E27: How Agencies are becoming Software with Bennett Spooner
In this episode of The SaaS Operators Podcast, we sat down with Bennett Spooner, Co-founder of OperatorOS, to explore how he’s turning an automation agency into something that looks and feels like software. Bennett shares how his team built a customer service automation system that handles up to 90% of support without humans, and why Foreplay’s API is at the center of how he's productizing agency work. He explains how APIs, agents, and e-commerce automation are reshaping the line between service and SaaS, and why scalability isn’t about code, it’s about leverage. We dive into the debate over service versus software models, how to think about margins and multiples when you’re in the middle, and what the next phase of AI-powered automation means for operators at both marketing agencies and software companies. The episode wraps with a bigger conversation on what it takes to build a business that compounds while the rest of the industry tries to catch up.
This week on The SaaS Operators Podcast, Bennett joined us to talk about the gray zone between agency and software. He’s building an automation company that’s starting to behave like SaaS, using Foreplay’s API to productize client work, scale support automation, and push creative research into code. It’s a look at what happens when service businesses start to think like software companies.
The agency that behaves like software
Over the last year, Bennett’s team took every custom automation that hit their inbox. Different stacks. Different industries. A fresh audit every time. It grew revenue, but it killed leverage. You cannot delegate a new process ten times a week.
So he is narrowing to e-commerce and productizing the work. The flagship right now is customer service automation that resolves up to 90 percent of tickets. There is a small RAG, a chat interface, escalation logic to humans, and alerting into Slack for edge cases like fraud or angry repeat refunders. It works. Clients are using it.
Where it goes next is creative. He is wiring Foreplay’s API into a “creative ad specialist” role so brands can research, brief, and ship faster. This is exactly why we shipped an API at Foreplay. Features can land for power users first, then come back into the core product when the patterns are clear. The flywheel is obvious. Better API. Better integrations. Better data. Better core product.
Should every modern tool ship an API?
Short answer from the group was yes. APIs unlock weird edge cases and custom workflows that a product team should not bake into the main UI. They also let an ecosystem do your R&D in the wild.
Bennett added a second angle. Web agents are creeping toward parity with APIs for certain jobs. If an agent can navigate, paginate, and click reliably, it can use your web app the way a human would. It is not there yet. It might get there. In the meantime, APIs will win on speed and scale. Either way, tools that are callable win.
Service versus SaaS is the wrong fight
Bennett still needs humans to onboard. That makes him feel like a service. But the outcome he sells feels like software. It is recurring, measurable, and mostly automatic once installed.
Jeremiah’s point: pure SaaS often serves the mean. When you build for a thousand customers at once, you blunt the edges that make a solution great for one company. A custom implementation can go deeper and hit 90 percent automation today. A generic product usually cannot.
My view is simple. Sell the outcome. Nail retention. Call yourself whatever you want on the way to software margins. Palantir did heavy implementation for years, then the margins showed up. Labels do not matter. Unit economics do.
Zach framed it with a sanity check on valuation. If you want a SaaS multiple, you need subscription billing, high gross margins, and some IP that cannot be copied overnight. If your magic is mostly playbooks and people, it is a productized agency. That can still be a great business.
Will automation be automated?
Bennett is confident. In the next one to two years, he expects a platform that can generate a full app, backend and frontend, with automation baked in. One prompt. No scaffolding. No glue code. If that is true, the winners will be the best sellers and consultants, not the best tinkerers. He is already training his students to be “AI transition partners.” Learn enough about RAG, APIs, and voice agents to be dangerous. Spend most of your time on value, messaging, and closing.
Zach pushed on the obvious question. If setup becomes trivial, why would an operator capture the value instead of the platform owner? Bennett’s answer is adoption. People resist change. They are busy. They are skeptical. They need a human to say, “I will do this for you and be accountable for the outcome.” Even when the tool is self-serve, implementation partners tend to exist. HubSpot proved that. So did Salesforce.
My take: abundance does not auto-convert to outcomes. Information has been effectively free for a decade and most teams still botch the basics. There is always surface area on the margin for operators who turn tools into results.
Why e-commerce, and is it a sinking ship?
We pushed Bennett on his ICP. He is picking e-commerce because decision makers are younger, more open to automation, and live in fast feedback loops. He sees a shift coming from search and social to AI surfaces. If ChatGPT or Gemini can rank and recommend with real context, new demand will flow through those interfaces.
Jeremiah and Zach were less bullish on consumer recommendations from chat. The context degrades unless you keep it fresh. Meta still sees our emotional triggers passively, which is powerful for discovery. I asked if Google wins anyway because it is embedded everywhere. Zach’s counter was taste. A distribution moat without product taste makes forgettable consumer products. We all agreed APIs and data partnerships are the pragmatic way to compound while the big platforms fumble UX.
Operators versus visionaries
We ended on Apple, Google, Meta, and why product taste feels scarce. Tim Cook is an elite operator. He shrank the iPhone supply chain lag from 30 days to three. That is constrained excellence. It does not guarantee new category creation. Meanwhile, Meta shipped Ray-Ban and Oakley collabs that feel like the right move. If you lack taste, rent it from a brand people already love.
This matters for founders. Know which game you are playing. If you are not the unconstrained builder, be world-class at the constrained one. Both win when paired with timing and distribution.
What I am watching next
- Foreplay’s API as a proving ground. We will keep pushing features to builders first, then generalizing what works back into the product. Expect more data partnerships and more ways to pull paid creative into your workflow.
- Service margins drifting toward software. If Bennett keeps 90 percent automation on support and wraps creative intelligence around it, the onboarding pain will shrink and the gross margin will rise.
- Agent reality versus agent hype. APIs will carry the load near term. Agents will catch up where latency is tolerable and determinism is not critical. The winners will abstract both behind clean promises.
- The adoption gap. The tool that “does it for you” will still need humans to say “yes” and to own outcomes. There is a decade of work for AI transition partners before anything looks fully automatic.
Bennett came in as an agency guy who wants to be software. He left sounding like a builder who understands where leverage actually lives today. Ship outcomes. Wrap them in clean pricing. Use APIs to punch above your weight. Let the market call you whatever it wants while you compound.
