a business already uses. The AI engineer — knows what a model can be trusted with, and builds the guardrails to prove it.
Almost nobody holds all three. Consultancies have the operator. Software firms have the engineer. AI labs have the third. But look at which chair is scarcest — and which one can't be downloaded. A thirty-year career is operator training. You have sat through the meetings, worked the workarounds, and learned where every process actually breaks. The other two chairs? That's what your AI agent is for.
You already hold the one skill AI can't supply. The workflows are the way you cash it in — without going back to a job.
The second play: own the list
The Mine
There's a smaller, scrappier version of the same idea, built for one person and a laptop. Pick one niche in one city — a trade, a service, a recurring need. Then go where the demand is already talking: Facebook groups, Marketplace, Instagram. Buyers post "I need help this week." Sellers post cancellations — "one slot left today" — to audiences that never see them. Great supply, terrible distribution.
Your agent does the watching: it reads the posts, structures the names, needs and prices, drafts the messages, and books the match. You set the rules and approve. And here's the part that matters: you own the asset — a clean, priced, trusted list that lives off the platforms. Matches and bookings are the revenue. The list is the equity.
Mine
Find the gold in other people's groups, listings and comment sections. The demand is already there, unanswered.
Own the list
Take it off-platform into your own clean list — names, needs, prices, proof. This is the step that turns hustle into an asset.
Own the room
Your group, your page, your storefront — the place the niche checks first. Trust compounds here.
Build the technology
Only now ask: what would automate this end to end? Most people never need this step. The first three pay on their own.
Start here — this week, no AI required
Do this next week
One thing a day, on your own life, before any AI is involved. This is the Freedom 101 warm-up — bring your Friday result to the workshop.
Monday — list every app that holds your stuff
Five inboxes, four file stores, three logins for one restaurant app. Write down which one wins when two disagree. Most people have never seen their own system on one page.
Tuesday — pull 20 things you did last week
Paying a bill, booking a table, filing a receipt. Count how many went the same way twice. Almost none will. That variance is exactly what you're going to fix.
Wednesday — write one process down, step by step
How you pay a bill, from the email to the bank. Every click, every tab, every wait. On paper. It will be longer than you think — that's the point.
Thursday — sort every step
Delete it, give it to code, give it to a model, or keep it yourself. In that order. Most people start at the model — you now know better.
Friday — do it to a business you know
A dentist, a law firm, the shop on the corner. Time the work against the waiting. Then build — or sketch — the smallest thing that removes one step. One step. That's a business.
Honest limits
- This page teaches a method, not a get-rich-quick scheme. Real processes, real measurement, real work.
- The case figures are a vendor's (Varick Agents) — composites with details changed. Run the method and trust your own baseline.
- The Mine runs on other platforms' rules. Accounts get flagged, terms change — own your list off-platform, and never build on rented ground alone.
- Conal is not a licensed advisor, and this isn't business or financial advice for your situation. It's how to think — the prime rule applies here too.