AI Agent Testing: From Unit Tests to Production Evals

AI agent testing is not one test suite. A tool schema can be deterministic while the final task depends on retrieval, model behavior, timing, and user context. The right strategy places each check at the layer that can detect its failure without making every release slow and expensive.

At a glance

This guide is for engineers building release practices for AI workflows. It helps you match failure modes to test layers so you can leave with fast boundary checks, realistic task coverage, and a feedback loop that catches behavior changing in production.

The testing loop

Use Isolate → Simulate → Evaluate → Observe → Improve:

1. Isolate: test deterministic contracts, transformations, permissions, and state transitions.
2. Simulate: control tools, dependencies, time, and failure responses.
3. Evaluate: run representative tasks against explicit quality criteria.
4. Observe: sample real behavior and connect failures to traces and outcomes.
5. Improve: turn confirmed mechanisms into the lowest useful durable test.

Decision rule: Test a failure at the lowest layer that can see it, then add a higher-layer case when integration or model behavior changes the outcome.

A layered testing strategy keeps fast deterministic checks close to code while reserving realistic evaluation for behavior that depends on the full workflow.

1. Start with the failure and the outcome

Write what must be true and what must never happen before choosing a testing tool. “The agent should be reliable” is too broad. “The renewal assistant must cite current account state and must not create a customer commitment without approval” gives each layer something observable to check.

The AI agent evaluation guide explains how representative cases become release evidence. This article focuses on the testing layers that produce that evidence, not on one test framework or provider.

2. Isolate deterministic boundaries

Unit tests are valuable wherever the expected result is stable: schema parsing, redaction, routing, retry classification, permission checks, idempotency keys, state transitions, and cost or timeout calculations. Keep these tests fast and specific so a developer can run them after every change.

// Illustrative pseudocode: adapt the assertion style to your test runner.
test("validation errors are terminal, not retryable", () => {
  const result = classifyFailure({ kind: "tool", code: "INVALID_ARGUMENT" });

  expect(result).toEqual({
    retryable: false,
    terminalState: "needs_correction",
    owner: "workflow",
  });
});

This is intentionally illustrative pseudocode, not a drop-in library example. The contract matters: deterministic invalid input should not silently enter a retry loop. Keep assertions focused on the behavior your system owns.

3. Simulate dependencies and unhappy paths

Integration tests should make external behavior controllable. Simulate a timeout, rate limit, malformed response, stale record, duplicate event, partial result, and permission denial. Assert not only that the agent returns a result, but that it stops, retries within budget, escalates, or compensates as designed.

The AI agent reliability guide shows how failure classes map to bounded recovery. Simulation is where teams can exercise those paths cheaply and repeatedly before production supplies the case for them.

4. Evaluate realistic tasks

Scenario evaluation covers behavior that unit assertions cannot fully see: retrieved context, model choices, multi-step plans, tool sequencing, ambiguity, and user-visible quality. Build a representative set with ordinary cases, difficult cases, and high-consequence edge cases. Record expected behavior, unacceptable behavior, input class, versions, and consequence.

Use deterministic checks for facts such as required fields, citations, schema validity, and policy results. Use rubric-based or human review for groundedness, usefulness, ambiguity, and judgment. Review disagreements rather than hiding them in an average. A candidate that improves overall score while regressing a high-risk slice needs a hold or explicit disposition.

Evaluation rule: A high average cannot compensate for an unbounded failure in a workflow where the consequence is material.

5. Observe production as a test surface

Production reveals distributions and interactions that fixtures miss. Sample successful, degraded, and failed runs. Connect the task outcome to release, prompt, model, retrieval, tool, retry, latency, cost, and policy evidence. Treat user reports and operator classifications as test inputs, not merely tickets.

The AI agent observability guide describes the evidence path. Keep production feedback safe and privacy-aware: redact secrets, restrict raw content access, and preserve the smallest trace context needed to reproduce the mechanism.

Turn incidents into durable cases

When a failure is confirmed, capture the smallest fixture that would have exposed it earlier. State the mechanism, expected outcome, unacceptable behavior, and the layer that can detect recurrence. Add a higher-layer scenario when the failure depends on interactions that isolation cannot model.

For a renewal assistant, stale account data may require a retrieval freshness case, while a malformed tool argument belongs in a contract test. If both appear only as “bad answer,” the suite will grow without improving detection.

6. Gate releases with layered evidence

A release gate should combine fast checks with representative evidence and a reversible exposure plan. Write the stop condition before the candidate runs. Keep a record of test versions, dataset or fixture versions, baseline, candidate, evaluator, sampled production window, and owner.

Do not make every test blocking. A check should block when its failure matters to the intended exposure and the team has a clear response. Otherwise, route it to review with an owner and expiry rather than teaching people to ignore the gate.

Operational checklist

• Write the expected outcome and unacceptable failure first.
• Match each failure mechanism to its lowest useful test layer.
• Keep deterministic boundary tests fast and specific.
• Simulate timeouts, malformed data, duplicates, stale state, and no-response paths.
• Stratify evaluation cases by task, consequence, and input class.
• Preserve versions, traces, evaluator results, and reviewer disagreement.
• Turn confirmed production mechanisms into durable fixtures or contracts.
• Define release stop conditions, exposure, owner, and rollback path.

Next step: make testing actionable

Choose one repeated failure and write its expected outcome. Add the cheapest test that can detect it, then add a realistic scenario if the failure depends on retrieval, model behavior, timing, or user context. Run the case through the release gate and keep the production trace that proves whether the fix worked.

The ai agent testing unit tests production evals operating model

Put ai agent testing unit tests production evals into practice

A useful article should help a reader move from an idea to a safe production decision. Start by stating the user-visible outcome, then identify the inputs, the owner, the evidence, and the terminal states. Do not begin with a tool or framework. Begin with the decision the system must support and the cost of getting that decision wrong. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Define the contract

Write the input and output contract before tuning the model. Name required fields, acceptable values, missing-data behavior, and the response a user receives when the system cannot proceed. A schema validates shape, but it does not prove meaning or authorization. Follow parsing with semantic checks, policy checks, and an explicit fallback. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Separate proposal from authority

A model can classify, extract, summarize, rank, or propose an action. Deterministic code should own permissions, limits, accounting, retention, and irreversible side effects. This boundary makes the system easier to test and easier to explain after an incident. If a model suggestion is rejected, route it to a visible state instead of silently rewriting it. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Build representative cases

Create a small case set before claiming that an implementation works. Include ordinary examples, empty inputs, ambiguous requests, stale data, conflicting evidence, malformed outputs, permission boundaries, and provider failures. Label the expected user outcome and the unacceptable failure. Keep the cases versioned so a later change can be compared against the same baseline. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Validate the full path

Unit tests are valuable for parsers, transitions, policies, and retry rules, but they do not prove the workflow works end to end. Add scenario tests that exercise retrieval, model calls, tool calls, timeouts, duplicate events, approvals, and partial completion. Preserve the trace for each important scenario and check both the final result and the intermediate evidence. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Make retries safe

Every external write needs a stable identity, a bounded retry policy, and a reconciliation path. A timeout is not proof that an action failed. Record unknown outcomes explicitly, query the provider when possible, and prevent duplicate work with idempotency keys. Never let a generic retry loop decide whether a financial, permission, or communication side effect is safe. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Design human review

Human review should be specific and actionable. Show the proposed action, normalized arguments, evidence, policy version, expected consequence, and expiration. Bind approval to that exact proposal. If a material field changes, invalidate the approval. Record the reviewer, decision, reason, and result so rejection patterns become improvements rather than repeated surprises. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Instrument useful evidence

A trace should connect the original request to each meaningful stage and the final outcome. Capture versions, latency, cost, status, error category, tool arguments after redaction, and the evidence used for the decision. Avoid collecting sensitive payloads merely because they are available. Store references or structured summaries when raw content is not necessary for diagnosis. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Measure outcomes

Completion rate is not enough. Measure correctness, user acceptance, review rate, repair rate, latency, cost, retries, and the frequency of unsafe or unauthorized proposals. Segment the results by version, model, customer cohort, input class, and failure mechanism. A higher automation rate can be a regression when it creates more corrections or delays discovery of mistakes. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Roll out in stages

Use observation, shadow mode, suggestions, prepared drafts, approved execution, and only then broader automation. Give each stage an owner, cohort, success threshold, and rollback condition. Preserve the manual path until the automated path has demonstrated normal and degraded behavior. Keep in-flight work recoverable when a release is paused. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Operate failure states

Make waiting, degraded, unknown, failed, cancelled, and completed distinct states. Tell the user whether the system is waiting for data, approval, a provider, a retry window, or manual repair. Give every waiting state an expiration or escalation path. Silent queues and ambiguous success messages are operational defects, not cosmetic issues. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Review changes

Prompts, models, schemas, retrieval indexes, tools, providers, and policies are all dependencies. Version them and record their versions with each run. When behavior changes, compare a representative cohort with the previous baseline. Review quality, safety, latency, cost, and correction work together; optimizing one dimension can damage another. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Example decision record

For a meaningful release, write a short record with the goal, affected workflow, expected improvement, known risks, evidence set, rollout stage, owner, rollback command, and review date. Link it to representative traces and the test cases that protect the most important boundary. This record turns a vague confidence claim into something another engineer can inspect. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Practical checklist

Before launch, confirm that the workflow has a named owner, a clear contract, policy enforcement, safe side effects, representative tests, useful traces, and a recovery path. After launch, sample successful and failed runs, inspect corrections, update the case set, and retire metrics that do not change a decision. The goal is not maximal automation; it is a dependable user outcome that the team can explain and improve. For ai agent testing unit tests production evals, use this as a concrete review criterion: connect the decision to its evidence, owner, and recovery path.

Failure scenarios in ai agent testing unit tests production evals

The happy path is the least informative path. Rehearse empty input, malformed output, stale context, unavailable dependencies, rate limits, duplicate callbacks, and a provider timeout after an external system may already have accepted the request. For each case, state the expected user-visible result, the durable record that proves what happened, and the person or process responsible for recovery.

A concrete runbook

When an incident starts, identify the affected workflow and version, determine whether the problem is global or cohort-specific, pause new side effects if needed, and preserve a representative trace. Compare a failed run with the last known healthy run. Look for the first meaningful divergence rather than the loudest error. Record the hypothesis, reversible change, verification signal, and owner before editing production behavior.

Data and permissions

Review every input source for freshness, scope, and authorization. A technically correct answer built from data the user should not see is still a failure. Propagate tenant and user context through retrieval and tools, filter before the model sees data when possible, and avoid treating a model instruction as permission. Test cross-tenant, revoked-access, and stale-credential cases explicitly.

Cost and capacity

Measure cost per successful outcome, not merely cost per model call. A cheap answer that triggers retries, review, or correction can be more expensive than a larger reliable call. Set token, time, concurrency, and provider budgets. Define what degrades first when capacity is constrained, and ensure that load shedding does not bypass authorization or create untracked work.

Quality drift

Quality can move when user questions, documents, providers, or model behavior changes. Keep a fixed regression set and a recent production sample. Compare not only averages but severe failures, cohort movement, abstention behavior, and correction work. If the system is uncertain, a safe review or explicit limitation is often preferable to a confident unsupported result.

Recovery and replay

Replay only when the workflow is designed for it. Separate read-only analysis from side effects, preserve idempotency keys, and mark whether a run was original, retried, reconciled, or replayed. A replay should not silently send a message, charge an account, or change permissions. Provide operators with a dry-run mode and a clear confirmation step for consequential actions.

What to document

Keep the contract, state machine, dependency versions, permissions, test cases, dashboards, alerts, rollback procedure, and escalation contacts near the workflow. Documentation should answer what the system does, what it cannot do, what evidence it records, and how to stop it safely. Update it after incidents and material changes, not only during an initial launch.

A review question

Ask: if this workflow is wrong for one customer tomorrow, can the team identify the affected runs, stop further harm, explain the mechanism, recover unknown side effects, and prove the fix? If any answer is no, the next improvement belongs in observability, authorization, state management, testing, or the runbook—not in more persuasive output wording.

Closing practice

Choose one representative workflow and perform the full exercise this week. Write its contract, rehearse two degraded cases, inspect one trace, measure one user outcome, and name one rollback condition. Keep the result small enough to use. Reliable systems are built through repeated evidence-based decisions, not through a single promise that the model will behave.

Final review questions

Before treating this guidance as complete, ask what a careful operator would need at the moment the system surprises them. Can they tell which version ran, what evidence was available, which policy applied, and whether an external action is still unknown? Can they stop new work without losing recoverable state? Can they distinguish a model-quality problem from a retrieval problem, a policy problem, a provider problem, or an integration problem? These questions turn a conceptual guide into an operational one.

A useful review compares three runs: one that succeeded cleanly, one that succeeded only after a retry or human correction, and one that failed. Record the timeline, state transitions, inputs, outputs, costs, and user-visible result. The comparison often reveals that the system’s most important missing feature is not another metric but a durable boundary: a missing owner, an ambiguous terminal state, an unlogged approval, an unsafe retry, or an output that was accepted without semantic validation.

Treat the result as a living contract. When a model, prompt, tool, policy, provider, data source, or workflow state changes, update the representative cases and rerun the relevant checks. Keep the old behavior available during staged rollout when the side effects matter. Retire a control only when its replacement covers the same failure and the evidence shows that the new boundary is understood.

The best implementation is usually the smallest one that makes the decision explicit. Prefer a narrow workflow with clear permissions, predictable recovery, and honest uncertainty over a broad agent that can do more but cannot explain what it did. Measure the correction work created for people, not just the work removed from the queue. If the new path saves time while increasing hidden review or incident cost, it has not improved the user outcome.

Use this guide to produce a concrete next action. Pick one workflow, name its owner, write the contract, collect representative traces, choose one quality and one safety signal, and set a rollback condition. Review the result with engineering, security, and the people who handle exceptions. Their questions are part of the system design. A dependable production workflow is not defined by a long prompt or a high demo score; it is defined by a user-visible outcome that remains explainable, bounded, and recoverable as conditions change.

From ai agent testing unit tests production evals draft to durable practice

A production article should leave the reader with a sequence they can execute and inspect. Start with a baseline: one representative input, one expected outcome, one failure case, and one trace or record that makes the result visible. Change one meaningful variable at a time. If the result improves, record why; if it regresses, preserve the case so the regression becomes a durable test instead of a forgotten anecdote.

When choosing between alternatives, state the tradeoff plainly. Simpler systems are easier to operate but may provide less coverage. More retrieval, more tools, or more model calls may improve one quality measure while increasing latency, cost, privacy exposure, and failure surface. The right choice depends on the user outcome and the risk budget. A recommendation is only useful when the reader knows when not to use it.

Make uncertainty visible. Use abstention, escalation, citations, confidence bands, or a clear “not enough evidence” state when the workflow cannot support a reliable result. Do not convert missing evidence into confident prose. Define which uncertainties are acceptable, which require human review, and which should stop execution. Then test those boundaries with examples that are deliberately incomplete or adversarial.

Plan ownership beyond launch. Someone should own the workflow contract, dependency updates, dashboards, incident response, and review cadence. The owner should be able to answer how a release is evaluated, how a bad change is rolled back, and how users are informed during degradation. Shared ownership without a named decision-maker usually becomes no ownership when the first failure crosses team boundaries.

Include the maintenance loop in the implementation. Sample production results, inspect correction work, refresh the regression set, review costs and latency, and remove metrics that do not guide action. Feed confirmed incidents into tests, policies, or state transitions. A fix that changes only wording may hide a symptom; a durable fix changes the mechanism that allowed the failure or makes the failure safe and visible.

Finally, define completion precisely. A workflow is not complete because the demo passed or because the model produced a plausible answer. It is complete when its inputs and outputs are bounded, its authority is explicit, its evidence is traceable, its side effects are safe to retry or reconcile, its degraded states are honest, and its owner can prove the result with representative cases. Use that standard to decide what to build next and what to leave out.

Operating checklist for ai agent testing unit tests production evals

Use a representative input and a deliberately difficult input. Confirm that the system identifies the workflow and version, applies the right scope and permissions, records the evidence used, and returns a state a user can understand. If it calls a tool, verify the arguments after normalization and confirm that the tool result is associated with the same run. If it retries, verify that the retry is bounded and cannot duplicate a consequential side effect.

Review one success, one corrected result, and one failure after every material release. Compare the first meaningful divergence, not only the final text. Update the regression case, runbook, policy, or state transition that would have prevented the failure or made it safe. Then record the owner and date for the next review. This small discipline keeps the implementation connected to production reality and gives the team evidence for deciding whether to expand, simplify, or pause the workflow.

Frequently asked questions