Mathieu Cognac

Mathieu CognacSep 18, 2026

Synthesia is growing 📈!!! I'm hiring for a Director of Sales Ops and Director of CS/Post-sales Ops, reach out if interested!

Mathieu CognacDec 19, 2025

If your AI only writes emails, you’re training a very expensive assistant — not a coworker. Agents that *act inside* our revenue stack change the work, not just the output. The biggest shift is from point automation to orchestration. When agents can update CRM fields, flag risk signals, and trigger playbooks, we move from marginal time savings to real improvements in forecast reliability and seller capacity. Forrester describes this as the agent-as-coworker model, and in my experience pilots that let agents operate in the stack produced mid-teens percentage lifts in seller capacity and noticeably cleaner pipelines within months. Two things determine whether that upside happens. First: governance. Autonomous updates need clear guardrails, immutable audit trails, and fast rollback paths. Without those, reps distrust the system and revert to manual work. Second: role and skill changes. We’re not removing judgment — we’re shifting it. Teams must learn to teach agents context, validate outputs, and handle exceptions. That requires new playbooks, adjusted KPIs, and focused verification training. Agents also make forecasting more actionable. They can surface buyer signals from meetings, emails, and contract threads and translate them into standardized "sales-ready" triggers. That tighter signal definition narrows subjective inputs and improves pipeline hygiene. I’ve seen early implementations reduce forecast variance materially — enough to make leadership meetings shorter and decisions faster. Build the process before you deploy the agent. Start with high-friction handoffs: proposal approvals, legal checks, pricing exceptions. Those are where rule-based autonomy proves safe and measurable. Prove safety with a small cohort. Measure impact. Then expand. Questions I’m asking myself: - How do we balance agent autonomy with immediate human override on mission-critical deals without slowing the seller? - Which RevOps metrics should prove agent value beyond time saved — forecast variance, adjusted commit accuracy, or quota attainment uplift? - If agents standardize buyer signals, how do we preserve seller creativity that wins complex, non-standard deals? If you’ve run a pilot, I’d love to hear one tactical result: did you go strict with guardrails and slow rollout, or loose rules and fast iteration? Share the one metric you tracked early that convinced leadership to scale. #RevOps #SalesOperations #ForecastAccuracy #RevenueTech #B2BSaaS

Mathieu CognacNov 11, 2025

AI doesn't fix broken processes — it multiplies them. That truth should make every RevOps leader pause before flipping the switch. The clearest takeaway: automation amplifies whatever workflow you already have. Weak qualification rules, fuzzy handoffs, or sloppy forecast hygiene won't be cured by models. They'll scale up the mistakes and make them more expensive. I've seen this play out: a lead score that promotes MQLs after two email opens floods reps with meetings that never convert. A predictive opp score can push deals into commit when no buyer intent exists. Those are process failures, not model failures. Forecast accuracy, territory fairness, and quota balance all suffer when signals are noisy. Bad process, faster. Start with process maps and ownership. Define which buyer actions actually matter and who owns each stage gate. Lock down every signal that moves a record. Require *human-in-the-loop* checks for any model action that affects quota, territory, or forecast numbers. Measure outcomes, not just model metrics. Track cycle time, win rate, and forecast misses after each change. Build rollback plans and rollback triggers so you can undo automations that worsen business results. In my experience, those guardrails separate a safe pilot from a full-scale disaster. Ask yourself • Are our lead-to-account mapping rules strong enough before we add predictive territory assignments — which signals should be authoritative first? • Should we require a buyer-engagement gate before a score can push a deal to commit — what exact metrics would prevent false forecast confidence? I'd like concrete approaches from other RevOps leaders: are you pausing rollouts until processes are defined? #RevOps #RevenueOperations #ProcessOptimization #ForecastAccuracy #SalesOperations