The Only Free Variable

Here's how I would set it up.

Stop selling hours. At [my hourly rate], every time AI makes you faster, you earn less. Your clients also don't want hours. They want the thing finished and working. So price the finished thing, and keep the hourly rate in the background.

Use three tiers that lead into each other.

1. The first purchase: a fixed-price delivery. One finished work product with a clear scope and a clear end, like a market analysis, a menu and pricing review, or a competitor scan. Charge $750 to $1,500. This tier isn't where your money comes from. It lets a skeptical owner try you without much risk, and it shows them what "finished" means from you.

2. The core offer: build it, then run it. This is the business. You put a working agent into their operation, like a phone receptionist, a daily scout that watches their market, or a morning briefing email. Charge a one-time setup fee of $2,500 to $7,500 depending on complexity, then $400 to $1,200 a month to host it, monitor it, fix it and keep improving it. The monthly fee builds the company. A shop that pays you $800 a month for three years is worth far more than a $5,000 project.

3. The top tier: a fractional AI department. Charge $2,500 to $5,000 a month for a set number of deliverables or builds each month, priority response, and a monthly review with the owner. Offer it only to clients who already pay you for a running agent and keep asking for more.

How to set the actual numbers:

-- Start with what the work replaces. If the receptionist agent covers calls that would need a part-time person at about $2,500 a month, then $900 a month is easy to say yes to. Aim to charge roughly 10 to 25 percent of the value you create in the first year.

-- Use your hourly rate as a floor, not as the price. Estimate your hours, multiply by [my hourly rate], then multiply by 2 to 3. If the value-based price comes in under that, the project is too small or the scope is wrong.

-- Put pass-through costs (API usage, phone numbers, hosting) into the monthly fee with a margin. Don't bill them line by line, because owners hate surprise usage bills. Write a reasonable usage cap into the agreement.

-- Bill monthly fees in advance. Take 50 percent of the setup fee upfront. Require a 3-month minimum, then month-to-month. Month-to-month sounds risky, but it's what gets a nervous owner to sign. If the agent keeps working, they'll stay.

-- Decide ownership up front: the client owns their data and outputs, and you own the templates and the system underneath. That's what lets you reuse your work.

Two strategic moves matter more than the exact prices:

Build templates for one industry. Every receptionist you build for a restaurant should make the next one faster. The third version should take a fraction of the time the first one did while selling at the same price. That's where your margin comes from. Hospitality is a good first industry because the pain is obvious: missed calls, missed reservations, staff stretched thin.

Raise prices on purpose. If the first five owners say yes without pushing back, you're too cheap. Raise prices 25 to 30 percent for the next five. Keep going until about a third of people hesitate.

Your physical space also makes a natural funnel: a free or cheap session there leads to a fixed-price delivery, which leads to a running agent. Most of your marketing should push people to that first step.

Keep [my hourly rate] for one thing only: work that falls outside a defined scope. Quote it so clients see that change requests cost money. That keeps the fixed-price work fixed.

Start with what a local owner actually buys. They aren't buying AI. They can't judge whether a model or prompt is any good, and they don't want to learn how. They can judge whether the phone got answered, whether the table got booked, and whether they got Saturday morning back. So the unit you sell is an outcome, described in the words they already use for their own problems. "AI agent" isn't on their list of problems. "We miss calls during the dinner rush" is.

Next, look at the owner's real cost of saying yes. Money is only part of it. The other costs are attention (meetings, logins, learning something new) and risk (looking foolish, a bot embarrassing them in front of a customer). Most AI vendors make all three worse. Your offer should push all three close to zero: you do the work, the owner gets a finished result, the price is fixed and known up front, and there's a guarantee.

Then look at your own costs. AI has collapsed the time it takes to produce the work. If you bill by the hour, you get paid less the better you get, so hourly billing works against you on anything you deliver. Keep [my hourly rate] only for advisory work that genuinely can't be defined in advance, and even there treat it as a floor.

Fourth, your marginal cost falls with repetition. The first AI receptionist for [a restaurant client] is a custom build. The fifth is mostly reuse. That argues for productizing by vertical: build once for a restaurant, a contractor or a marina, then sell the same proven thing to every similar business in town at a fixed price. Every repeat raises your margin without raising the client's risk.

Fifth, AI work doesn't stay finished. Models change, menus change, hours change, staff turn over, and edge cases show up. Real ongoing work exists, which is what makes recurring revenue honest instead of a toll. It's also where the value of your business builds up.

Last, local presence is your moat. A national SaaS company can't walk into the restaurant, sit with the manager and fix it on Tuesday. Trust travels by referral and by your face. Your pricing and packaging should make it easy for one happy owner to tell another exactly what they bought and what it cost.

Put those together and the structure follows. I'd set it up as a three-step ladder.

1. Front door: a fixed-price diagnostic that is itself a deliverable. Spend a few hours inside the business and hand back a finished artifact, not a slide deck: the specific workflow you would automate, a working prototype or recorded demo on their own data, and the dollar case. Price it around $500 to $1,500 and credit it toward the build if they go ahead. This filters out tire-kickers, gets you paid for discovery, and leaves the owner holding something real. Your public Intelligence Deliveries already do a free version of this as marketing. The paid diagnostic is the same idea done on their business.

2. The build: a productized, fixed-price deliverable per vertical. Price it against the value it creates, not the hours it takes. A useful rule is that the owner should be able to see a 5 to 10x return in the first year, and you should capture roughly 10 to 20 percent of the value created. Here's an illustrative example to show the method, not real data. Say a restaurant misses 15 calls on a busy night, a third of those would have booked, and the average party spends $150. That's well over $30,000 a year walking out the door. At that level, a $3,000 to $7,500 setup is an easy yes, and you haven't priced off your time at all. Publish a starting price for each vertical product. Publishing prices builds trust locally and saves you from rerunning the same sales conversation.

3. The run: a monthly operating fee for every build. Somewhere between $250 and $1,500 a month depending on complexity, covering monitoring, tuning, updates when the business changes, and a short monthly report that shows what the system did ("answered 412 calls, booked 61 reservations, flagged 3 complaints"). That report matters more than it looks. Recurring fees die when the value becomes invisible, so make the value visible every month. Bundle the underlying software and API costs into this fee rather than passing them through line by line. Owners want one number.

A few rules to hold onto:

Guarantee the outcome on the build. Something like: if it doesn't do X within 30 days, you fix it or refund the setup. Your confidence removes the owner's risk, and for a small local business that's often what tips the decision.

Don't discount. Change the scope instead. If the price is too high, offer a smaller version of the outcome, not a cheaper version of the same one.

Let the client own their accounts and data. You own the know-how and the build pattern. That fits the Personal AI and BYOAI position you already argue for, and it builds trust because nobody feels locked in. They stay because the service is good, not because leaving is hard.

Say no to one-off custom projects at fixed prices unless they could become a product you sell again. Every build should either pay well on its own or teach you something you can resell to the next ten businesses like it.

Size the business by run clients, not builds. Work out how many monthly clients you and your team can serve well, multiply by your average run fee, and check whether that covers the business before the build revenue. If it does, builds are profit and growth. If it doesn't, either raise the run fee or narrow your verticals until delivery is efficient enough.

In short: sell outcomes rather than AI, price on value rather than time, productize by vertical, and turn every build into a recurring relationship that shows its value every month.

Here is the procedure run on "success" exactly as you gave it, with no category assumed.

STEP 1 -- IRREDUCIBLE PARTS

1. A target condition. This is a state of the world that can be checked as true or false. Without it, "success" can't be told apart from anything else.
2. A holder of the target. Someone's expectation decides what counts. Success is always success for someone.
3. A gap. The current state has to differ from the target state. If there is no gap, nothing has to happen.
4. Action. Something has to change in the world to close the gap.
5. Time. The change takes place over some duration, however short.
6. Cost. Energy, money, or materials get used up by the action.
7. Attention. Someone's attention has to point the action at the target, even if only once to set something running.
8. Verification. The holder has to learn that the target state now exists. Before that, the outcome has happened but success hasn't been registered.
9. Correction. Under uncertainty the first action may miss, so there has to be a way to see the miss and adjust.

Things that usually come along with it:

10. A written plan or strategy
11. A large, distant target
12. A target set by someone else (family, market, peers, institution)
13. Comparison to others as the measure
14. Recognition or approval from others as the proof
15. Effort or sacrifice in proportion to the result
16. Motivation, discipline, or mindset as a prerequisite
17. Credentials, permission, or a mentor before starting
18. Failure first, as a required stage
19. A single success per domain or per life
20. Success treated as a permanent state you reach and then hold
21. Success treated as something you then spend to get happiness, security, or rest
22. A target fixed before acting and never revised

STEP 2 -- LAW OR CUSTOM

1. Target condition -- LAW (logic). You can't hit what isn't defined.
2. Holder -- LAW (human attention). A judgment needs a judge.
3. Gap -- LAW (logic). No gap means no event.
4. Action -- LAW (physics). States don't change without something acting.
5. Time -- LAW (time).
6. Cost -- LAW (physics and money). Action uses up something.
7. Attention -- LAW (human attention). At least one aiming act is required.
8. Verification -- LAW (attention). Unobserved success is just an outcome.
9. Correction -- conditional LAW. It's required only where results are uncertain. Since that is nearly everywhere, treat it as law, with a size that depends on how uncertain things are.
10. Written plan -- CUSTOM. It's a tool for coordination and memory, not a requirement.
11. Large distant target -- CUSTOM. Physics sets no minimum size.
12. Target set by others -- CUSTOM. The holder in part 2 can be anyone, including you.
13. Comparison to others -- CUSTOM. Part 1 only needs a checkable condition. A rank is one option.
14. Recognition by others -- CUSTOM. Verification (part 8) only requires the holder to know.
15. Effort in proportion to virtue -- CUSTOM. Cost scales with the size of the gap, not with how deserving you are.
16. Motivation and discipline -- CUSTOM. They're needed only when the action has to be sustained and voluntary. A short action or an automated one needs neither.
17. Credentials or permission -- CUSTOM, unless an actual regulation applies to the specific action.
18. Failure first -- CUSTOM. It's a story told afterward. Correction (part 9) only requires the ability to adjust.
19. One success per domain -- CUSTOM (usage assumption).
20. Permanent state -- CUSTOM (usage assumption). Under law, success is a match between state and target at one moment. The world keeps moving after that.
21. Something to spend later -- CUSTOM (usage assumption). Nothing in the parts says what success is for.
22. Fixed target -- CUSTOM. Part 1 requires a target to exist when it's checked, not that it never changes.

STEP 3 -- SIMPLEST ARRANGEMENT USING ONLY LAW PARTS

The laws leave exactly one free variable, and that is the target. Once the target is chosen, everything else follows from it: the gap is set by the target, the cost and time are set by the gap, and verification is set by how checkable the target is.

So the minimal arrangement is:

- Hold your own target, phrased as a condition you can check directly with your own senses or a simple measurement.
- Size the gap so the cheapest single action you can take today closes it.
- Spend one act of attention to aim, then act.
- Check it. If it hit, success is registered. If it missed, adjust the action, not the story, and act again.
- When it's registered, it's finished. Set a new target if you want another one.

Repeated, this becomes a loop of small, checkable, self-held targets, each closed and verified on its own timescale. Big outcomes come from this loop compounding, not from one large target held for years.

The honest limit: by law alone, the cheapest possible success is a target that is already true, with zero gap and zero cost. Nothing in physics, money, time, or regulation rules that out. What rules it out is that you wouldn't care. So caring about the target is a real requirement, but it belongs to the holder in part 2, not to the size of the target.

STEP 4 -- HOW IT'S NORMALLY DONE, AND EVERY DIFFERENCE

The normal way works like this. Pick a large, distant target, usually one defined by the surrounding culture: money, title, status, a recognized achievement. Write a plan. Sustain years of disciplined effort, fueled by motivation. Measure progress against peers. Count failures along the way as proof of seriousness. Declare success when other people recognize it. After that, treat it as a permanent position you defend and draw down on.

The differences:

1. Who holds the target. Normally it's inherited from others, and the person executing often isn't the person whose expectation defines success. Mine puts the holder and the actor in the same person. That removes a whole class of failure: closing the gap and still not feeling it was success, because the target belonged to someone else.

2. Target size. Normally it's large on purpose, because size is read as ambition. Mine is sized to the cheapest action that closes it. The law parts carry no premium for size, and a large gap only adds time, cost, and uncertainty, which means more correction cycles before any verification at all.

3. How checkable the target is. Normal targets are often vague ("successful," "secure," "made it"), so verification never quite happens. Mine requires a target you can check directly, because verification is a law part. A target you can't check means success can't be registered, however much has actually been achieved.

4. The plan. The normal way depends on a long written plan. Mine has none, only aim, act, check. The plan is a custom that stands in for frequent verification. When the target is close, the check replaces the plan.

5. Motivation and discipline. The normal way treats these as the fuel. Mine barely needs them, because short actions don't require sustained voluntary effort. Discipline is a cost the normal way creates by making targets distant.

6. Comparison and recognition. The normal way measures against others and needs others to confirm. Mine needs only the holder to verify. Social proof is a second, optional target layered on top. It often becomes the real target without anyone noticing.

7. Effort and sacrifice. The normal way treats effort as earning the success. Mine treats cost as whatever the gap requires and no more. Paying extra doesn't increase the success. It only raises the price.

8. Failure. The normal way treats failure as a stage on the path. Mine treats a miss as information for the next action. Same physical fact, but no identity or narrative is attached.

9. Permanence. The normal way treats success as a place you arrive at and keep. Mine treats it as a moment when state and target match. Holding a state is a new target with its own gap and cost, usually called maintenance.

10. What it's for. The normal way assumes success gets converted into happiness, security, or rest later on. Mine makes no such assumption. That leads to the most common complaint about normal success: arriving and finding the conversion didn't happen. Nothing in the law parts promised it would.

11. Count. The normal way imagines one big success, or a few. Mine produces many, continuously. The law parts put no limit on how many times the loop can run.

12. Fixed target. The normal way holds the target constant for years and treats revising it as giving up. Mine lets the target be redefined at each check, because the only law requirement is that a target exists at the moment it's checked.

Once the customs are removed, success isn't scarce or hard. It costs exactly as much as the gap you chose. Most of the difficulty people associate with success comes from choosing large, vague, borrowed, permanent targets, then paying for the motivation, plans, and recognition those choices require.

Run this again on a specific outcome and the target, gap, and cost turn into concrete numbers. That's where the exercise is most useful.

I've kept the category out of steps 1 to 3 and named it only in step 4, as you asked.

STEP 1 AND 2 -- IRREDUCIBLE PARTS, MARKED

1. A job exists that the owner has been avoiding. LAW (human attention). They avoid it because, to them, doing it costs more attention or dread than having it done seems worth. That gap is the only reason they would pay anyone.

2. The owner has to know you exist at the moment they feel the avoidance. LAW (attention). How they come to know you is CUSTOM.

3. The inputs move from the owner to you: information, materials, access, and the one or two judgments only they can make. LAW (information has to move). How it moves is CUSTOM: meetings, calls, forms, email threads.

4. The handoff has to cost the owner much less attention than the job itself. LAW (attention). If handing it over feels like doing it, they will avoid the handoff too. People usually assume a little friction at intake is fine. For this person it isn't.

5. Both sides need a shared definition of "finished" before work starts. LAW. Without one, the deadline and the price both float, and a floating price is where negotiation comes from. Writing it down as a proposal or statement of work is CUSTOM.

6. The owner trusts you enough to hand it over. LAW (human risk calculus). They are exposing access, data, money or reputation. How you earn that trust (meetings, references, contracts) is CUSTOM.

7. You do the work in 6 hours or less of your attention. LAW (time). It follows that only jobs you can bound in advance qualify. A job you haven't done before can't be promised in under 6 hours. People usually assume the owner defines the job. That's CUSTOM.

8. Elapsed time is 7 calendar days or less. LAW (time). Every question you send the owner spends days of that week, because they reply when they reply. How many round trips you take is CUSTOM.

9. The work comes back already in place, needing nothing more from the owner. LAW. If they still have to install, implement, file, send or decide something, the avoided job still exists and it isn't finished. Delivering a report, recommendations or a draft is CUSTOM.

10. The owner can confirm it's finished cheaply, in minutes. LAW (attention).

11. A price exists, and the owner believes it is below what the finished outcome is worth to them. LAW (money).

12. The price is known and accepted before the work starts. LAW, given your "without negotiating" condition. People negotiate when the price is uncertain or comes after the fact. Quoting after discovery is CUSTOM. Hourly billing is CUSTOM.

13. Money actually changes hands. LAW (money). Invoices, net-30 terms and deposits are CUSTOM. Chasing payment spends your hours, and an unpaid invoice invites the owner to reopen the price.

14. The price covers your floor: price divided by 6 hours is at least your hourly floor. LAW (money). At [my hourly rate], that's [six times that] at minimum. Value to the owner sets the ceiling, not your hours.

15. Taxes, liability and any licensing that applies to the work. LAW (regulation). Long engagement contracts and master agreements are mostly CUSTOM. Enforceable terms and income tax are not.

Assumptions about how it's used, all CUSTOM:
-- the owner must meet you
-- each job is new and scoped from scratch
-- the owner reviews drafts and asks for revisions
-- the owner needs to understand how it was done
-- payment comes after an invoice
-- the job leads to an ongoing relationship
-- your 6 hours are done by hand, in one sitting
-- you take whatever job the owner brings

STEP 3 -- THE SIMPLEST ARRANGEMENT THAT SATISFIES ONLY THE LAW PARTS

1. A fixed list of finished outcomes, possibly just one. Each is named in the owner's words for the thing they're avoiding, not in your words for the method. Each has a posted price, a written definition of done, and proof from earlier runs that it fits inside 6 hours. This one choice settles parts 5, 7, 11, 12 and 14 at once.

2. Intake is one exchange that takes the owner under 15 minutes. It collects everything needed: access, materials, and the few decisions only they can make. It's designed so you never need to come back with a question. You allow at most one round trip, and that slack is planned into the week.

3. Trust comes from making the downside small, not from meetings. The price is fixed and you don't charge until it's done. The owner gives a payment method at handoff, and you charge it only when they accept delivery. That removes negotiation, collection and most of the trust problem together.

4. Your 6-hour budget, set in advance:
-- 30 minutes reviewing intake
-- 4 hours of work
-- 30 minutes checking it against the definition of done
-- 30 minutes delivering it in place
-- 30 minutes held back for the one allowed question

5. The calendar: start within a day of handoff and deliver by day 3 or 4. That leaves days 5 to 7 as slack for the owner's slow reply. If you need them, the week is safe. If you don't, they get it early.

6. Delivery is the job finished where it lives (sent, filed, live or installed), plus one screen showing it's done. The owner checks that screen in 5 minutes and accepts, and the card is charged.

7. You decline anything not on the list. A new job type goes on the list only after you've done it once, timed it, and fixed its definition of done.

STEP 4 -- HOW IT IS NORMALLY DONE, AND EVERY DIFFERENCE

Now I can name it. This is a done-for-you service engagement: freelance, contractor or consulting work for a small business. The usual sequence:
1. Discovery call
2. Scoping meeting
3. Custom proposal or statement of work
4. Quote, then back-and-forth on price
5. Deposit or contract
6. Kickoff
7. Drafts and revision rounds
8. A deliverable, often advice or a draft the owner still has to act on
9. A net-15 or net-30 invoice
10. Follow-up to collect
11. An attempt to turn it into a retainer

The differences, one by one:

1. Custom scoping versus a fixed list. The normal way scopes each job because the provider sells hours and has to price the risk of the unknown. Yours moves that risk onto you, which you can only carry because every item on the list is already proven to fit in 6 hours. The cost is that you turn work away. The normal way takes it, then runs over time and over budget. Novelty is what breaks the 6-hour limit, and the fixed list is what keeps novelty out.

2. Discovery meetings versus one intake exchange. Meetings exist to build trust and to sell. But they spend the one resource the owner is already short of, attention. That's the same reason the job was avoided in the first place. You get trust from a fixed price plus no charge until done.

3. Quote, then negotiation, versus a posted price. People negotiate when they can't tell whether a price is fair or when the scope is unclear. A posted price, a written definition of done and no-charge-until-accepted remove both reasons. Haggling isn't a personality trait of the owner. The way the price is presented produces it.

4. Hourly billing versus a fixed price. Hourly billing punishes you for being fast and gives the owner a reason to question every line. Under 6 hours only pays well if speed is your margin, so the price has to come from the value of the outcome, not your time.

5. Revision rounds versus delivering once against a fixed definition of done. Each revision round costs days and owner attention. A definition of done agreed upfront makes the review a yes-or-no check rather than a matter of taste.

6. Delivering advice or a draft versus finishing it in place. This is the biggest gap. The normal way often hands back recommendations, a plan or a file to implement, which means the owner still has the job they were avoiding. By your own definition of the outcome, that isn't finished.

7. An invoice on net terms versus a card charged at acceptance. Chasing payment eats your hours and reopens the price after the work is done. Taking the payment method at handoff, before any work, means the money moves the moment they accept.

8. A long contract versus short terms attached to the offer. The legal parts are real law and stay: tax, liability, and any licensing the specific job needs. A multi-page agreement for a sub-$2,000 fixed-scope job is custom.

9. Pushing for a retainer versus none. A retainer is a business-model preference, not part of this outcome. Repeat business comes from the owner coming back for the next avoided job on your list, not from a monthly commitment.

10. Marketing on credentials or method versus naming the avoided job. The owner has to recognize their own avoided job at the moment they feel it. They search for and remember the job, not the tool or method you'll use.

11. Taking whatever the owner brings versus declining what isn't on the list. This custom is the hardest to break and does the most damage. Saying no is what keeps every engagement under 6 hours and inside the week.

Where the normal way is right: genuinely new, one-off, high-stakes work can't be bounded in advance, so it needs custom scoping and probably a meeting. Some owners also won't hand over access to anyone they haven't met. Offer a 10-minute call as an option, never a requirement, and keep the price and definition of done fixed either way. Anything that doesn't fit the list is a different product with different terms, not an exception to this one.