You've been shipping features and polishing dashboards. The market is quietly deciding it would rather rent workers.
SaaS was a brilliant model. Build once, sell to many. Give everyone a login, charge per seat, ship updates continuously. For two decades, this worked beautifully.
Then agents showed up.
Not chatbots. Not glorified autocomplete. Actual agents that take a goal, break it into steps, execute those steps, and deliver a result. The kind of software that doesn't wait for a user to click buttons. It does the work.
You've built your product on the SaaS playbook. You've got a feature roadmap, a pricing page with three tiers, maybe an enterprise plan. And now there's this nagging feeling that the ground is shifting underneath you.
It is. Because once your software does the work, you're no longer selling a tool. You're renting out a worker.
The Uber Shift
Think about car rentals vs. Uber. Both get you from A to B. Both charge you money. But the relationship is completely different.
A car rental gives you the vehicle. You drive. You navigate. You find parking. You're paying for access to a machine.
Uber gives you the ride. You say where you're going. Someone else drives. You're paying for the outcome.
SaaS is car rental. Here's the software. Here are the controls. Here's the dashboard. You figure out how to get where you're going.
Agent-based software is Uber. Tell us what you need done. We'll handle it.
Both cost money. One gives you a tool. The other gives you the work, finished.
The Swiss Army Knife Problem
When you sell a tool, you compete on features. More dashboards, better analytics, another integration. The product roadmap is a list of things users can do.
When you rent out agents, you compete on capability. How well can your agent handle edge cases? How fast does it learn from mistakes? How reliably does it deliver the result?
The difference matters. Feature competition is a race you eventually lose, because someone always adds more blades to the knife. Capability competition is about depth. A master carpenter with five tools builds better cabinets than a hobbyist with fifty.
Your moat stops being a feature list. It becomes training data, domain expertise, and the accumulated judgment your agents develop from handling thousands of cases. Just like a staffing agency's moat is the quality of its people.
When Shelfware Becomes Impossible
Here's what's quietly radical about this shift: it forces pricing to reflect actual value.
Seat-based SaaS pricing was always a polite fiction. You paid per user per month whether that user logged in daily or never. The seat price was a proxy for value, never a measure of it. Customers knew this. It's why "shelfware," software you pay for and never use, became its own category.
Agent-based pricing can't work that way. When your agent schedules 200 meetings, you can charge per meeting. When it processes 50 invoices, per invoice. When it generates zero value, zero.
But here's the wrinkle. Salesforce launched Agentforce with $2-per-conversation pricing, and customers pushed back hard. They wanted predictability. Salesforce ended up offering three pricing models simultaneously: per-conversation, per-action credits, and (yes) seat-based licenses.¹ The market wants outcome-based pricing in theory. In practice, it also wants a number it can budget for.
Deloitte's research predicts this tension will define 2026: SaaS vendors experimenting with hybrid models that blend usage and outcome-based pricing, while customers push for simplicity.² The direction is clear even if the destination is messy.
For customers, the trend is liberating. For software companies, it's terrifying and exciting in equal measure. Terrifying because recurring revenue from unused seats disappears. Exciting because the ceiling on per-customer revenue blows open. An agent that saves a company $100,000 per month can command pricing that no traditional SaaS seat ever could.
Bottling Expertise
This is where it gets interesting for indie hackers.
Building traditional SaaS as a solo founder meant competing on features with funded teams. You'd build a to-do app and immediately face a dozen competitors with bigger engineering budgets adding features faster than you could.
Building agents flips the game. A solo founder who deeply understands, say, e-commerce returns processing can build an agent that handles returns better than any generic SaaS tool. Domain expertise becomes the product, and engineering scale stops being the bottleneck.
Intercom is already showing what this looks like at scale. They built Fin, an AI agent that resolves customer support tickets, and priced it at $0.99 per resolution. Not per seat. Not per month. Per problem solved. If Fin doesn't resolve the conversation, you don't pay. That's the staffing agency model in action: you're renting a support rep, and you only pay when they close the ticket.
Now picture the indie hacker version. A solo developer who spent ten years in insurance claims builds a claims-processing agent and charges per claim resolved. A former recruiter builds a candidate-screening agent and charges per qualified shortlist. A photographer builds an image-culling agent that works the way a professional editor would, billed per shoot.
Each of these people is doing the same thing: bottling their expertise into an agent and renting it out.
The Business Is in the Matching
Staffing agencies figured something out decades ago that software is just now learning: the value is in the matching and the maintenance.
Recruit the right person. Train them. Place them with the right client. Handle the payroll, the scheduling, the performance reviews. The client just gets the work done.
Agent companies will look the same way. Develop the agent. Train it on domain-specific data. Deploy it for the right customer. Handle the updates, the edge cases, the monitoring. The customer just gets the outcome.
The interesting question is what happens when agents from different providers need to work together. Gartner's research on multiagent systems³ found that inquiries surged 1,445% from Q1 2024 to Q2 2025. That's the market asking: how do we coordinate these rented workers across company boundaries?
Staffing agencies solved this too. They call it managed staffing. One agency provides the team lead, another the specialists, a third handles admin. They all show up Monday morning and get the job done.
The agent economy will need the same coordination layer.
Tool or Worker?
If you're building SaaS right now, the question worth sitting with is simple: am I building a tool, or am I building a worker?
Some software genuinely needs a human at the controls. Creative tools, design software, anything where human judgment is the point. But for the vast category of software that exists to get specific tasks done, the future looks less like "here's your dashboard" and more like "it's handled."
Software companies are becoming staffing agencies. Salesforce is already fumbling toward this with three pricing models and a product literally called "Agentforce." The transition won't be clean. But the direction is set.
Rabbit Hole
If you're thinking about agent economics, "Agents Learned to Talk. Now They Need to Learn to Pay." digs into the payment infrastructure this new economy will need. For more on how pricing changes when AI does the work, "Pay per result might be the unit test for pricing AI SaaS" explores outcome-based pricing in depth. And for the architectural shift underneath all this, "Every Action Is an Agent" asks what happens when every piece of software becomes agentic.
- Salesforce launched Agentforce at $2/conversation in late 2024, faced customer pushback over unpredictable costs, and by 2025 had introduced Flex Credits ($0.10/action) and seat-based licenses ($125/user/month) alongside the original model.
- Deloitte, "SaaS meets AI agents: Transforming budgets, customer experience, and workforce dynamics," TMT Predictions 2026. The report tracks how agentic AI is pushing SaaS vendors from seat-based licensing toward hybrid usage and outcome-based models.
- Gartner, "Multiagent Systems in Enterprise AI: Efficiency, Innovation and Vendor Advantage." The report tracks the spike in enterprise interest and outlines how orchestrated specialist agents handle complex workflows more effectively than single general-purpose systems.