The operating model: how the team runs

The machine optimises.
The team builds.

This isn't a proposal, it's a description of my current team. Everything on this site (the pages, the experiments panel, the location demos) was built and shipped the way my team ships every week: idea in a sprint, built by the person who had it, reviewed with AI, live the same day.

The shipping pipeline: real and running today

Marketing initiatives don't need product-grade guardrails, so we don't pay product-grade shipping costs. The whole team plugs into the same pipeline.

1
GitHub
Anyone pushes
The whole team works in GitHub, every specialist included. Have an idea in the sprint? Build it and push it. The platform is flexible enough that anyone can create anything.
2
AnthropicClaudeGemini
AI reviews
Every change is reviewed with the best models available. Does it break anything, does it meet the bar? Guardrails are scaled to marketing risk, not product risk.
3
Cloudflare
Live on Cloudflare
Approved changes deploy straight to the edge. Shipping takes a day, not a quarter, so the cost of trying something rounds to zero.
4
The system learns
Everything ships into the measurement layer, the league table and the experiments panel. Winners scale, losers retire and nobody has to defend a corpse.
Proof in your hands: this very site, 20+ pages, an experiments back end and two location demos, was built exactly this way in days, in Lodgify's own design system. The pipeline you're reading about produced the thing you're reading.

"You're not just a channel specialist."

The transformation
In my current role I took a team of content writers and turned them into builders. They ideate in sprint 1:1s, build it themselves, push it live. The job title stopped being the ceiling. Here the team is paid-media specialists, and the same unlock applies: an AdWords specialist stops being “the person who manages campaigns” and becomes someone who ships landing systems, experiments and agents.
My job: push the boundaries
I prototype the new thing myself first. I run my own local AI models at home, so I know what's possible before I ask for it. Then I teach it into the team, set challenges just past comfortable and give people a prod when they're stuck or thinking too small.
Live example: the team's current stretch project, agentic pipelines
Scanner agent: works city by city, collects the raw local data → Verifier agent: checks every claim → Publisher agent: loads the CMS → Loop: every town, UK & US
Content writers six months ago, building agent flows today, targeting a ~50× increase in site coverage and per-location relevance. The same trajectory is open to a PPC specialist. At Lodgify, this pipeline is the location-flywheel factory: it's how 400+ city pages get built from live inventory without a content-team bottleneck.

No silos, no dead budget

Channel ownership breeds budget protectionism: dead spend survives because it's tied to someone's salary and someone's story, and they'll keep flogging that dead horse rather than admit it. Release people from "owning a channel" and the incentive flips. The only question left is "what's the highest-impact thing I could build right now?"
Case in point: the near-zero-ROAS Display and Demand Gen spend in this account is textbook protected budget. Nobody defends it once nobody's role depends on it.

The question changes

"How is this keyword performing? Can we squeeze 3% more?"

That keyword is well-optimised. The bidding is automated. The league table runs the messages. The machine has that covered.

"What are you bringing that's new, or that improves the situation completely?"

That's the sprint question, every cycle, for every person.
Back to the cover→The experiments back end →The message league table →