The next great software company won't sell software

For twenty years, we sold tools.

We called it SaaS. Software as a Service. We packaged code into a login screen, charged a monthly fee, and watched as every category of work got its own dashboard. Salesforce for sales. HubSpot for marketing. QuickBooks for accounting. Workday for HR. The dashboards multiplied. The work, mostly, did not go away.

A company today spends a few hundred euros a year on accounting software. And tens of thousands on the accountant who actually closes the books.

That gap is the entire opportunity.

For every dollar spent on software, six are spent on services. Six dollars of human work, sitting next to one dollar of code, doing the things the software couldn't quite finish. Reading the email. Reconciling the spreadsheet. Drafting the contract. Filing the claim. Triaging the ticket. The dashboard told you what to do. A person still had to do it.

That era is ending. And what's coming next is not better dashboards.

What's coming next is the work itself, sold as a product.

Service as a Software

Sequoia put a name on it last month. They called it the autopilot. The shift from selling the tool to selling the outcome. From "we'll help your accountant work faster" to "we'll close your books." From "we'll make your lawyer more productive" to "we'll draft the NDA." From copilot to autopilot.

The new SaaS isn't Software as a Service. It's Service as a Software.

You don't buy access to a CRM. You buy qualified leads in your inbox.

You don't buy a contract drafting tool. You buy the contract.

You don't buy a support helpdesk. You buy support that already happened, while you slept, in three languages, with a satisfaction score attached.

This is not a feature update. This is a different business.

The old SaaS company wrote code, charged a subscription, and counted seats. The new one writes agents, charges for outcomes, and counts work units. The old moat was switching costs. The new moat is the proprietary data on what good work looks like in your category, accumulated one job at a time, compounding faster than any competitor can catch up.

For thirty years, software ate the world. Now software is going to do the world's work.

Why now, and why not before

People have been promising "AI that does the work" for a decade. Most of it was theatre. RPA bots that broke when a button moved. Chatbots that fell over when a customer typed something the script didn't predict. Agencies hiding humans behind a software-shaped login screen and calling it AI.

What changed is not marketing. What changed is that the models can finally do the intelligence work.

Sequoia's framing here is the cleanest I've seen. Two kinds of work. Intelligence and judgement.

Intelligence is rule-following. Complex rules, sometimes thousands of them, but rules. Translating a clinical note into an ICD-10 code. Drafting an NDA from a template. Reconciling a bank statement. Matching a CV to a job spec. Reading an insurance policy and pulling the deductible. The rules are hard. The work is repetitive. There is a right answer.

Judgement is taste. It's the gut call that comes from doing something a thousand times. It's choosing which feature ships next. It's deciding whether to take the deal. It's reading the room.

For decades, we built software that handled neither. It just held the data while humans did both.

Today, the models do the intelligence work. Not perfectly. Not in every domain. But enough that, for the first time, you can sell the outcome of intelligence work and actually deliver it. The judgement layer still belongs to humans. But the intelligence layer, the six-dollar layer, is up for grabs.

The companies that figure out how to sell that layer as a product are going to be very, very large.

What this means for the rest of us

If you're building an AI tool right now, you already know the question that keeps you up at night. What happens when the next version of Claude makes my product a feature?

If you sell the tool, you're in a race against the model. Every six months, the model gets better and your wrapper gets thinner. You're a feature waiting to be absorbed.

If you sell the work, every improvement in the model makes your service faster, cheaper, and more defensible. The model getting better is your roadmap, not your obituary.

This is the cleanest strategic test in AI right now. Are you selling the tool or the work?

If you're selling the tool, you'd better have a very good reason. Distribution. A regulated workflow. A wedge into the customer that nobody else can match. Otherwise, the math is brutal.

If you're selling the work, the question changes. Now you have to decide which work, for whom, and how you replace what's already being outsourced.

Why outsourced work is the wedge

The Sequoia piece makes one more point that I think is the most important of all. The autopilot opportunity starts where outsourcing already exists.

If a company already pays an external accountant to close the books, three things are true. The work can be done externally. The budget is already a line item. The buyer is already paying for an outcome, not a tool.

Replacing an outsourced contract with an AI-native provider is a vendor swap. Replacing a full-time employee is a reorg. One of those happens in a week. The other takes a year and a board fight.

So the playbook is not "go build an AI accountant for the Fortune 500 finance team." The playbook is "find the small business that's already paying a bookkeeper four hundred euros a month, and offer to do it better, faster, and cheaper, with a humans-in-the-loop layer for the messy parts."

The wedge is the outsourced job. The long game is the insourced one.

How we're playing this at LayerX

We've been building LayerX as an AI studio for the last year. We ship agents. We do AI consulting for companies that want to figure out where the leverage is in their own business.

For a long time we described what we do as "AI consulting" and "AI products." Two lines of business sitting next to each other.

I don't think that framing holds anymore.

Here's what I actually think we are. We are an autopilot company in disguise.

When a Portuguese mid-market staffing agency comes to us, they don't want a WhatsApp tool. They want their inbound candidate flow handled. They want the screening, the qualification, the scheduling, all of it, done. Ardaven is the delivery mechanism. The product they're buying is "your candidate pipeline runs."

When a hospitality group hires us, they don't want a chatbot. They want their guest communication, in five languages, twenty-four hours a day, with a measurable conversion rate. The bot is the surface. The product is "your guests get answers and you get bookings."

When a B2B distributor calls us, they don't want a "conversational catalogue." They want orders coming in over WhatsApp from customers who used to phone the office at five PM and never get through. The product is "your trade counter is open all night."

We are not selling tools. We are selling work that gets done.

That's the bet. That's why we're moving fast on consulting as a wedge into specific verticals. Consulting is how we learn what good work looks like in staffing, in hospitality, in distribution, one client at a time. The product is what we ship after we know.

The proprietary data we accumulate, the playbooks, the prompts, the integrations, the edge cases, becomes the moat. Not the code. Anyone can write the code. The code gets cheaper every month. What doesn't get cheaper is knowing exactly how a Portuguese staffing agency talks to a Brazilian candidate at eleven PM on a Tuesday, and what makes that candidate actually show up to the interview.

That's the asset. That's what compounds.

What I tell founders who ask me what to build

Pick a vertical you understand or can learn fast.

Find a job that is already being outsourced in that vertical. Not the strategic, judgement-heavy work. The repetitive, intelligence-heavy work that companies hand off because they don't want to do it themselves.

Sell the outcome of that work. Not a dashboard. Not a tool. The outcome.

Charge for the work, not for seats.

Use AI to deliver it. Use humans for the parts AI can't do yet. Watch which parts those are. That's your roadmap.

Compound your data. Every job you do teaches the system. Every job your competitors don't do is a job you do better next time.

That's it. That's the whole playbook. The companies that execute it well are going to build the next generation of category-defining businesses. Not because they have better models, but because they have better answers to the question "did the work get done?"

A closing thought

I have been around long enough to remember when SaaS was the heretical idea. People said you couldn't sell software as a subscription. People said companies wouldn't put their data in the cloud. People said the desktop was forever.

Now we're at the other side of that curve. The dashboards have multiplied. The work, mostly, has not gone away.

What's coming next is not another dashboard.

What's coming next is the work, done.

If you're a founder building in this space, I'd love to compare notes. If you're an operator at a Portuguese mid-market company looking at a six-figure outsourcing line and wondering what AI does to it, even more so. We're shipping autopilots across verticals right now, and we're learning fast.

The next great software company won't sell software.

It will sell the answer.