Gentrace’s Doug Safreno on Escaping POC Purgatory with Collaborative AI Evaluation
The Chief AI Officer ShowShow Notes
The reliability gap between AI models and production-ready applications is where countless enterprise initiatives die in POC purgatory. In this episode of Chief AI Officer, Doug Safreno, Co-founder & CEO of Gentrace, offers the testing infrastructure that helped customers escape the Whac-A-Mole cycle plaguing AI development. Having experienced this firsthand when building an email assistant with GPT-3 in late 2022, Doug explains why traditional evaluation methods fail with generative AI, where outputs can be wrong in countless ways beyond simple classification errors.
With Gentrace positioned as a "collaborative LLM testing environment" rather than just a visualization layer, Doug shares how they've transformed companies from isolated engineering testing to cross-functional evaluation that increased velocity 40x and enabled successful production launches. His insights from running monthly dinners with bleeding-edge AI engineers reveal how the industry conversation has evolved from basic product questions to sophisticated technical challenges with retrieval and agentic workflows.
Topics discussed:
Why asking LLMs to grade their own outputs creates circular testing failures, and how giving evaluator models access to reference data or expected outcomes the generating model never saw leads to meaningful quality assessment.
How Gentrace's platform enables subject matter experts, product managers, and educators to contribute to evaluation without coding, increasing test velocity by 40x.
Why aiming for 100% accuracy is often a red flag, and how to determine the right threshold based on recoverability of errors, stakes of the application, and business model considerations.
Testing strategies for multi-step processes where the final output might be an edit to a document rather than text, requiring inspection of entire traces and intermediate decision points.
How engineering discussions have shifted from basic form factor questions (chatbot vs. autocomplete) to specific technical challenges in implementing retrieval with LLMs and agentic workflows.
How converting user feedback on problematic outputs into automated test criteria creates continuous improvement loops without requiring engineering resources.
Using monthly dinners with 10-20 bleeding-edge AI engineers and broader events with 100+ attendees to create learning communities that generate leads while solving real problems.
Why 2024 was about getting basic evaluation in place, while 2025 will expose the limitations of simplistic frameworks that don't use "unfair advantages" or collaborative approaches.
How to frame AI reliability differently from traditional software while still providing governance, transparency, and trust across organizations.
Signs a company is ready for advanced evaluation infrastructure: when playing Whac-A-Mole with fixes, when product managers easily break AI systems despite engineering evals, and when lack of organizational trust is blocking deployment.
More from The Chief AI Officer Show
98% of IT leaders say they're using AI, yet 43% aren't measuring ROI and 80% admit they're not using it to its potential. Olga Lagunova, Chief Innovation and Technology Officer at GoTo, says the industry won the adoption battle but hasn't reached the outcomes it needs. Olga breaks down GoTo's federated AI steward model, where functional leaders own roadmaps and a central enablement team provides infrastructure. She explains why SMB pragmatism shaped their practical AI principles and how 39% of employees already report AI is eroding their skills.
Eric Ries built the Lean Startup methodology, helped advise on Anthropic's long-term benefit trust structure, and is now making an argument most enterprise leaders aren't ready to hear: deploying AI agents inside a company with bad governance isn't just a compliance risk, it's an existential one. His new book, Incorruptible, makes the case that the same slow-moving forces that corrupt companies over decades will be dramatically accelerated by AI, and that the standard governance playbook most executives have been handed is the actual source of the problem. In this conversation, Eric and Ben cover the legal and structural traps that silently strip companies of mission control long before anyone notices, why Anthropic walking away from a $200M Pentagon contract turned into an unexpected competitive advantage, and what AI leaders specifically need to do before they deploy agents at scale. Eric is direct about what he thinks leaders are getting dangerously wrong right now, and he pulls no punches. Topics Discussed: Why standard "best practice" corporate documents are structurally designed to separate mission from control
Intercom launched outcome-based pricing before the market had a framework for it, before inference costs made it profitable, and before customers knew how to budget for it. Fergal Reid, Chief AI Officer, was inside that decision and he shares exactly how they modeled their way through it, what two bets he personally owned, and why they went to $1 per resolution knowing it was a loss. That pricing story is inseparable from their model strategy. Fergal walks through the production data that led them to conclude that Opus 4.5 didn't outperform Sonnet 4.0 on their RAG customer service task, what that told them about the limits of general intelligence at the application layer, and why it pushed them to build Apex, their own model trained via reinforcement learning on an open-weight base specifically for customer service. With 85% of Intercom's own support volume now fully automated, the bets held. Topics discussed: Outcome-based pricing mechanics: the $2 beta, the loss-leader move to $1, and the two assumptions Fergal had to own
What happens when a journalist turned Amazon product manager becomes the Chief AI Officer of one of the world's largest international broadcasters? You get someone who sees the AI threat to media not just as a distribution problem, but as a full production chain crisis that requires a fundamentally different organizational architecture. Marie Kilg, Chief AI Officer at Deutsche Welle, makes the case that legacy media's survival depends on something most AI transformation conversations ignore: data interoperability across systems that were never designed to talk to each other. With 32 languages, siloed editorial teams, and decades of layered organizational structure, Deutsche Welle's path to an AI-powered content flywheel starts at the infrastructure layer, not the model layer. Topics Discussed: Why AI threatens the full media production chain, not just distribution
Most security leaders treat AI as a new threat category requiring new defenses. Rohit Parchuri, SVP and Chief Information Security Officer at Yext, pushes back hard on that. His argument: if your foundational controls are solid, AI does not require you to rebuild anything. What it does is amplify whatever you already have, gaps included, which makes the real question not "what new controls do we need?" but "how well are we actually executing on what we already built?" Rohit walks host Ben Gibert through how Yext is operationalizing this at scale: threat-modeling AI as just another system with inputs, processing, and outputs; building AI security testing directly into the existing CI/CD pipeline rather than standing it up as a separate track; investing heavily in data classification and taxonomy to solve DLP before deploying any AI tool internally; and establishing an AI Excellence Committee with cross-functional representation to run a single governance funnel across every AI request in the company. He also makes the case that the CISO who earns a seat at the AI strategy table is the one who deeply understands the business value chain, not just the threat landscape. Topics discussed: Threat-modeling AI as a system instead of a threat category