Greg Isenberg just said the quiet part out loud: building AI agents is the new SaaS.
Not "agents are a cool feature". Not "add AI to your product". A whole new wave, with the same life-changing upside SaaS had 15 years ago, when 21-year-olds were stumbling into ideas that printed money.
His argument is simple. SaaS sells software. Agent businesses sell work. And work (labor) is a multi-trillion dollar market, way bigger than the market for tools.
I watched the full episode and pulled out the entire playbook: how to find the niche, pick the workflow, build the first agent, price it, and sell it. With his actual pricing numbers, because that's the part everyone skips.
The mental model: the product is the job
One sentence from the episode explains the whole shift.
→ SaaS says: "here's a tool your team can use."
→ Agent SaaS says: "here's a job your team no longer has to do by hand."
Sounds like a small change. It's not. You're not selling seats anymore, you're selling a service that used to require a salary.
Greg's examples make it concrete. Slang AI sells an AI superhost for restaurants: it answers inbound calls during the dinner rush, handles guest questions, manages reservations, routes VIPs, and plugs into OpenTable and Yelp. The restaurant isn't buying software. It's buying the missed reservations back.
Sameday does the same thing for home services. Plumbers, HVAC, roofers, pest control. AI receptionists and dispatchers that answer calls 24/7, respond to texts, book jobs, reschedule. More revenue from the same demand.
The filter for your own idea: "I handle this one annoying job better than a junior employee, faster than an agency, and cheaper than adding headcount."
If your idea doesn't pass that sentence, it's a tool, not an agent business.
Pick a workflow with a paycheck attached
This is where most people mess up. They start with what AI can do. Greg starts with what businesses already pay humans to do.
If someone is paying an employee, an agency, a receptionist, or a dispatcher to do the work, the paycheck already exists. You're just competing for it.
A good agent workflow has 5 traits:
- It happens all the time. Daily is good, hourly is better. Every inbound lead, every call, every ticket, every quote request.
- It has a clear finish line. The job got booked. The ticket got categorized. The refund got approved.
- It touches software that already exists. Gmail, Slack, Shopify, HubSpot, Zendesk, Stripe. Agents need tools they can use and context they can read.
- The edge cases are annoying but learnable. Too basic and a Zapier zap kills you. Pure human judgment and your v1 breaks. The sweet spot is repetitive work with just enough judgment.
- The buyer can feel the loss. Missed calls, slow replies, dropped leads, empty calendar slots.
His first rep for finding one: pick a niche and write down 20 jobs people in it complain about.
Roofers: missed calls, financing questions, insurance paperwork, appointment reminders. Med spas: lead qualification, no-show recovery, membership upsells. Shopify brands: returns, exchanges, wholesale follow-ups.
Then score each job on 5 things: how often it happens, how expensive the pain is, how easy it is to know when the job is done, what tools it needs access to, and who already owns the budget.
If you want more raw material for this exercise, I broke down a full list in AI agent business ideas.
Shadow the human before you build anything
This is the step almost everyone skips, and Greg calls it the unfair advantage.
Before you prompt anything, before you write a line of code, watch a human actually do the job. 10 to 20 real reps. Ask them to screen record it. Ask them to narrate.
What makes a case easy? What makes a case weird? What do they check before deciding? Where do mistakes happen?
His restaurant host example nails why this matters. On paper, the job is "answer: what time are you open?". The real workflow is deeper: the host knows when the kitchen closes, which tables fit strollers, when the patio is shut, how to spot a VIP, and when a call is actually a private dining inquiry worth routing to a manager.
The detail is the product.
Then write the agent spec. Seven parts:
- What wakes the agent up?
- What context does it need?
- What tools can it use?
- What is it allowed to do on its own?
- Where does it need approval?
- When should it escalate to a human?
- What does success look like?
Answer those seven and you're building something real. Skip them and you're building agent slop for a Twitter demo.
Build the smallest useful agent (not an AI employee)
Most people hear "agent" and imagine a fully autonomous employee. That's how you end up with demos that impress on X and fail with real customers.
Greg's version: the Minimal Useful Agent. Four good first versions, in a ladder:
- Draft and approve. It reads context, drafts the reply or quote or summary, and a human approves it. Great when there's risk in getting it wrong.
- Triage. It classifies inbound work and routes it: maintenance request, billing issue, refund.
- Coordinator. It moves between systems and people: checks availability, sends reminders, chases missing info.
- Bounded action. It does one specific thing under clear rules, like processing refunds under $50. This is why your Uber Eats refund shows up before you finish typing the complaint.
He also quoted a point from Anthropic's agent guidance that I think about a lot: many agent problems should start as workflows. A workflow follows a predictable path. An agent decides dynamically. You earn autonomy by starting predictable and adding judgment only where it creates value.
So day one is one workflow and one promise. "We answer missed calls for roofers and book qualified jobs." That's enough. Your customer is buying an agent for the first time in their life. They don't want all of it at once, especially from someone who isn't Salesforce.
The wrapper is the SaaS
Here's what separates a cool automation from a company: the agent does the work, but the wrapper creates the trust.
Customers need to see what happened: logs, approvals, handoff rules, a way to test the agent before it goes live, and a reason why it did what it did.
The agent lives in the phone system, the inbox, the Slack channel. The wrapper is the control room, and the control room is what they're subscribing to.
This is also where evals come in, and Greg framed them in a way I haven't seen before: as a sales asset.
Take 50 real examples of the job (50 calls, 50 leads, 50 maintenance requests). Mark the right answers. Run your agent against them. Now imagine the pitch:
"We tested this on 50 of your old maintenance requests. It routed 42 correctly, flagged 6 for human review, and made 2 mistakes. Here are the two mistakes, and here's how we fixed them."
Try losing a deal after saying that to the owner of a boring business. Transparency closes.
Price it like labor, then productize
The fastest path isn't building a polished product and hunting for users. It's the reverse: sell a pilot where you do the work with AI, then productize the parts that repeat.
Start with 3 customers in one niche. Same niche, same workflow, same pain. Sell the outcome: "we will answer and qualify your missed calls."
His actual pricing examples:
→ $1,500 setup + $1,000/month for one workflow
→ $2,000 setup + $30 per qualified appointment
→ $3,000/month for up to 500 handled tickets
Notice the anchor. That's employee money, not software money. A $29/month tool competes with other tools. A $1,000/month agent competes with a $3,500/month salary and wins.
Greg thinks outcome pricing is where all of this ends up, and I agree, but he's clear: don't jump straight there. The exact price matters less than what the pilot teaches you. What does the customer value? Where does the agent break? What needs approval? What would they miss if you took it away?
Then you productize the repeated pattern. If every roofer needs the same emergency call script, service area check, financing answer, and estimate follow-up: boom, you have a product. You earn the software by doing the work first.
Distribution: make fun of the old way
The content format working right now, according to Greg: workflow teardowns.
Show the old way. A call comes in, nobody answers, the customer calls the next company. Or the CSR answers, asks five questions, checks the calendar, books the job, writes the notes, and forgets the follow-up.
Then show the agent way. Call answered, right questions asked, service area checked, appointment booked, CRM updated, confirmation sent, weird cases flagged to a human.
The owner watching that feels the pain in their chest. You're selling painkillers, not vitamins.
Pick one workflow and make the internet associate you with it. Make the checklist, the benchmark, the teardown, the "50 examples of this workflow" post. Then take the assets that work and put paid ads behind them. One platform to start.
I run my whole content operation on agents, and this teardown format is the same reason founder stories work on this blog. Before and after, with numbers, beats any feature list.
The 30-day plan, day by day
If Greg were starting from zero, this is the calendar:
- Day 1: pick a niche where missed work costs money. Home services, property management, insurance agencies.
- Day 2: interview 10 operators. Ask them to screen share the workflow. Pay them if you have to.
- Day 3: pick one workflow with frequency, pain, software access, and a clear success metric.
- Day 4: write the 7-part agent spec.
- Day 5: run it manually with AI. Copy-paste context into Claude, draft outputs, have a human approve. You're testing whether AI helps before you build software.
- Day 6: build the smallest useful version. Draft-and-approve or triage is usually enough.
- Day 7: build the eval set from 50 real examples.
- Week 2: sell two pilots in the same niche.
- Week 3: add the wrapper: logs, approvals, settings, handoffs. Use AI to build it.
- Week 4: publish workflow teardowns, turn the pilots into proof, double down on content.
And the whole time, you're building an audience. By month two and three you're figuring out LTV and which channel deserves paid spend.
One warning from my side: once you sell a few of these, the bottleneck moves from building agents to running them. I wrote about that in managing AI agents, because "agent manager" is quietly becoming the actual job.
FAQ
Are AI agents really a bigger opportunity than SaaS?
The argument is about market size. SaaS competes for software budgets. Agents compete for labor budgets, and labor is a multi-trillion dollar market. A business that pays a receptionist $3,500/month will happily pay $1,000/month for an agent that does the same job 24/7. That budget was never available to a normal SaaS tool.
What's the difference between an AI agent and an automation?
An automation follows a fixed path: trigger, steps, done. A Zapier zap. An agent handles judgment inside the workflow: it classifies a weird request, asks for missing info, decides when to escalate to a human. Greg's rule (borrowed from Anthropic) is to start with the predictable workflow and add judgment only where it earns its keep.
Do I need to be technical to build an agent business?
Less than you think. The 30-day plan has you running the job manually with AI (copy-paste into Claude, human approves) before any software exists. The hard parts aren't code: they're picking the right workflow, shadowing the humans who do it, and building the 50-example eval set. Those are founder skills, not engineering skills.
How should I price an AI agent service?
Start simple: a setup fee plus flat monthly, like $1,500 setup and $1,000/month for one workflow. Once you understand the value, layer in usage or outcome pricing, like $30 per qualified appointment or $3,000/month for up to 500 tickets. Anchor against the cost of the human doing the job today, not against other software.
Steal these playbooks every week
The founders who win this wave will be the ones who hear the playbook early and actually run it.
That's literally why the Profitable Founder Podcast exists. Every week I sit down with bootstrapped founders doing $100K to $10M a year and pull out the exact numbers, pricing, and channels behind their businesses. The agent-SaaS wave is minting a new batch of these stories right now.
