AI tools for your work
Limited betaChoose an AI tool by starting with the task you want to complete. usedby connects work areas and tasks to tool categories, with public examples to examine.
Your role is not listed? Describe the task you want to do.
Train and evaluate models
These results show public use in this category. They do not establish use by your profession or suitability for this exact task. Your requirements have not been evaluated.
Category, industry and company size
Company size describes the employer. It does not establish access for a solo practitioner.
LLM observability & evals
158 companies, 6 tools
About these numbers
Counts cover companies recorded on usedby. They are not market shares. All four proof levels count. Missing size data is excluded only from size-based views. Your requirements have not been evaluated.
Companies with known size: 60%. Vendor companies excluded from this category: 0.
LangSmith
46 companies
1 confirmed43 observed2 signals
Harvey runs offline evaluations of its agentic search system in LangSmith, using OpenTelemetry traces from the OpenAI Agent SDK. The evaluations track metrics such as hallucination, tool recall, retrieval recall, formatting, and answer quality.
Original source text
Centralized evaluation infrastructure that we built using the OpenAI Agent SDK's OpenTelemetry traces enables us to run offline evaluations in LangSmith. We track metrics like hallucination, tool recall, retrieval recall, formatting, and answer quality.
Harvey · Read the source 
Braintrust
42 companies
13 confirmed28 observed1 signal
Box's product managers use Braintrust as the system of record for the datasets, experiments and metrics behind evals of the Box Agent.
Original source text
Box's path to launch was determined by datasets. Before opening the agent to internal dogfooding, the team built a set of roughly eighty questions covering core functionality and decided there would be no internal release until the agent scored above 80%.
Box · Read the source 
Arize AI
32 companies
9 confirmed23 observed
Booking.com uses Arize to trace, monitor and evaluate its GenAI agents and traditional ML models (such as ranking and fraud detection) in production and during development.
Original source text
In one case, an Arize monitor fired when latency for a specific agent suddenly spiked several standard deviations above its baseline.
Booking.com · Read the source 
Comet Opik
17 companies
10 confirmed7 observed
Zencoder uses Opik to trace and evaluate the outputs of its LLM-based coding agents, debug unexpected model behavior, and compare prompts and models across experiments.
Original source text
Opik was built to scale and supports the team's high volume of experimentation, as they are currently running thousands of experiments daily.
Zencoder · Read the source 
Fiddler AI
14 companies
3 confirmed11 observed
