Mediafly sells to sales teams: the content they carry, the tools that model the value of a deal, and the analytics underneath. By the time I joined as Chief Product Officer it had grown by acquisition into six products on six separate stacks, poorly joined, with no single experience across them. Each product had its own team, its own stack and its own history, and those teams had already been through several reductions.
It was the hardest situation I have walked into. Within weeks there was another reduction to plan, with about two weeks to plan it, because the company needed financing and investors needed a plan they could believe.
I knew what it would cost. People leave after a reduction.
I knew from Unbounce that after a reduction the best people leave first. There, where a rebrand kept the people we needed. Here, nobody would have believed a brand story. So the retention plan was autonomy.
Everyone who stayed chose the team they wanted to join. I asked what they cared about, and what they would do first if they were the CEO, and gave them room to run the experiment quickly, with clear expectations. We still lost good people. Nobody had been cross-trained, and the people who remained wore many hats. That was the cost, and I signed the plan knowing it.
Joining six stacks would have taken years we did not have.
Nothing on this list worked on its own.
Three teams, not six. One on core infrastructure, the shield for everyone else, doing the maintenance and incremental work, staffed by people who wanted that work. One building the features customers asked for most, across products. One around a product manager with a marketing background who was new to AI, with the people who wanted to build what came next.
Fewer products, with Intelligence at the centre. We focused where we could win, and put the analytics brain underneath everything instead of beside it.
The agentic revenue operating system. It is very hard to tell salespeople what to do. It is much easier to hand them the best story at the moment they meet a customer: an assistant in the field, not more tasks. Value models built from a few inputs in minutes instead of weeks with a specialist, slides that rebuilt themselves for the person in the room, a background brief on a public company to start from, and a loop back to marketing on which content actually closed. What closes deals is less time between interactions, so that was the measure.
A channel, not a product gap. When the board wanted a learning platform, we bought Appinium. Most people said it filled a product gap. I was for it for a different reason: it was built on Salesforce and gave us a way into that ecosystem. The plan was one layer across all the stacks, built on Intelligence, rather than years of rebuilding them into one. There were flaws in that idea. It was the best one we could execute in the time we had.
The runway to do it. I helped raise $25 million across two rounds.
The first thing we built was an automatic requirements document for the value-selling tool: work that took a specialist weeks, done in minutes. One customer who had implemented a version saw, in his own testing, what less time between interactions did for his sales team's conversion. Search built for AI, and the data architecture for the layer across the stacks, were underway when I left at the end of 2025.
People, then focus, then product. I would choose that order again.