checklist

AI chatbot provider outage and model failover checklist for small teams

A practical checklist for handling AI provider outages, model errors, degraded latency, fallback routes, customer communication, and safe recovery without losing control of chatbot data or actions.

Audience: Founders, product owners, engineering owners, support leads, security owners, platform owners, and incident responders operating AI chatbots Risk: High Evidence: NIST AI RMF Core and Manage Playbook, CISA secure AI system development guidance, OWASP LLM10:2025 Unbounded Consumption, OWASP LLM06:2025 Excessive Agency, and Cybergiz chatbot operations templates

Use this checklist when a model provider, gateway, retrieval service, tool, queue, or region becomes unavailable or materially degraded. It is designed for a small team that needs a clear decision path without creating a second security incident during recovery.

The short answer: define the chatbot’s safe state before an outage, detect provider and route degradation with independent signals, switch only to an approved fallback, keep data and tenant boundaries unchanged, disable high-impact tools by default, communicate a bounded service status, and verify the original route before restoring it. A fallback is not safe merely because it returns an answer; it must preserve authorization, data handling, output validation, quotas, logging, and human escalation.

Start with the AI Tool Risk Checker and attach its result to the failover record. Use the Small Team AI Security Checklist for baseline ownership, access, data, logging, and incident controls. Pair this page with the AI chatbot pause and fallback plan template, the AI chatbot abuse monitoring and rate-limit checklist, and the AI chatbot output validation checklist.

Bottom line

An AI chatbot can enter a controlled fallback only when the team can answer these questions:

  1. Which signal proves that the primary route is degraded, and which independent signal confirms it?
  2. Which users, tenants, data classes, tools, and destinations are allowed on the fallback?
  3. Does the fallback have its own limits for input, output, concurrency, queue depth, retries, and cost?
  4. Which responses must become read-only, draft-only, delayed, or human-reviewed?
  5. Who can switch, pause, communicate, and restore the service, including outside normal hours?
  6. What evidence proves that the original route and the fallback stayed within their authorization and data boundaries?

NIST AI RMF Manage calls for post-deployment monitoring, incident response, recovery, and change management. CISA and the UK NCSC also frame secure AI as a lifecycle responsibility that includes deployment and operation. Treat failover as a controlled production change with a known safe state, not as an improvised provider swap.

This checklist is operational guidance, not legal advice, a compliance certification, or a guarantee of provider availability.

When to use this checklist

Situation or changeUse this checklist?Minimum review
Primary model provider outage or elevated errorsYesTrigger, safe state, fallback eligibility, owner, and communication.
Provider rate limit or account quota pressureYesQuota evidence, cost controls, retry behavior, and approved fallback.
Model, region, gateway, or API version changeYesRoute comparison, data boundary, regression set, and rollback.
Latency or queue degradation without full outageYesDegraded-service thresholds, queue shedding, and customer notice.
Retrieval, vector store, or source-system failureYesNo-source behavior, source scope, cache age, and human handoff.
Tool or downstream action dependency failureYesDisable or draft-only action path and downstream authorization.
Internal prototype using synthetic data onlyMaybeRecord owner and safe state before any broader use.
Static content with no AI routeUsually noUse ordinary availability and deployment controls instead.

If the team cannot name the fallback route and its data, action, and ownership limits, keep the chatbot paused or read-only until the gap is resolved.

Outage review intake form

Copy this into the incident, change, or provider review record.

FieldEntry
Incident or change ID
Detected date and timezone
Chatbot and environment
Primary route, model, provider, and region
Affected tenants, users, and endpoints
Failure signal and independent confirmation
Data classes in the route
Retrieval sources and cache behavior
Tools and downstream actions
Approved fallback route
Fallback data and tenant scope
Input, output, retry, queue, and cost limits
Switch owner and incident owner
Customer and internal communication owner
Restoration criteria and next review date

Do not paste private transcripts, API keys, provider credentials, or raw customer records into the incident channel. Record route identifiers, aggregate signals, redacted samples, and an access-controlled evidence location instead.

Failure signal inventory

Use several signals because a provider status page, a client error, or one slow request is not enough to prove the safe response.

SignalUseful dimensionsWhat it can showEvidence to retain
Provider error rateProvider, model, region, endpoint, and statusDependency or route failureAggregate counts and time window.
LatencyPercentiles, queue wait, route, and tenantDegraded service before errors risePercentile trend and alert decision.
Timeout rateClient, gateway, provider, and routeRequests that may trigger retriesTimeout class and retry outcome.
Retry velocityClient version, route, and identityAmplification or retry stormRetry counts and backoff behavior.
Queue depth and ageRoute, priority, tenant, and workerCapacity pressure and stale workBounded queue snapshot.
Provider quotaAccount, model, plan, and time windowExternal capacity exhaustionQuota signal and owner response.
Cost velocityModel, provider, route, and tenantUnexpected spend during fallbackAggregate cost and budget decision.
Retrieval healthSource, index, freshness, and authorizationMissing or stale contextSource status and no-source behavior.
Tool dependency healthTool, target, approval, and errorUnsafe or unavailable action pathAction outcome and authorization result.
Customer reportsTenant, feature, region, and timeImpact not visible in metricsRedacted report and impact scope.
Monitoring healthCollector, alert, dashboard, and timestampWhether the team can trust the viewMonitoring status and safe-state choice.

Keep provider status pages as one input, not the only evidence. A provider can be healthy while a tenant quota, gateway, region, route, or application dependency is failing.

Incident severity matrix

ClassExample patternDefault responseEscalation
AdvisoryIsolated slow request within the route’s service objectiveObserve and review the trend.Escalate if the signal persists or spreads.
DegradedRepeated timeouts, elevated errors, or queue growthLimit retries, reduce load, and use the approved degraded route.Notify the incident owner and affected support team.
HighProvider quota, tenant isolation, cost, or shared availability is at riskPause expensive or nonessential work and restrict the route.Open an incident and protect unrelated tenants.
CriticalCross-tenant data risk, credential compromise, unsafe tool action, or uncontrolled spendFail closed, disable affected actions, revoke access if needed, and preserve minimal evidence.Incident owner leads containment and recovery.

Severity should consider impact and confidence. Do not make a high-impact tool available again just because the provider responds to a health check.

Safe-state matrix

Decide what the chatbot can do when the primary route or dependency is unavailable.

CapabilityPrimary route unavailableSafe stateRestore gate
Public, low-impact answerUse approved fallback if source and policy checks passRead-only answer with uncertainty and source limitsRecent test and monitoring pass.
Customer-specific account answerDo not substitute an unscoped modelHuman handoff or no-answer responseIdentity, tenant, and source scope verified.
Retrieval-backed answerDo not invent a current answerBounded cached answer or handoff with freshness noticeCache age, authorization, and source health verified.
Draft ticket or CRM updateDo not write during an unverified dependency failureDraft-only or queue for human reviewSchema, idempotency, and owner review pass.
Message or email sendDo not send automaticallyPrepare nothing or require human approvalRecipient, content, and approval checks pass.
Permission, refund, deletion, or account actionDo not executeFail closed and route to authorized staffIndependent authorization and action test pass.
Expensive model routeDo not retry without boundCheaper bounded route or pauseCost, quality, and rate-limit guardrails pass.

Fail closed for high-impact actions. For ordinary availability pressure, a concise refusal, read-only route, or human handoff is usually safer than an unreviewed provider swap.

Fallback route approval matrix

Approve fallbacks before the incident. The fallback may be another provider, an earlier model, a cached source, a static answer set, or a human workflow, but every choice needs explicit boundaries.

Fallback optionData allowedActions allowedMain riskApproval evidence
Same provider, alternate modelTeam-defined low-risk classesRead-only or draft-onlyDifferent behavior or retention termsRoute comparison and test set.
Alternate approved providerOnly classes contractually and technically allowedRead-only by defaultNew data residency, retention, or subprocessor boundaryVendor and data review.
Earlier known-good modelSame scope as the tested versionExisting allowlist onlyQuality, refusal, or context changesRegression and output review.
Cached or static answer setApproved, freshness-bounded contentNo state-changing actionsStale or incomplete answerFreshness and source review.
Human support handoffMinimum context needed for the caseHuman-owned actions onlyQueue and disclosure pressureHandoff capacity and notice.
Full pauseNo model or tool processingNoneCustomer availability impactIncident decision and notice.

Never route private data to a fallback solely because it is online. Confirm the fallback’s data terms, access scope, retention, logging, action boundaries, and owner before use.

Data and tenant boundary checks

  • The fallback has an explicit approved data classification and tenant scope.
  • Customer, account, ticket, billing, health, employment, and identity data are not sent to an unreviewed route.
  • Retrieval filters, authorization checks, and tenant binding run independently of model output.
  • Cached responses have an owner, freshness rule, invalidation path, and access check.
  • Logs and alerts use correlation IDs without storing provider credentials or raw private prompts by default.
  • Provider, region, subprocessor, retention, and deletion assumptions are recorded for each route.
  • A fallback cannot broaden the original user’s access or change the intended data destination.
  • Cross-tenant and wrong-account tests are included before restoration.

Retry, queue, and cost controls

An outage can become an availability and cost incident when every layer retries independently. Set route-specific values and document the source of each threshold.

ControlDecision to recordFailure behavior
Retry eligibilityWhich errors may be retried and by which layer?Stop non-retryable and repeated failures.
Retry capWhat is the team-defined maximum across client, gateway, and worker?Use a safe fallback or handoff.
Backoff and jitterHow is synchronized retry avoided?Delay or shed work when the dependency is unhealthy.
ConcurrencyWhat is the per-identity, tenant, route, and provider cap?Queue within a bound, then reject or defer.
Queue depth and ageHow much work may wait?Shed low-priority work and communicate degradation.
Input and output budgetWhat size is accepted on the fallback?Reject or shorten safely without changing meaning.
Cost guardWhich spend or token velocity requires action?Disable expensive models or pause the feature.
IdempotencyWhich actions must not repeat after timeout?Require a stable operation key and verify state.

Do not use a fallback to hide a retry storm. Stop amplification first, then decide whether a bounded degraded route is safe.

Tool and action controls

Provider failure is not a reason to grant a fallback more autonomy. Keep tool actions narrow, independently authorized, and reversible.

ActionDuring outage or failoverRequired evidence before restore
Search or retrievalAllow only approved sources with bounded calls and data.Source health, scope, and freshness test.
Ticket or CRM draftDraft-only with field and tenant checks.Schema, idempotency, and human review test.
Email or messageNo automatic send unless explicitly approved.Recipient, content, and approval record.
External API readValidate endpoint, target, and parameters outside the model.Authorization and dependency test.
Permission or account changeDisabled by default.Independent authorization and human approval.
Delete, refund, or irreversible actionFail closed.Explicit approval, target check, and recovery evidence.

Customer and internal communication plan

Communicate the user-visible impact without exposing provider credentials, internal detection logic, or another tenant’s status.

AudienceMessage should includeAvoid
Internal respondersTime, scope, signal, safe state, owner, next update, and evidence locationUnredacted prompts or unverified root cause.
Support teamUser impact, approved workaround, escalation path, and status languagePromising a restoration time the team cannot support.
Affected customersFeature status, safe alternative, data-handling statement when relevant, and next updateBlaming a provider or disclosing another tenant.
Security or privacy ownerData scope, route change, tool status, and containment decisionTreating a fallback as pre-approved without evidence.
Business ownerAvailability, cost, customer, and risk trade-offHiding degraded controls to preserve a metric.

Test set and evidence record

Use a non-production route, synthetic traffic, approved masked fixtures, and clear stop conditions. Do not test failover with real customer data merely to create evidence.

Test IDDefensive scenarioExpected resultLast result
FAL-01Primary route returns provider errorsDetection confirms the issue and the defined safe state activates.
FAL-02Primary route is slow but not failingBounded timeout and retry rules prevent amplification.
FAL-03Alternate model is selectedData, tenant, output, and action gates remain unchanged.
FAL-04Fallback receives private or high-impact dataRoute rejects or hands off instead of broadening scope.
FAL-05Retrieval source is unavailableBot uses a freshness-bounded cache or safe no-answer path.
FAL-06Queue depth exceeds the recovery triggerLow-priority work is shed or deferred safely.
FAL-07Tool action is requested during failoverAction is disabled or independently authorized.
FAL-08Provider quota or cost alert firesExpensive route pauses or switches to an approved bounded option.
FAL-09Client retries after timeoutIdempotency prevents duplicate durable actions.
FAL-10Monitoring or alerting is unavailableTeam enters the documented restricted safe state.
FAL-11Customer-facing notice is neededApproved message and human escalation path are ready.
FAL-12Primary route recoversRestoration gate verifies tests, metrics, authorization, and owner sign-off.

For each run, record the route and version, test input class, data class, threshold source, aggregate result, tool decision, reviewer, timestamp, and evidence location.

Staged failover drill

StageScopeEntry gateExit signal
1. DesignOne chatbot routePrimary, fallback, safe state, owners, and data boundaries documented.Reviewers agree the fallback is bounded.
2. OfflineSynthetic requests and masked fixturesError, timeout, retry, queue, cost, and tool tests exist.Critical tests pass.
3. InternalOne trained teamMonitoring, handoff, and communications are ready.No unresolved high-risk finding.
4. Limited drillOne low-risk route or internal tenantPause and restore are observable and reversible.Recovery gates pass without cross-tenant impact.
5. Production readinessApproved customer scopeOn-call, provider, cost, and rollback owners sign off.Next review trigger is recorded.

Keep the first drill read-only or draft-only. Add automatic writes, sends, or account actions only after the independent authorization and idempotency tests pass.

Recovery and restoration plan

StepOwnerEvidenceExit gate
Confirm scope and safe stateIncident ownerSignal timeline and affected routeScope is bounded and communicated.
Stop amplificationPlatform ownerRetry, queue, and concurrency decisionPressure is within team-defined limits.
Restrict data and actionsSecurity or product ownerFallback scope and tool stateHigh-impact actions fail closed.
Activate approved fallback or pauseEngineering ownerRoute version and decision recordFallback tests and monitoring pass.
Communicate impactSupport or business ownerApproved notices and timestampsUsers have a safe next step.
Verify primary recoveryProvider and platform ownersHealth, error, latency, and dependency evidenceRecovery is stable for the defined observation window.
Run restoration testsEngineering and security ownersRegression, tenant, output, and action resultsCritical tests pass.
Restore in stagesProduct ownerCanary scope and rollback triggerNo unresolved critical finding.
Close and learnIncident ownerTimeline, root cause, actions, and next reviewOwners and due dates are recorded.

Do not restore all traffic at once if the route has changed. Restore in a controlled scope, keep the fallback or pause path available, and confirm that monitoring is healthy before closing the incident.

Decision table

FindingDecisionRequired evidence
Primary route is healthy and gates passContinue normal serviceHealth trend, tests, and owner review.
Provider is degraded but fallback is approved and boundedActivate degraded routeData, action, cost, and monitoring checks.
Fallback changes data, retention, or tenant assumptionsRestrict to safe data or pauseRoute comparison and approval.
Monitoring or quota enforcement is unavailableRestrict or pauseSafe-state decision and recovery test.
High-impact action is requested during failoverFail closed and hand offAuthorization and human review evidence.
Recovery is unclear or repeated failures persistKeep restricted and investigateIncident timeline, owner, and next decision time.

Sign-off record

RoleNameDecisionConditionsDate
Business owner
Product owner
Engineering or platform owner
Security reviewer
Privacy or legal reviewer, if needed
Support or incident owner

Approval should identify the primary and fallback routes, provider and region, data classes, tenant scope, tool state, team-defined limits, communication plan, restoration gate, unresolved risks, and next review trigger.

Action tracker

ActionRiskOwnerDue dateStatusEvidence
Document primary route and independent health signals
Approve fallback routes and data boundaries
Define safe state for each capability
Set retry, queue, concurrency, input, output, and cost limits
Disable or constrain high-impact tools during failover
Test tenant, authorization, output, and idempotency controls
Run customer communication and recovery drills

Final provider outage and failover checklist

  • Primary and fallback routes have named owners, versions, providers, regions, and review dates.
  • Independent monitoring covers errors, latency, timeouts, queues, retries, provider quota, cost, and dependency health.
  • A documented safe state exists for public answers, private data, retrieval, drafts, messages, and tool actions.
  • Fallback data, tenant, retention, source, and destination boundaries are approved.
  • IP, identity, tenant, endpoint, provider, and tool limits remain in force during failover.
  • Retry caps, backoff, queue bounds, concurrency limits, and idempotency are tested.
  • High-impact actions fail closed or require independent authorization and human approval.
  • Expensive models and provider routes have budget alerts and a pause owner.
  • Customer and internal communication uses approved, bounded language.
  • Synthetic failover tests cover provider errors, timeouts, retrieval failure, quota pressure, tool actions, and restoration.
  • Monitoring failure has its own restricted safe state.
  • Restoration is staged, observable, reversible, and backed by critical test results.
  • The AI Tool Risk Checker result is attached to the review.
  • The Small Team AI Security Checklist baseline controls are reviewed.
  • Residual risk, rollback trigger, owners, and next review date are signed off.

Metrics to track

Track aggregate trends and redacted summaries rather than collecting raw private prompts by default.

MetricWhy it mattersReview cadence
Primary and fallback error rateShows whether the switch improved availability.Near real time and daily
Latency and timeout percentilesFinds degradation before a full outage.Near real time
Retry and queue amplificationDetects load multiplication and stale work.Every incident and weekly
Fallback traffic by tenant and routeChecks fairness and scope.Daily and weekly
Data or authorization blocksTests boundary preservation.Every alert and weekly
Tool actions blocked or approvedShows action risk during recovery.Every incident
Cost and provider quota velocityPrevents denial of wallet during failover.Daily and monthly
Handoff volume and response timeMeasures whether the human fallback is usable.Weekly
Restoration test pass rateShows readiness to return to normal.Each drill and quarterly
Time to detect, restrict, communicate, and restoreMeasures incident control quality.Every incident and monthly

Evidence checked

FAQ

Should a chatbot automatically switch providers during an outage?

Only when the fallback was approved and tested for the same data, tenant, authorization, action, retention, and monitoring requirements. Otherwise use a read-only response, human handoff, or pause.

Is a provider health check enough to restore traffic?

No. Check application errors, latency, queues, retrieval, authorization, output validation, tool actions, cost, and tenant impact. Restore in stages after critical tests pass.

What is the safest fallback for private customer questions?

Usually a human handoff or a route with independently verified identity, tenant, source, and data controls. A general model or stale cache should not guess at account-specific facts.

Should cached answers be used during an outage?

Only with an explicit freshness bound, access check, source owner, invalidation path, and user-facing limitation. Cached content is not automatically current or authorized.

How do we avoid making an outage worse with retries?

Cap retries across every layer, use backoff and jitter, bound concurrency and queues, and make state-changing work idempotent. Stop amplification before activating a fallback.

What should happen to tool actions during failover?

Keep high-impact actions disabled or draft-only until independent authorization, target validation, approval, and idempotency checks pass. Fail closed for deletion, refunds, permissions, and other irreversible actions.

How much outage evidence should be stored?

Store the minimum redacted evidence needed to reconstruct the decision: route and version, aggregate signals, data scope, action state, threshold source, owner, and timestamp. Restrict raw transcripts and credentials.

Is this checklist enough for compliance?

No. It is practical operating guidance, not legal, privacy, compliance, certification, or security assurance advice. Map it to applicable contracts, laws, standards, provider terms, and risk tolerance.

Run the AI Tool Risk Checker for the primary and fallback routes, attach both results to the review, and keep customer-specific data and high-impact tools restricted until the failover, authorization, cost, and restoration tests pass.