TheZebra appoints Daniel Herrington as chief AI officer to modernize insurance shopping
TheZebra.com has named Daniel Herrington as chief AI officer to accelerate a simpler, smarter insurance shopping experience. Herrington brings more than a decade of AI product leadership with roles at Google, Vroom, Capital One and Priceline. At Google, he worked on training and serving infrastructure for Gemini while leading the CoreML team.
Based in Austin, TheZebra is positioning AI at the center of its product and operations. Herrington will guide partnerships with leading AI labs and drive AI use across developer experience, customer empowerment and advisor enablement.
What Herrington's charter means for product development
- Developer experience: roll out AI tooling, model infrastructure and CI/CD for prompt, agent and feature iteration across engineering.
- Customer empowerment: build search and selection flows that clarify trade-offs, reduce cognitive load and close the gap between quote and bind.
- Advisor enablement: give human advisors copilots that summarize, suggest next best actions and flag compliance risks in real time.
- Company-wide lift: apply AI to product, marketing, finance and legal workflows to shorten cycle times and improve accuracy.
Quote from leadership
"AI enables us to make the process of buying and managing insurance clearer, faster and more personalized for consumers, while also empowering our advisors with smarter tools to deliver better results. Daniel's experience building sophisticated AI products at-scale, paired with a deep understanding of how technology should serve real people, is exactly aligned with our values and vision," said CEO Keith Melnick.
How this likely shows up in the product
- Search that surfaces intent, not just keywords-real-time extraction of risk factors, preferences and constraints.
- Matchmaking that weighs coverage needs, price sensitivity and carrier appetite, then explains the "why" behind recommendations.
- Advisor tools that pre-fill context, detect gaps and propose compliant scripts or follow-ups.
- Faster experiments: prompt versioning, offline evals and online guardrails so teams can ship, learn and roll back safely.
A practical roadmap for product teams
- Data foundations: unify policy, quote and interaction data with clear PII boundaries. Log prompts, responses and user actions for later evals.
- Model strategy: blend retrieval + lightweight models for speed, with pathways to larger models for complex reasoning. Keep a fallback path.
- Evaluation: define offline suites (accuracy, relevance, safety) and online KPIs (bind rate, time-to-quote, NPS, advisor handle time).
- Safety and compliance: apply the NIST AI Risk Management Framework. Add PII redaction, consent tracking and hallucination checks.
- Cost control: cache, distill, and trim context. Route by difficulty. Monitor token and inference budgets per feature.
- Human-in-the-loop: require advisor or QA review for high-risk actions. Capture feedback to retrain prompts and policies.
- Explainability for users: show the factors behind each recommendation, alternatives considered and expected trade-offs.
- Tooling: standardized prompt packs, feature flags for AI components, and golden datasets for regression on every release.
Org implications
- Create a small AI platform team to own model access, evals and guardrails; let product squads ship on top of that surface area.
- Set a weekly rhythm: ship small, measure, review failures, update prompts/policies, and document learnings.
- Make "explain the decision" and "measure the win" mandatory in PRDs for any AI-backed feature.
What to watch next
- Partnerships with top AI labs and how that affects latency, cost and feature velocity.
- Upgrades to search and match UX-clearer coverage explanations and fewer steps to bind.
- Advisor copilots moving from assistive to trusted co-editors with guardrails.
- Cross-functional lifts in marketing, finance and legal from applied AI tooling.
For more hands-on frameworks and examples relevant to this move, explore AI for Insurance and AI for Product Development.
Context on the model work referenced: Google's Gemini platform overview is here for background on capabilities and constraints Gemini.
Your membership also unlocks:
