For thirty years the hard part of software was the building, so power routed through engineers and the idea lost fidelity at every handoff. We come from a different tradition. Handed a blank run of pages, a budget that never covered it, and too few people for the job, we were told to go make something people wanted and something that paid for itself. We did. We moved millions to try something new, to get better at what
“The cost of execution has collapsed, but the returns to taste and point of view have gone way up.”
Olivia Moore · a16z · February 2026
they already did, to live a little happier, and yes, to buy what our advertisers were selling. We won awards, and we built brands that meant something. Making remarkable things for other people is already in our DNA. What is new is that the tools finally let us build them ourselves, from the first idea to production.
The hard part of software used to be building it. AI flipped that, so the scarce resource now is vision, and we're a creative-led studio that ships working AI systems into the mid-market, not another slide deck.
01 The problem
The world will spend $2.59 trillion on AI this year and almost no one is getting value from it. The bottleneck was never the building. It is knowing what is worth building.
02 Worse for the mid-market
Companies from $10M to $1B in revenue want it, can't staff it, and are too small for the enterprise machine to serve. The adoption gap is already compounding into a competitive one.
03 What we provide
We don't advise, we build. Working AI systems shipped into your stack and running, plus Maasv, our cognition layer, deployed alongside them so they keep getting smarter after we leave.
04 How we work
A repeatable, in-and-out program: observe, present, build, hand off. Three founders and a small paid build crew. No embedding, no bodies for hire, fixed scope and fixed fee.
05 How the money works
Two revenue streams: the Build, a fixed-fee project from $100K to $1M, and Maasv, licensed for as long as the system runs. The conservative base case ramps to $7.2M by year three.
06 Why us
Creatives who can finally build, a market our network already owns, and our own cognition layer. The competition staffed up on engineers and went shopping for vision; we started with the vision and the tools caught up.
The Ledger
Everyone is spending like it works · Nº 01
Almost no one is getting it to work.
According to Gartner, the world will spend $2.59 trillion on AI in 2026, up 47% in a single year. And it isn't only the data-center build-out: Menlo Ventures reports that enterprises tripled their spending on generative AI alone last year, to $37 billion, what it calls the fastest-scaling category in the history of software. The return on all of it is close to nothing.
39%
see any impact
Only 39% of companies see any measurable impact on the bottom line, according to McKinsey's 2025 State of AI – and most of those under 5% of EBIT.
42%
have abandoned most of their AI initiatives, up from 17% a year earlier, says S&P Global's survey of 1,006 companies.
89%
saw no productivity impact at all in the most rigorous study to date – roughly 6,000 firms surveyed by four central banks, per the NBER.
The Diagnosis
It isn't the technology · Nº 02
It's the absence of vision.
The tools work. What's missing is judgment about what to point them at.
i
The wrong project.
Picked because it was easy to name, not because it mattered.
ii
No vision.
Nothing connecting the work to the actual business.
iii
Caution dressed as strategy.
Rebuilding the thing they already had and stapling a “now with AI” badge on it.
The post-mortems say it in drier language. RAND's 2024 study of failed AI projects traces them to “misaligned purpose” and “chasing technology over business outcomes,” while Gartner finds the ones that die are killed for “unclear business value,” which is the tell. The bottleneck was never the building. It is knowing what is worth building.
The Market · Nº 03
The mid-market is the worst-served slice of all.
According to the National Center for the Middle Market at Ohio State, companies between $10 million and $1 billion in revenue number nearly 200,000 in the US, employ about 48 million people, and produce roughly a third of private-sector GDP. If they were a country, one of the largest economies on earth. Big enough to need real systems. Too small for the enterprise machine to serve well.
25%
have actually integrated generative AI into how they operate, RSM found in its 2025 middle-market survey – even though 91% are already using it.
70%
say they need outside help to get value from it, and in the same survey 39% blamed a lack of in-house expertise above all else.
74%
plan to increase AI spend over the next two years, RSM's 2026 follow-up found, increasingly through outsourced and co-sourced help.
They want it. They are trying. They know they can't do it alone. And the people they'd hire to help, if they can even find them, mostly sell them a plan, not a working thing.
The team they can't hire is exactly the team we are.
The Stakes
The gap is compounding · Nº 04
+5.5%
productivity for the S&P 500 since ChatGPT launched, against −12.3% for small caps, by Wells Fargo's reckoning in an analysis reported by CNBC.
2×
the AI adoption rate at large firms versus small ones – the OECD counts 40% of firms over 250 employees against 11.9% of those under 50, a gap fresh US Census data confirms.
5×
the revenue gains, and 3× the cost reductions, that BCG found separate the firms which have cracked it from everyone else.
For a mid-market company, getting AI wrong is no longer a wasted budget line. It's falling behind a competitor who got it right.
The Incumbents
Built to bill, not to ship · Nº 05
The help that exists answers demand with advisory.
Accenture booked $5.9 billion in generative-AI work in FY2025, its own SEC filings show, nearly double the year before. And by industry estimates, a typical Big-4 engagement prices the mid-market out before the work even starts. What it buys, too often, is a roadmap and a slide deck.
Strategy$0.5–1M
Pilot$1–2M
Full build$3–10M
One Big-4 AI engagement, by tier – industry estimates
“Buying from a partner who ships beats building it in-house roughly three to one.”
MIT NANDA, 2025
The market is starving for people who deliver a working system instead of a recommendation.
The Position · Nº 06
We don't advise. We build it.
A small, creative-led studio that ships working AI systems into the businesses that need them. Not assessments. Not roadmaps. The actual thing, running in your stack, doing the job.
Vision-first
We start by understanding the real problem, the way an editor or a designer would, not by selling a methodology.
Build-native
AI lets the people with the vision build the thing directly. No fidelity lost in translation.
Vertically deep
We go narrow on purpose. We learn one industry cold and become the obvious choice in it.
The strategy is free. It happens in the conversation, the way it should. The build is what we charge for, because the build is what actually changes anything.
The Edge
Two things almost no one has at once · Nº 07
01
Creatives who can now build.
The whole industry staffed up on engineers and went shopping for vision. We started with the vision, and the tools finally caught up. Our network is editors, designers, and domain operators, the people who can see what's worth building. No engineering shop can recruit that pool, because it was never theirs.
02
Our own cognition layer.
We run the studio on Maasv. It remembers every client, every system, every decision we've ever made, which makes a small team operate like a much larger one. And it isn't just internal. We deploy it into client work as a product, so the systems we build keep getting smarter after we leave.
Why Now
The build cost collapsed · Nº 08
AI didn't make expert engineers type faster. In the most rigorous study to date, METR found it made them 19% slower on code they already knew. What it did was collapse the cost of getting from an idea to a working system, which is exactly why the scarce resource moved upstream to judgment and experience.
Base44
An eight-person, six-month-old, profitable company. Sold to Wix for $80 million in cash.
Lovable
Zero to $200 million in annual revenue in twelve months, with around 45 people.
“The constraint was never the vision. The vision was always ours.”
We always knew what to build. What we never had was the means: the engineers, the budget, the headcount, the infrastructure to make it real. That constraint is gone.
The Competition
“Build, don't advise” is table stakes · Nº 09
We are not the only ones who build.
The serious players already lead with shipped systems, not slideware. We studied them closely, and the honest finding is that “we build, not advise” has become the price of entry, said by everyone credible. Their sites even run the same playbook: a three-word promise, a wall of logos, one testimonial, case studies in the client's own numbers. We know the moves. The question every buyer is really asking is who has the goods.
A fast-growing tier renting out a part-time head of AI by the month.
“‘We build, not advise’ is no longer an edge. It's the cost of entry.”
So our edge cannot be that we build. It is who does the building – creatives with vision, not only engineers – how narrow we go, and the cognition layer we build on. The boutique tier sits at $25K–$100K+ per project; we price at the top of it and up, to the outcome rather than the hour. Which leaves one player big enough to change the math.
The Field · Nº 10
$1.5B
put behind a new AI-native services firm aimed at the mid-market by Anthropic, Blackstone, Hellman & Friedman, and Goldman Sachs, Bloomberg reported this spring. They immediately acquired Fractional AI, one of the best build shops in the market, to run it.
The category is real enough that institutional money is rolling it up at the top. That is the best news we could get. They are building a Big-4-style deployment factory. We are the opposite shape.
Them
Us
Engineering scale
Vision and domain experience
One model, one playbook
Model-agnostic, problem-first
Horizontal, everyone
Vertically deep, the obvious choice in one lane
A machine you hire
A team you trust
They proved the market. They can't be us. The window to plant the flag on the creative, vertical, independent version of this is open now, and it won't stay open long.
The Method
Come in, build, move on · Nº 11
01
Observe
We come in, learn the business, and find the system actually worth building.
02
Present
We leave, think, and come back with the plan and a fixed price. No surprises.
03
Build
We build it in production, in your stack. No embedding, no bodies for hire.
04
Hand off
We hand over a working system and move on. It keeps running on Maasv.
Two ways we earn, and only two: the build, priced as a project, and Maasv, the cognition layer it runs on, licensed for as long as the system is working. We start where we already have an edge: trades, through a live relationship and an existing pilot, and healthcare and member engagement, adjacent to work we've already shipped. We are not a body shop; we are building a name.
The Numbers
A business, not a freelance gig · Nº 12
The build pays the bills. Maasv, the product it runs on, is where the value compounds. Below, both lines and their total.
Revenue ramp – the build, Maasv licensing, and the total
Maasv estimated at ~$10K / mo per client deployment – a placeholder until real pricing is set
How we price
We price the outcome, not the hour. Fixed scope, fixed fee, nothing under $100K. A day-rate shop passes its AI speed back to the client as fewer billable hours; we keep it. So as the build gets faster, the speed becomes our margin, not the client's discount.
The Build
$100K–$1M
A fixed-fee project, priced to the outcome and the value at stake. Observe, present, build, hand off – weeks to a few months, never a two-week sprint.
Maasv
~$10K / mo
Per client deployment, licensed for as long as the system is working. Recurring, high-margin, and it compounds. A placeholder until we set real pricing.
The build engine – base case, deliberately conservative
Year 1
Year 2
Year 3
Founders + employees
3 + 1
3 + 3
3 + 5
Build revenue
$1.2M
$2.4M
$4.8M
Salaries + overhead
$370K
$990K
$1.7M
To founders + product
$830K
$1.4M
$3.1M
Revenue per head climbs from $300K to $600K as the studio matures and the builds get bigger, above the $160K–$220K a generalist agency clears (Iota Finance, HRBench), because we bill senior, value-priced work rather than leverage. The team stays lean: three founders, who aren't a salaried cost, plus a small build crew (loaded at roughly $250K each) that grows with the work. After those salaries and the light overhead – inference, software, formation, a little marketing, no office – roughly two-thirds of revenue is left for the three founders and for reinvesting in the product. The business funds itself and Maasv from day one.
What it means for the people in the room.
This pays from day one. Year 1 is a real income, not a someday-payoff, and the engine fills fast. A business that pays its founders a living while the equity builds underneath them.
The product upside – where the real money is.
Maasv rides on top of the client work at SaaS economics: roughly 76% gross margin, recurring. At the ~$1 million of Maasv ARR the chart reaches by year two, the standard SaaS multiple (~6.7×) puts $6 to $7 million of enterprise value on the product line alone, and it only climbs from there; the AI-premium band roughly doubles it. For the ceiling, not the plan: Base44 exited at $80M; Lovable reached a $6.6 billion valuation inside a year.
The SaaS economics – roughly 76% gross margin and a ~6.7× ARR multiple – come from CloudZero and the SaaS Capital Index. Services pricing is directional; the margin and revenue-per-head anchors carry the model, and the headline valuations are top-decile outcomes used to size the upside, not the base.
It's much more fun to have helped build than to have watched get built.
Everything this needs is already here. We can see the problem clearly and, for the first time, we have the tools to build the answer ourselves. We have a year of shipped, running systems behind us, and a library of components that makes each new build faster than the last. We have a cognition layer that lets a small team operate like a much bigger one. And the market is right there, asking out loud for help it can't find.
So the obvious version isn't something we grow into over two years. It's the three of us, now. We already trust each other, and between us we already have the range it takes. The window is open, the rest of the field is still writing decks about it, and the only people who can start this are us. The sharp, independent studio known cold in its verticals, with a product compounding underneath it, is just what this becomes once we begin.
The only thing left is us.
ABC
We build.
A creative-led AI build studio. We ship production systems into mid-market companies.