Services as Software: The Arbitrage Almost Nobody Is Talking About

ANNOUNCEMENT: We Built You a Local AI Agency. It's $27.

In 2020, Alex Hormozi emailed me to help scale his ads. I've built two seven-figure agencies and sold one of them.

I've now packaged all of that into an agency in a box. We build it for you in 24 hours. Named, branded, website live, offer priced, demo script written word for word, proposal and contract pre-filled in your name.

You sell local service businesses an AI receptionist at $297 a month plus $497 setup. You never build anything. The platform delivers.

$27 one time. 180 day money back guarantee.

Build my AI agency

Why the gap between what services cost to deliver and what the market still pays for them is the biggest opportunity in business right now

October 11, 2026
Read Time: 9 minutes

There's a number that explains almost everything about where the opportunity sits right now.

For every one dollar a business spends on software, they spend six dollars on services.

Six to one. That ratio has held for decades and it isn't moving. Businesses buy services relentlessly, consistently, and without much resistance, because services solve problems that software alone never could.

Foundation Capital put a figure on what happens when AI collides with that ratio. They called it services-as-software, and they valued the opportunity at $4.6 trillion.

A16Z, Sequoia, and Y Combinator are all backing companies built on this thesis right now. Harvey AI and Lawyaa in legal. Dozens more across accounting, healthcare, and operations. These are some of the fastest growing startups in the world, and none of them are selling software.

They're selling outcomes, delivered by AI, priced like human labour.

That's the arbitrage. And it's available to you at a fraction of the scale with none of the venture capital.

The Four Quadrants, and Why One of Them Is a Gold Mine

Here's the clearest way I've found to think about this.

You can sort almost any business into one of four positions based on two questions: is the work delivered by humans or AI, and is the client buying a tool or an outcome?

Classic software. Sold as software, delivered by software. You buy it, your team uses it, you're priced per seat. Predictable, competitive, margin-compressed.

AI-delivered tooling. Still sold as a tool, but AI is doing the work underneath. Co-pilots, agents, assistants. Better margins, but you're still selling a thing rather than a result.

Traditional services. Sold as an outcome, delivered by humans. Consultants, agencies, lawyers, accountants. The client buys more leads, a finalised divorce, a clean set of accounts. High prices, but your margin is permanently capped by salary costs.

And then the fourth quadrant. Sold as an outcome. Delivered by AI.

That's where the arbitrage lives. You charge what the market pays for outcomes, which is high, and you deliver at the cost of AI, which is almost nothing.

Nobody owns this quadrant yet. It barely existed eighteen months ago.

What the Numbers Actually Look Like

Let me use real figures from my own history rather than hypotheticals.

My last agency, in the real estate space, sold between 2020 and 2021. We charged $1,500 to $3,000 a month, typically closer to $3,000. Cost to acquire a customer was around $1,500. Close rate around 30%. Delivery took ten hours or more per client per month across media buying, landing pages, and a call centre.

Gross margin landed somewhere between 50 and 60%. And there were months where we actually lost money on a new client in month one, between acquisition cost and delivery cost. The money only arrived in months two, three, and four if they stayed. Typical agency churn in local sectors runs three to four months, so you're looking at nine to twelve thousand pounds of lifetime value if you're lucky.

Now compare that to my current local services business.

We charge $300 to $800 a month. Margin is 99%. There is one customer success manager doing onboarding calls. Everything else, the build, the fulfilment, the delivery, is handled by an AI system my technical co-founder built end to end.

Cost to acquire a customer sits around $300 to $400. Add a setup fee of $250 to $300 and acquisition is effectively free. Close rates are substantially higher because at that price point it's much closer to an impulse decision. Delivery is under an hour a month per client, mostly the onboarding call and occasional check-ins handled by text.

Lower price. Dramatically higher margin. Lower acquisition cost. Higher close rate. Better retention.

That's not a small improvement. That's a different business.

The Two Plays, and They Both Work

Once delivery costs collapse, you have a genuine strategic choice, and both options are legitimate.

Play one: hold the price.

Keep charging $3,000 a month. Deliver entirely with AI. Your margin goes from 50% to 95% overnight.

The advantage is obvious. Same revenue per client, vastly more profit. The disadvantage is that you inherit all the existing problems of the traditional model. The same churn rates, the same acquisition costs, the same ceiling on market size because most businesses can't afford three grand a month.

You can probably run this arbitrage for the next twenty-four months before the market catches up.

Play two: compress the price.

Drop from $3,000 to $800, or $300. Your margin stays above 95% because delivery costs almost nothing.

What you unlock is enormous. A segment of the market that was never addressable before. Not just in the US and UK, but internationally, where $3,000 a month was never realistic but $300 absolutely is.

You get much higher retention, because at that price the service is sticky and the decision to cancel never feels urgent. You get a lower cost to acquire. You get a higher close rate.

The trade-off is that you make less per client, which means the entire game becomes acquisition velocity. How fast can you put new customers into the system. That's the constraint you're now optimising against.

Which Play You Choose Depends on the Deliverable

This is the part people get wrong, and it matters.

If the deliverable is visible, compress the price.

Content, ads, social media management, bookkeeping. Work the client can see and count. Thirty pieces of content a month. Fourteen ad variations. A set of reconciled accounts.

Visible deliverables invite line-by-line scrutiny. A moderately savvy client will eventually look at thirty social posts and think, I could probably do this myself now. I've watched people in this space go from charging a couple of thousand a month for social media management down to around four hundred, because their clients were technologically literate enough to work out what was actually involved.

How fast that happens depends entirely on the sophistication of the sector. Go after home service businesses and you can comfortably charge a grand a month for content that costs you ten dollars to produce. Go after tech-forward clients and the compression happens much faster.

If the buyer is paying for trust, hold the price.

Legal, advisory, consulting. Outcome-based work where the client fundamentally cannot see inside the process. They don't have the expertise to evaluate how the result was produced, and they're not buying the process anyway. They're buying certainty.

AI consulting sits squarely here. There's a genuine body of industrialised knowledge in how you run an assessment, structure the diagnostic, build the opportunity matrix, and present the findings. The client isn't evaluating your method. They're evaluating whether they trust you to get it right.

That's defensible pricing, and it holds far longer.

The Fourth Force: Guilt

Three of the forces lining up behind this are straightforward.

Market, because services spending dwarfs software spending and always has. Timing, because delivery costs have collapsed and the market hasn't repriced. Purchasing power, because you're selling something businesses are already buying, so no budget needs to be created.

The fourth one is newer and more interesting.

Right now, a small minority of companies have genuinely adopted AI. The vast majority have not. And every month that passes, the owners of those businesses feel it more acutely. They read about competitors moving faster. They hear about margin improvements they're not seeing. They know they should have started already.

That accumulated guilt is a budget line waiting to be spent.

Which is exactly why you sell the outcome and not the homework. You're not selling them an AI system. You're not asking them to learn anything or change how they work. You're saying: thirty leads in ninety days. Every call answered. No more missed revenue.

They don't want to understand AI. They want someone to make the guilt go away.

What Actually Kills This

Being honest about the risks matters.

Clients DIY it. The theoretical risk is that savvier clients realise they can do it themselves. In practice I've run into this far less than you'd expect. Plenty of clients have told me they're already using AI and could handle it internally. Then we compare their output to ours and the gap is night and day. Knowing the tools exist and being good at deploying them commercially are entirely different skills.

Race to the bottom. This always happens, and it'll happen here too, particularly from lower-cost regions. You saw exactly this on Fiverr, where a hundred dollar an hour service in the US would be advertised at five dollars by someone overseas claiming identical output.

The defence is the same as it's always been. Case studies, proof, testimonials, and relationship. Those compound, and they can't be undercut. Which is precisely why starting now matters, because the moat is built over time.

The 5% human bottleneck. You still need skilled people. The best services businesses from here will be small, sharp, and genuinely AI-native. Not large teams doing manual work, but a handful of people who are exceptional at deploying these tools.

Quality drift. If you let AI run unsupervised and stop checking output, churn rises because you stop delivering. Worth noting though that in the local business segment, delivery is simple enough that minor slippage often isn't even noticed by the client. We notice it before they do.

How to Find Your Own Play in Four Steps

If you're already running a services business, this is the exercise.

List every deliverable you currently sell. Every single one, broken out properly.

Scope how much of each one AI could realistically handle today. Not in two years. Today.

Choose the most valuable deliverable and decide which play fits it. Compress if it's visible. Hold the price if the client is buying trust.

Rebuild delivery first, then reprice second. Never the other way round.

And if you're starting from nothing, the instruction is even simpler. Go find a model that already works, from any era. Social media marketing circa 2010. Lead generation for real estate agents. Local SEO. Then rebuild it so that AI handles the ads, the funnels, the creative, the follow-up, the reporting.

The model doesn't need to be new. Your delivery does.

A funnel that works for one real estate agent is copy and paste for the next hundred. The difference is that the same business that ran at 60% gross margin in 2019 now runs at 95%.

The Window

I'd estimate we have until roughly 2028 in the US, UK, and Canada before the market broadly understands what delivery actually costs and prices adjust accordingly.

Internationally, and across other languages and regions, the window stays open considerably longer.

But the honest framing is this. Market pricing is still anchored to what services used to cost. Delivery pricing has already fallen off a cliff. The gap between those two numbers is the entire opportunity, and it is the widest it will ever be right now.

Every month you wait, it narrows slightly.

I broke all of this down in full detail in a video this week, including the full four-quadrant framework, the real numbers from both agencies, and exactly how to position the model. It's doing well, and it covers more ground than I could fit here.

And if you'd rather skip the build entirely and just have a local AI agency handed to you, named, branded, website live, offer priced, demo script written, that's the $27 option at the top.

See you next week,

– Andrew