Turning manual AdOps into software.
80× ad spend growth in two years. The team never grew past four people. I build the systems that make campaign management unnecessary — so operators supervise exceptions instead of executing repetition.
I don't optimise campaigns. I engineer the machine that runs them.
Turn messy AdOps workflows into explicit state machines, rules and decision trees — the prerequisite to automating anything.
Locate the 10% of operational logic that produces 90% of execution quality, and spend the engineering there.
Replace checklists with QA gates that cannot be skipped. Quality becomes a property of the system, not the operator.
Build agents that run campaign workflows end to end, and define the small set of decisions still worth a human.
Four things I believe that most of this industry doesn't say out loud.
Manual work in AdOps is not expertise. It's evidence of missing software.
Campaign setup, QA, pacing checks, reporting, client updates — the industry calls these craft. They're symptoms of a system nobody designed.
The future isn't AI-assisted operators. It's software-operated campaigns with human governance.
Every vendor sells assistance, because assistance is comfortable to buy. Execution is the harder and more valuable thing to build.
Headcount is the most expensive way to buy throughput — and the default way this industry buys it.
I've built the thirty-person version of an ad operations department. Then I built the same capability with four people and software.
Campaign quality is a constraint problem, not a skill problem.
Errors aren't a training gap. They're a property of system design. Find the bottleneck, kill it with software, find the next one.
ABM, retargeting, audience onboarding and DSP execution on first-party audience data — across thousands of campaigns. Not just media buying: designing the repeatable operational systems underneath it.