Building a SaaS company after running an agency is the expected path. It’s seen as the logical graduation: swapping retainers for recurring revenue, services for scalability. I did it. And I can tell you it’s a trap.
It’s a halfway house, a comfortable delusion that you’re building a product when you’re really just building a better shovel. The endgame isn’t a more elegant user interface or a slicker dashboard. The endgame is to get rid of the user entirely. The endgame is agentic AI.
My journey isn't a neat story of a clean exit and a triumphant second act. I haven't sold anything. Instead, I’ve spent the last few years dismantling a perfectly functional, profitable agency model from the inside out to build its replacement. This is the only way to survive what's coming. We're building autonomous AI agents because the SaaS model, just like the traditional agency model before it, is already obsolete. It just doesn't know it yet.
For decades, the marketing agency model has been a bundled product. You, the client, pay a single retainer. For that, you get a slice of a strategist’s brain, a creative team’s ideas, and an account manager’s time to execute the work. It’s inefficient, opaque, and fundamentally broken.
Clients don’t want to pay for your overheads, your office ping-pong table, or the junior executive’s learning curve. They want outcomes. For years, SaaS was pitched as the answer. Don't pay a £5,000 monthly retainer for social media management; pay £500 a month for a scheduling tool and do it yourself.
It was a powerful narrative. It unbundled the cost but not the labour. The client, or their in-house team, now had to operate a complex stack of 10, 20, sometimes 30 different SaaS tools. They became the systems integrator, the human API gluing together platforms that were never designed to speak the same language. The result? CMOs drowning in logins, chasing data across a dozen dashboards, and still relying on human effort to pull the levers.
Agentic AI finishes the job. It unbundles the execution. An 'agent' is not a tool you operate. It's an autonomous system you task with a goal. 'Increase our share of voice for non-brand keywords in the North West' or 'Generate a 15% uplift in qualified leads from our existing email database'. The agent then marshals the resources—the data, the creative assets, the channel APIs—to achieve that objective without direct human command.
Look at a brand like Gymshark. Their marketing is a masterclass in community and content. But behind the scenes is a huge operational effort of content scheduling, analytics, and channel management. The agency of the future doesn’t offer to do this work for them. It provides an agent that executes it autonomously, at a fraction of the cost, 24/7. The agency’s role shifts from ‘doing’ to ‘directing’. You become a portfolio manager of AI agents, with your value residing in strategy, exception handling, and architecting the overall system.
Automation is a Feature. Autonomy is the Product.
Let's be clear. Most of what is being sold as 'AI marketing' today is just workflow automation with a fancy label. It’s connecting Mailchimp to a CRM and calling it intelligence. It’s a chatbot running on a decision tree. This is not defensible. It's a feature, not a business.
Automation performs a predefined task. Autonomy achieves a goal. This is the critical distinction. An automated system will send an email when you press a button. An autonomous system will decide who to email, what to send them, when to send it, and will continue to adapt its approach based on the results until it hits a target, like 're-engage 10% of our lapsed subscribers'.
This is what we've been building. When we started developing Competable, our competitive intelligence platform, the initial impulse was to build a better dashboard. We would show clients more data, prettier graphs, deeper insights into their competitors' pricing and product catalogues. We fell into the classic SaaS trap.