Sherlocks is an AI SRE in Slack that investigates production incidents, identifies root causes, and returns evidence and next steps in Slack.
For SRE and DevOps teams running Slack-based incident management, Sherlocks adds automated investigation and root cause analysis to the tools and workflows you already use.
From alert to root cause, directly in Slack.
AI SRE for Slack-Based Incident Management
Sherlocks acts as an AI SRE agent inside your existing Slack incident workflow.
When an alert fires—or an engineer asks a question—Sherlocks:
- Automatically investigate alerts
- Start investigations from Slack
- Correlate telemetry across tools and services
- Test likely root causes against live infrastructure data
- Identify impacted services and blast radius
- Recommend remediation
- Post RCA findings directly in Slack
- Answer follow-up questions during the incident
Engineers can ask:
“Why is the API slow?”
“What caused this deployment failure?”
“Is this Kubernetes alert related to the latest release?”
Investigations can also be triggered with /investigate.
Slack-First Incident Investigation and Root Cause Analysis
When an alert fires, Sherlocks can begin investigating automatically instead of waiting for an engineer to open dashboards, query logs, inspect infrastructure, and manually reconstruct what changed.
Sherlocks combines:
- Alert context
- Slack incident context
- Infrastructure topology
- Live telemetry
- Deployment history
- Code changes
- Previous incidents
- Operational knowledge
It then builds an investigation plan, generates hypotheses, retrieves evidence, and tests likely causes.
Typical alert analysis: 2–3 minutes Complex multi-service investigation: 5–6 minutes
Automated Root Cause Analysis in Slack
Sherlocks performs automated root cause analysis in Slack rather than simply summarizing the alert.
Investigations can correlate:
- Metrics — latency, errors, CPU, memory, throughput
- Logs — application logs, errors, stack traces
- Traces — distributed traces and APM
- Infrastructure — Kubernetes events, pods, deployments, cloud state
- Databases — query performance, connections, replication
- Queues — consumer lag, depth, broker health
- Deployments — CI/CD history and release events
- Code — commits and source code when enabled
- Incident history — previous failures and resolutions
- Slack context — incident discussions, decisions, and runbooks
The resulting RCA can include:
Likely root cause Confidence level Supporting evidence Contributing factors Incident timeline Affected services and blast radius Recommended remediation Links to relevant logs, metrics, traces, deployments, and commits
The investigation is delivered back to the team in Slack.
AI Incident Response and Troubleshooting in Slack
Sherlocks turns Slack from an alert destination into an interface for AI-powered incident investigation and troubleshooting.
During an incident, Sherlocks can provide:
- Real-time investigation updates
- Automatic RCA posts
- Impacted service and endpoint context
- Correlated logs and traces
- Investigation timelines
- Recommended next steps
- Supporting dashboards and evidence
- Follow-up investigation controls
Troubleshooting examples
Sherlocks can:
- Recommend a rollback after correlating degradation with a deployment
- Identify a missing environment variable causing Kubernetes crash loops
- Trace Kafka consumer lag to an OOM-crashing consumer
- Identify a long-running database query exhausting connection pools
- Connect a production regression to a specific deployment or commit
Sherlocks focuses on automated investigation and remediation guidance rather than unrestricted production changes.
Works With Your Slack Incident Management Stack
Sherlocks works across the systems engineers already use during production incidents.
Observability
Datadog · Prometheus · New Relic · Sentry · Grafana · Elastic APM · Coralogix · ELK · Loki · CubeAPM · CloudWatch
Cloud and infrastructure
AWS · GCP · Azure · Kubernetes · ECS · Helm
Databases
MySQL · PostgreSQL · MongoDB · MongoDB Atlas · Redis · Cassandra
Queues
Kafka · RabbitMQ · Amazon SQS · Azure Service Bus
Code and CI/CD
GitHub · GitHub Actions · Jenkins · Azure Pipelines
Incident response
Slack · PagerDuty
Sherlocks handles the investigation and RCA layer while your existing monitoring, alerting, paging, and incident-management tools continue to perform their existing roles.
Cross-Tool Incident Correlation From Slack
Production incidents rarely exist inside one observability tool.
Sherlocks' Watson data agent can make on-demand, read-only queries across connected systems during an investigation.
Its Awareness Graph gives those signals additional context by mapping:
- Service dependencies
- Infrastructure topology
- Normal and abnormal behavior
- Deployment correlations
- Previous incidents and resolutions
- Slack discussions
- Runbooks
- Technical documentation
This allows Sherlocks to connect symptoms across tools and services instead of investigating each alert in isolation.
Incident Memory From Slack, Past RCAs, and Runbooks
Sherlocks preserves operational knowledge that would otherwise disappear into old Slack threads, postmortems, and individual engineers' memory.
Its incident memory can incorporate:
- Previous investigations and RCAs
- Past root causes
- Successful remediation strategies
- Slack incident conversations
- Runbook references
- Team decisions
- Deployment history
- Service behavior
When a similar incident occurs, Sherlocks can retrieve relevant historical context before investigating it.
Incident analysis retention: 1 year Default telemetry metadata retention: 90 days
Safe, Read-Only AI SRE for Production
Sherlocks' Watson data agent operates with read-only permissions across production systems.
It cannot:
- Modify infrastructure
- Modify databases or queues
- Deploy changes
- Execute destructive commands
- Access secrets or credentials
Teams can also configure:
- Safety instructions
- Service-specific operating rules
- Escalation conditions
- Approval gates for critical actions
- Team-specific deployment and traffic policies
This keeps the core AI SRE workflow focused on investigation, evidence, and recommended remediation.
AI SRE for Incident Handoffs and On-Call Context
For distributed and follow-the-sun teams, Sherlocks can investigate an incident before the next engineer takes over.
The incoming engineer can enter the Slack thread with:
- Investigation already underway
- Relevant evidence collected
- Likely causes identified
- Impacted systems mapped
- Previous incident context available
- Recommended next steps prepared
This reduces the need to reconstruct an incident from dashboards and long Slack threads at every handoff.
Sherlocks can also tag responsible owners in Slack and work alongside existing paging workflows.
Sherlocks AI SRE Performance
| Metric | Performance |
|---|---|
| Agent success rate | 35.5% → 74.8% |
| Tool-call success | 57.3% → 72% |
| p75 investigation time | 15 min → 8 min |
| Conclusive RCAs | 55% → 61% |
| Alert classification speed | 30% faster |
| Classification cost | 70% lower |
70% MTTR reduction.
Add an AI SRE to Your Slack Incident Workflow
Automatically investigate alerts, correlate production signals, identify likely root causes, and give engineers actionable incident context without leaving Slack.