/ Pricing
How I price
I do not have a price list. That is a deliberate choice, and this page is me showing my working so you can hold me to it.
Here is the short version. A fixed-price Discovery is $3,000. A bounded build runs $5,000 to $25,000. Ongoing work runs one of two ways: a baseline retainer from $4,000 a month, or an embedded engagement from $9,500 a month. Where you land comes down to three things: how complex your systems are, how much risk the work carries, and how much you actually want AI to do in your organisation.
The rest of this page is what those three things mean in practice, and why the step between $4,000 and $9,500 is not what it looks like.
Why there is no fixed price list
Almost nobody in this market publishes how they price. The usual reason given is that every business is different, which is true and also a cop-out, because it tells you nothing.
The real reason is that a fixed price list only works when you are selling a fixed thing. A website has a shape. An accounting package has a shape. Wiring intelligence into the way a business already runs does not, because the work is defined by your systems, not by mine.
So a price list would do one of two things. It would be padded high enough to cover the worst case, which means most clients overpay for the privilege of certainty. Or it would be set at the average, which means the complicated jobs get quietly under-scoped and then go wrong halfway through.
Neither of those is a good deal for you. What I publish instead is the method: the two questions that put you in a shape, the things that move you inside it, and the point at which the number stops being an estimate and becomes fixed. That happens when there is a written scope, whether that comes out of a Discovery or out of a conversation with someone who already knows what they need.
The three things that set your number
Every quote I write comes down to these. Read them and you can place yourself before we ever speak.
Complexity
How many systems have to talk to each other, and how well behaved are they. Two modern tools with proper APIs is a straightforward job. Six tools, one of them a desktop application from 2011 with no API, and a spreadsheet somebody maintains by hand, is a different job entirely. Complexity is the single biggest driver, and it is usually the thing owners underestimate.
Risk
What happens if this breaks at 2am. Automating an internal reporting pack is low risk, because the worst case is a stale report and a mildly annoyed manager. Automating the intake path that touches client money, patient records, or a legal deadline is not, and the extra cost is testing, fallbacks, audit trails, and the review cycles that go with them. You are paying for the work that makes it safe, not just the work that makes it run.
Ambition
How far you actually want to take this. Some owners want one process off their plate and are done. Others want the domain knowledge in their business captured in systems that keep learning, so it stops living in three people's heads. Both are legitimate. They are not the same size of job, and pretending otherwise is how engagements go sideways.
Nothing in that list is your revenue, your headcount, or how much I think you can afford. Price follows the work.
Two questions put you in a shape
Rather than a menu, it is two questions. Answer both and you know roughly where you sit.
/ First question: do you already know what you need built?
Yes
We scope it and go
Plenty of engagements start here, and I am not going to invent a discovery phase to slow you down. Bring me the problem, we spend a call or two getting specific, and you get a fixed-price proposal for a bounded build. No mapping exercise, no toll gate. If it turns out mid-conversation that the thing you asked for is not the thing you need, I will say so then, which costs you nothing.
Not yet
Buy the map before you buy the build
A Systems Snapshot is free and takes about ten minutes: one process named, what it is costing you, and the first move, emailed as a one-pager. A Discovery is $3,000 fixed over one week and produces the real thing: a systems map, your top three opportunities scored by payback, and one costed proposal. It is credited in full against a build that goes ahead inside 60 days, so on any build over $3,000 it is effectively free if you proceed.
That credit is the honest reason I recommend it. A week of mapping usually finds either a cheaper way to do what you asked for, or a more valuable thing to do instead, and either one pays for the week. It is the best-value week most owners will buy from me. It is still your call.
If you are a founder with a new product idea rather than an owner with a messy process, the equivalent first step is a two-week demand test at under $1,000, then a Validation sprint at under $4,000 if the demand is real. Different question, same principle: cheap evidence before expensive building.
/ Second question: is this one job, or an ongoing capability?
One job with a clear finish line is a bounded build, $5,000 to $25,000, fixed price, scope locked before it starts. Product and MVP work runs the same way at a different size, $10,000 to $30,000 for most builds and around $50,000 for genuinely complex multi-sided ones. Details are on the product page.
An ongoing capability is a retainer, and there are two of those. They are not a small tier and a big tier, so the next section is about what actually separates them.
Personal AI OS is a separate thing again: $1,500 for the first seat, $750 each after that, $3,500 for a team workshop. It gets your own AI daily driver running on your machines and your data, and it pairs naturally with Discovery.
The two retainers, and why the step exists
This is the part worth reading properly, because the gap between $4,000 and $9,500 looks like a quantity of time and it is not.
Baseline retainer
from $4,000 / mo
A steady cadence through your backlog, re-prioritised against results each month. You own the backlog. You decide what is next, I work through it, and the commitment I am making is that the list keeps moving.
There is no day count on this tier and no timesheet, on purpose. What you are buying is that things keep getting done at a reliable pace, not a quantity of me. Bigger programmes sit above the floor, and the number follows the scope rather than a tier chart.
Embedded
from $9,500 / mo
A minimum of six days a month, usually one fixed day a week, scaling to around two and a half days for a busy portfolio. Month to month, thirty days' notice either way.
Here the relationship inverts. I am not working a list you maintain; I am holding part of the portfolio with you, inside your decision-making, running build streams and carrying the context between them. That means being reachable on a schedule other people in your organisation can plan around, which is what the six days is for.
So why more than twice the price for something that is not twice the work?
Because it is not the same work. A baseline retainer is build capacity against priorities somebody else owns. An embedded engagement is somebody helping own the priorities, which is a different job with a different failure mode: if I get the sequencing wrong on a baseline retainer, you have wasted a month of throughput. If I get it wrong while embedded, you have pointed your team at the wrong thing.
The other half of the answer is boring and honest. A fixed day a week is capacity held for one organisation, and held capacity cannot be sold twice. That reliability is most of what separates the two numbers.
The six days is a scheduling commitment, not a unit of sale. You can divide the fee by it and you will get a number, but that number will not tell you anything about what you are buying, and it is not how the work is priced or invoiced.
Which one you want, plainly. If you have a list of manual processes and somebody internally who owns it, take the baseline. If the constraint is that nobody senior has the room to run this properly, and you need somebody in the room on a predictable rhythm, take embedded. Most clients start on baseline. Some move up. Moving back down is fine too, and it happens.
What moves your number
Worth being blunt about these, because every one of them is something I would rather you knew before the proposal than after.
/ Up
Legacy systems without APIs
If a system will not talk, something has to make it talk, and that something takes real engineering time.
Data that has never been cleaned
Automating on top of a messy dataset produces confident, well-formatted nonsense. Cleaning it first is usually unglamorous and usually necessary.
Regulated or high-consequence workflows
Health, legal, financial, anything with a statutory obligation attached. The safety work is not optional and it is not free.
More than one decision-maker
Not a criticism, just arithmetic. Every additional approver adds review cycles, and review cycles are calendar time.
A hard external deadline
Compressing a schedule means holding capacity for you, and held capacity costs more than scheduled capacity.
Scope that keeps growing
Handled up front by fixing the definition of done, which is why I insist on writing one.
/ Down
A clean, modern stack
Xero, Shopify, a proper CRM, tools with real APIs. Most of the win is wiring what you already pay for.
Doing the process cleanup first
Sometimes the honest answer is that a checklist fixes it and no software is needed. Discovery will tell you that, and I would rather say it than sell you a build you do not need.
Starting narrow
One process, done properly, usually pays for the next one. Zero to one, then one to ten.
An owner who is decisive
The fastest engagements I have run were not the simplest ones. They were the ones where somebody could make a call in a day.
Knowing what you want already
If you have done the thinking, or had it done for you, you can skip straight to a scoped build and keep the cost of the mapping in your pocket.
How the number actually gets set
Three steps, and you can stop after any of them.
- 1
A short call
Thirty minutes. We work out whether this is even the right problem to be spending money on. Free, and genuinely no obligation.
- 2
A written scope
A real number, not a band, with what is being built and a written definition of done. If you arrived knowing what you need, this comes straight out of the call and costs you nothing. If you did not, this is what a $3,000 Discovery produces, along with the map and the scoring that got there.
- 3
The work
Fixed price for a bounded build, or one of the two retainers for a programme. Either way you knew the number before it started.
The thing I will not do is quote a number at step one for work nobody has specified yet. That is not a rule about Discovery, it is a rule about scope: a price is only honest once somebody has written down what it buys. Some clients get there in a call, some need the week. Both are fine.
What is not included, and what you can hold me to
/ Not included
Software subscriptions run $30 to $250 a month for most stacks, billed to you directly by the vendor so you own the accounts and the data. I do not resell tooling or take margin on it. Model and API usage is a separate running cost, small for most businesses but real. Work outside my lane, brand positioning, visual identity, pure design, goes to someone who is genuinely good at it rather than getting a passable job at your expense. Every figure on this page excludes GST.
/ What you can hold me to
A fixed price before any build starts, which only moves if you ask for something new, in writing. A written definition of done with one metric that matters, hours back or revenue earned, not activity. No lock-in: both retainers are month to month, thirty days' notice either way. You keep every artefact, the map, the scores, the proposal, whether we build together or not. And an honest no: if AI is not the answer to your problem, I will say so at the point I work it out, whether that is on the first call or in the middle of a Discovery, rather than selling you the build anyway.
/ Common questions
How much does Garage 30 charge?
Garage 30 does not publish a fixed price list, because the work is defined by the client's systems rather than by a standard package. A fixed-price AI Systems Discovery is $3,000 NZD. A bounded build runs $5,000 to $25,000 NZD. Ongoing work runs as either a baseline retainer from $4,000 NZD a month or an embedded day-rate engagement from $9,500 NZD a month. The exact figure is fixed in writing once the scope is written down, which for some clients comes out of a Discovery and for others comes straight out of a conversation.
What is the difference between a baseline retainer and an embedded engagement?
They are different jobs, not different amounts of the same job. A baseline retainer, from $4,000 NZD a month, is a steady cadence through a backlog the client owns and prioritises: they decide what is next, Garage 30 works through it, and the commitment is that the list keeps moving. An embedded engagement, from $9,500 NZD a month, puts Casey inside the organisation's decision-making, holding part of the portfolio and running build streams alongside it, with a minimum of six days a month so the availability is something colleagues can plan around. Baseline is build capacity against someone else's priorities. Embedded is helping own the priorities.
Why is an embedded engagement more than twice the baseline retainer?
Because it is not twice as much of the same work. A baseline retainer delivers against a backlog the client maintains, so a sequencing mistake costs a month of throughput. An embedded engagement helps set the sequence itself, which carries a different kind of risk and a different kind of responsibility. The second half of the answer is capacity: a fixed day a week is time held for one organisation and it cannot be sold twice, and that reliability is most of what separates the two figures.
Does Garage 30 charge by the hour?
No. Work is priced as a fixed fee for a bounded scope, or as a monthly retainer for an ongoing programme. The baseline retainer has no day count and no timesheet at all. The embedded engagement states a minimum of six days a month, but that is a scheduling commitment so the client can plan around the availability, not a unit being sold, and the work is neither priced nor invoiced by the day. Hourly billing rewards the wrong thing: a consultant who takes three times as long earns three times as much.
Do I have to do a paid Discovery before Garage 30 will work with me?
No. Discovery is recommended when the problem is not yet well understood, not a prerequisite for working together. Clients who already know what they need go straight to a scoped, fixed-price build after a call or two, and that costs nothing extra. Discovery exists for the other case, where a week of mapping usually finds either a cheaper way to do the thing that was asked for or a more valuable thing to do instead. At $3,000 NZD, credited in full against a build that goes ahead inside 60 days, it is effectively free on any build larger than that if the client proceeds.
Why will Garage 30 not give me a price on the first call?
Because a price is only honest once somebody has written down what it buys. That is a rule about scope rather than about discovery: some clients specify the work well enough in a single call and get a fixed-price proposal straight away, and others need a week of mapping first. What Garage 30 will not do is quote a number for work nobody has specified, because a price set before anyone has seen which systems are involved and how clean the data is describes a standard package rather than the client's business.
What does an AI Systems Discovery cost and what do I get?
$3,000 NZD, fixed, over one week. It produces a systems map of how the business actually runs, the top three automation opportunities scored by payback against effort, and one costed build proposal with scope and timeline. The client keeps every artefact regardless of what they decide next, and if a build goes ahead inside 60 days the full $3,000 is credited against it.
What makes an AI automation project more expensive?
Four things, in rough order of impact. Legacy systems without APIs, because something has to be built to make them talk. Data that has never been cleaned, because automation on top of a messy dataset produces confident nonsense. Regulated or high-consequence workflows, where testing, fallbacks and audit trails are not optional. And a compressed deadline, because holding capacity costs more than scheduling it.
Should I take a fixed-price project or a retainer?
A fixed-price build, $5,000 to $25,000 NZD, suits one clearly bounded problem with an obvious definition of done. A baseline retainer, from $4,000 NZD a month, suits a business with a backlog of automation work that wants it worked through at a steady pace without re-scoping every few weeks. An embedded engagement, from $9,500 NZD a month, suits an organisation running real build streams that needs someone senior inside the decision-making on a predictable rhythm. Most clients start with a bounded build or a baseline retainer and move up only when the work justifies it.
What is the cheapest way to start working with Garage 30?
The Systems Snapshot, which is free. It is a conversational audit of about ten minutes that names one process worth automating, what it is costing, and the first move, emailed as a one-pager. From there, a client who already knows what they want can go straight to a scoped fixed-price build, and a client who does not can buy a Discovery at $3,000 NZD, credited in full against a build that goes ahead inside 60 days.
What does it cost to test a product idea before building it?
A two-week demand test runs under $1,000 NZD and tells you whether strangers will act, not just whether they like the idea. If the demand is real, a Validation sprint at under $4,000 NZD produces a prototype, a landing page, or structured user conversations, and ends in a recommendation to build, pivot, or park. Most MVP builds that follow land between $10,000 and $30,000 NZD, with genuinely complex multi-sided products closer to $50,000.
Are software costs included in the price?
No. Software subscriptions run $30 to $250 NZD a month for most small business stacks, and they are billed directly by the vendor so the client owns the accounts and the data. Garage 30 does not resell tooling or take margin on it. Model and API usage is a separate running cost, small for most businesses but real. All quoted figures exclude GST.
What happens if the project ends up costing more than quoted?
It should not, and that is the point of fixing scope before the build starts. The quote follows a written definition of done, so the invoice at the end matches the number on the proposal. The price only moves if the client asks for something outside that definition, and when that happens it gets re-quoted in writing before any of it is built. Retainers are month to month with thirty days' notice either way, so an ongoing engagement can be stopped rather than absorbed.
The takeaway
The number is not a secret. It is just downstream of the scope.
Ask any consultant what will be live in your business at the end of the first month, and what they need to know before they can tell you. If they can answer both specifically, the price will make sense. If they cannot, the price is fiction no matter how precise it looks.
If you want to know roughly where you would land, a short call is the fastest way to find out.
Not ready for a call? Start with the free Systems Snapshot, ten minutes.