Eilon Reshef: Why Gong pods use dozens of design partners

Each Gong pod includes a PM, designer, engineering lead, 5–7 engineers, and fractional analysts/writers, and owns a defined problem space (e.g., forecasting, conversation intelligence) rather than a metric or a narrow feature. This gives them end‑to‑end ownership and clarity while avoiding over‑centralized control.

Pods work closely with 6–24 existing customers in the target ICP, meeting regularly, demoing partial builds, and iterating based on real workflows. This "extreme" design‑partner model leads to ~95–100% of built capabilities being significantly used, far above typical enterprise SaaS hit rates.

Gong created a “research coordinator” role and a micro‑CRM to source, segment, and schedule design partners for all pods. Offloading operational overhead makes it realistic for dozens of pods to maintain deep, ongoing customer contact.

Teams are expected to decide when to ship, when to ask for help, and which customer requests to act on, and leadership deliberately accepts lower visibility in exchange for higher velocity and engagement. For most 51/49 decisions (where both options are acceptable), Reshef pushes for fast calls rather than weeks of analysis.

Gong combines LLMs with bespoke models (e.g., deal prediction) and maintains data-scientist and prompt-engineering expertise to decide what’s feasible, design evaluation metrics, and systematically improve prompts and models. Without this rigor, AI features quickly hit a quality ceiling and stall at “V1 demo” quality.