To improve their AI recruiting search, the founders created a Slack bot that notified them of every user search. They would then manually recreate each search—up to 100 per day—to qualitatively assess the results, identify failure patterns, and methodologically fix the long tail of edge cases.
Related Insights
To ensure AI reliability, Salesforce builds environments that mimic enterprise CRM workflows, not game worlds. They use synthetic data and introduce corner cases like background noise, accents, or conflicting user requests to find and fix agent failure points before deployment, closing the "reality gap."
Countering the idea that AI sacrifices quality for speed, Honeybook's recruiting agent found four net-new, high-quality candidates the team had missed manually. The fifth candidate it found was one the team was already pursuing, validating the AI's quality and ability to augment human efforts.
Traditional recruiting tools rely on keyword searches (e.g., "fintech"). Juicebox uses LLMs to semantically understand a candidate's profile. It can identify an engineer at a payroll company as a "fintech" candidate even if the keyword is absent, surfacing a hidden talent pool that competitors can't see.
To ensure product quality, Fixer pitted its AI against 10 of its own human executive assistants on the same tasks. They refused to launch features until the AI could consistently outperform the humans on accuracy, using their service business as a direct training and validation engine.
The common mistake in building AI evals is jumping straight to writing automated tests. The correct first step is a manual process called "error analysis" or "open coding," where a product expert reviews real user interaction logs to understand what's actually going wrong. This grounds your entire evaluation process in reality.
Developers often test AI systems with well-formed, correctly spelled questions. However, real users submit vague, typo-ridden, and ambiguous prompts. Directly analyzing these raw logs is the most crucial first step to understanding how your product fails in the real world and where to focus quality improvements.
Instead of seeking a "magical system" for AI quality, the most effective starting point is a manual process called error analysis. This involves spending a few hours reading through ~100 random user interactions, taking simple notes on failures, and then categorizing those notes to identify the most common problems.
Early versions of AI-driven products often rely heavily on human intervention. The founder sold an AI solution, but in the beginning, his entire 15-person team manually processed videos behind the scenes, acting as the "AI" to deliver results to the first customer.
Instead of generic benchmarks, Superhuman tests its AI models against specific problem "dimensions" like deep search and date comprehension. It uses "canonical queries," including extreme edge cases from its CEO, to ensure high quality on tasks that matter most to demanding users.
Reviewing user interaction data is the highest ROI activity for improving an AI product. Instead of relying solely on third-party observability tools, high-performing teams build simple, custom internal applications. These tools are tailored to their specific data and workflow, removing all friction from the process of looking at and annotating traces.