I build and run go-to-market and operations, now with AI agents.
14 years in GTM and operations. CRO & COO at SEOmonitor, VP of Global Sales at ContentKing through its sale to Conductor, and enterprise sales at Botify, where I finished #1 of 16 reps worldwide. In 2026 I co-founded Flip Education and ran it with one other person and 70+ scheduled AI agents: 18 countries, 11 languages. Advisor since 31 Aug 2026.
What's published here is one system, extracted from running a real company on agents. The thesis: as agents get smarter, you stop managing them by task and start managing them by mandate — a goal with a number on it, a constitution of operating rules, and a daily loop against results. The judgment lives in the rules, not the prompts, because prose decays and encoded rules hold.
The pieces, as they publish:
| Layer | What it does | Repo |
|---|---|---|
| Evaluation | Decide which model runs each job on evidence, not vibes: rubric judging, judge panels, cost/quality Pareto reports | pareto-evals |
| Enforcement | Quality gates at the write/deploy boundary: secrets, spend, sends, scope, slop, silent failure | agent-gates |
| Memory | Async follow-up queue so deferred verification survives the session that deferred it | follow-up-queue |
| Law | The agent constitution: goal + operating rules + loop, per function, on a shared base | (memo in progress) |
Behind them at Flip Education: written agent charters, ~50 automated quality gates, continuous evals, pre-spend cost controls, silent-failure tripwires, a daily cost ledger. None of it education-specific.
raianpollock.com · raian@raianpollock.com · Barcelona