Sequoia's AI Ascent 2026 put a phrase on stage that service-company operators should not dismiss as Silicon Valley vocabulary:
"Services is the new software."
The easy reading is that investors have found a new category to fund. That is true, but it is not the important part for an operator running a service business.
The important part is that the old boundary between software economics and service delivery is starting to move.
For the last twenty years, software had the beautiful business model. Build once. Sell many times. Improve the product centrally. Let customers use the same underlying system at scale.
Services had the harder model. More customers meant more people, more handoffs, more supervision, more tribal knowledge, more quality variation, and more management load. Growth usually added complexity faster than it added leverage.
AI is changing that equation. Not because every service company becomes a software company in what it sells, but because every serious service company will need to become software-like in how it operates.
That is the real shift.
The signal under the phrase
Sequoia's published piece on the category argues that services are being rebuilt around AI in a way that makes the economics look more like software. The service still sells an outcome. The customer still wants the work done. But the delivery system behind the work can start to compound.
Y Combinator is pointing in the same direction. In its 2026 Request for Startups, YC describes a future where agencies use software internally to deliver finished outcomes, with "software companies with margin and scale" as the destination.
That sentence matters because it reverses the usual SaaS posture.
The old software company sold tools to the service company.
The new AI-native service company uses tools as internal infrastructure, then sells the finished result.
That distinction sounds subtle until you run the numbers inside an operation. If the customer is paying for the outcome, and the company can deliver more of that outcome through systems instead of manual coordination, the margin profile changes. The customer does not care whether the work was completed by a person, a workflow, an AI assistant, or a combination. The customer cares whether the service was correct, fast, trustworthy, and easy to buy.
That is why this is not just a venture-capital story. It is a service-company operating story.
The OffDeal pattern
John Hwang's OffDeal analysis gives the cleanest case study from the latest batch of material.
OffDeal started as SaaS for search funds. Then it moved in the opposite direction most founders are trained to prefer. Instead of only selling software to people doing the work, it used its own software to do the service.
That is the pattern worth noticing.
The tool became internal infrastructure. The business model became service delivery. The value capture moved closer to the outcome.
For a service-company CEO or COO, the question is not whether your company should become OffDeal. The question is which parts of your own operation are still priced and staffed as if every unit of work must be manually pushed from step to step.
If lead intake improves but quoting does not, quoting becomes the bottleneck.
If document processing improves but exception handling does not, exception handling becomes the bottleneck.
If customer communication improves but scheduling does not, scheduling becomes the bottleneck.
Every partial automation creates a new constraint somewhere nearby. That is why AI projects that look successful in demo form often disappoint in production. They improve one motion inside a system that still moves at the speed of its slowest unmanaged handoff.
The operating model has to change with the tool.
Public-company evidence
This is also showing up outside startup case studies.
Intuit's Q3 2026 update showed AI moving from a feature layer into workflow execution. Coverage of the earnings call noted that Intuit's accounting AI agents are already powering recommendations across more than 50 million transactions each week.
That is not a chatbot story.
It is a workflow story.
Intuit is also moving toward consumption-based AI pricing and autonomous workflows. Again, the direction is clear: price moves closer to work completed, value delivered, or workflow executed.
This matters for service companies because the same pricing pressure will move downstream.
Customers will not use the same words investors use. They will not ask whether you are "AI-native." They will ask why another provider can respond faster, quote faster, process faster, follow up faster, or handle exceptions with less friction.
The competitive comparison will arrive as a service expectation, not as an AI conversation.
The missing middle
The dangerous interpretation is that service companies should now chase automation everywhere.
That is how the next wave of expensive disappointment gets created.
Florian Schroeder's piece, Stop Building AI Agents. Build AI Operations Systems Instead, names the missing middle with unusual clarity. AI does not remove the need to define the work. It makes the definition of work unavoidable.
If the process is unclear, AI scales the confusion.
If the approval path is weak, AI creates more work that needs approval.
If the customer promise is vague, AI produces faster versions of vague output.
This is the part many service companies will miss. The economic opportunity is real. Services can inherit some software-like leverage. But the leverage does not come from installing AI on top of the current mess. It comes from making the work legible enough that AI can safely participate in it.
That means naming the workflow.
Naming the handoffs.
Naming what good output looks like.
Naming which decisions can be automated, which need human judgment, and which should never be delegated.
Naming where the result is stored so the next run improves.
In other words, the company needs an operating layer, not a collection of disconnected AI tools.
The service-company translation
For a 60-person service company, this shift will not begin with a grand transformation program.
It will begin with a specific area where the work is important, repetitive enough to learn from, and painful enough that improvement matters.
Claims intake.
Invoice processing.
Policyholder feedback.
Dispatch exceptions.
Training from customer calls.
Daily operational check-ins.
The first move is not to ask, "Which AI tool should we buy?"
The first move is to ask, "Which part of the company should become more knowable to itself?"
That is where the software-like economics begin. Not in the model. Not in the dashboard. In the company's memory of how the work actually happens.
Once calls are captured, documents are structured, decisions are recorded, exceptions are tagged, and outcomes are fed back, the company starts to build an asset that compounds. Every future workflow can query the same foundation. Every improvement has somewhere to land. Every repeated issue becomes visible earlier.
This is what software companies have always had: the ability to improve the system centrally and let the improvement propagate.
Service companies have usually had the opposite: improvement trapped in a supervisor's head, a spreadsheet, a training session, or a process document no one opens.
AI does not fix that by magic. AI makes the cost of leaving it unfixed much higher.
The new management question
The question for service-company leadership is changing.
It used to be: how many people do we need to deliver this volume of work?
The better question now is: how much of this work can become a managed system that improves every time it runs?
That does not mean removing people from the company. In many service businesses, the value is still human judgment, trust, escalation, and relationship. The point is not to pretend people are unnecessary.
The point is to stop using people as the glue for work the company should have systematized years ago.
When people are the glue, growth adds drag. When the operating layer is the glue, growth can start adding leverage.
That is the practical meaning of "services is the new software" for a service company. It is not a slogan about valuation. It is a warning about operating discipline.
The next competitor may not have better people than you.
It may have the same category of people, working inside a better operating system.
The position
Sequoia is naming the category shift. Y Combinator is naming the margin logic. OffDeal is showing the operator pattern. Intuit is showing the public-company version. Schroeder is naming the work required underneath it.
Read together, the message is simple:
Services are not becoming software because customers want another app.
Services are becoming software because the delivery system behind the service can now be instrumented, remembered, improved, and priced closer to the outcome.
The companies that understand this will not look like AI labs. They will look like better-run service companies.
Faster where speed matters.
More consistent where consistency matters.
More visible to management where visibility matters.
More able to learn from the work they already do every day.
That is the service-company version of the shift.
The work is not to bolt AI onto the business.
The work is to make the business ready to operate with intelligence inside it.