Series: AI-Native SRE

Table of Contents

• The Problem Every On-Call Engineer Knows
• What FRIDAY Does
• Architecture Overview
• Key Design Decisions
• The Tool-Use Loop: How FRIDAY Reasons
• The Training System: Pre-Built Knowledge
• Handling Edge Cases
• Results
• Lessons Learned
• Try It Yourself

The Problem Every On-Call Engineer Knows

It's 2:47 AM. Your phone buzzes and it's a P1 alert. You open your laptop, bleary-eyed, and begin the familiar ritual:

1. Open PagerDuty → read the alert title
2. Open Datadog → search for the service, find the error spike
3. Open GitHub → check if someone deployed something
4. Cross-reference timestamps between all three tools
5. Form a hypothesis
6. Drill deeper — check affected tenants, error paths, queue depths
7. Write up findings for the team

This process takes 15–45 minutes for an experienced engineer. For a junior on-call? Sometimes hours. And the cognitive overhead of context-switching between 3-4 tools while sleep-deprived leads to missed signals, false conclusions, and longer outages.

I asked myself:

So I built one. It's been running in production for months, investigating real incidents on a platform serving 30+ million end users across multiple AWS regions. We call it FRIDAY.

What FRIDAY Does

When a PagerDuty alert fires, FRIDAY:

1. Receives the webhook in real-time via API Gateway
2. Locks the target region from the alert metadata (never investigates the wrong region)
3. Checks GitHub first — finds what changed before the alert (deployments, config changes, PRs)
4. Queries Datadog — error rates, affected tenants, application exceptions, queue depths
5. Synthesizes findings — correlates code changes with observability signals
6. Delivers a structured report to Microsoft Teams as an Adaptive Card

The entire investigation takes under 2 minutes. The on-call engineer wakes up to a complete analysis instead of a raw alert.

Architecture Overview

┌──────────────┐     ┌────────────────┐     ┌─────────────────────┐
│  PagerDuty   │────▶│  API Gateway   │────▶│  Lambda (Sync)      │
│  Webhook     │     │  (Validate)    │     │  Parse + Self-Invoke│
└──────────────┘     └────────────────┘     └─────────┬───────────┘
                                                       │ Async
                                                       ▼
                                            ┌─────────────────────┐
                                            │  Lambda (Async)      │
                                            │  Investigation Agent │
                                            │                      │
                                            │  ┌────────────────┐ │
                                            │  │ Amazon Bedrock  │ │
                                            │  │ Claude Opus     │ │
                                            │  │ (Tool-Use Loop) │ │
                                            │  └───────┬────────┘ │
                                            │          │          │
                                            │    ┌─────┼─────┐   │
                                            │    ▼     ▼     ▼   │
                                            │ GitHub Datadog  S3  │
                                            └─────────┬───────────┘
                                                      │
                                                      ▼
                                            ┌─────────────────────┐
                                            │  Microsoft Teams    │
                                            │  (Adaptive Card)    │
                                            └─────────────────────┘

Key Design Decisions

1. Two-Lambda Architecture (Sync + Async)

API Gateway has a 30-second hard timeout. A thorough AI investigation takes 60–180 seconds. The solution: the sync Lambda validates the webhook, parses the alert, and immediately self-invokes asynchronously returning 200 OK to PagerDuty within 2 seconds.

# Sync handler: validate, parse, self-invoke, return immediately
lambda_client.invoke(
    FunctionName=context.function_name,
    InvocationType="Event",  # Fire and forget
    Payload=json.dumps({
        "_async_investigate": True,
        "alert_payload": alert_payload,
    }),
)
return {"statusCode": 200, "body": "Investigation started"}

The async Lambda runs the full investigation without timeout pressure.

2. GitHub First, Datadog Second

This is counterintuitive. Most engineers and most AI systems jump straight to observability data when an alert fires. But in my experience, 80%+ of acute incidents are caused by a preceding change: a deployment, a config update, a replica count change, a memory limit modification.

FRIDAY is instructed to check GitHub before touching Datadog:

MANDATORY FIRST STEP — GitHub (Step 0):
Before touching Datadog, you MUST run these calls in parallel:
1. github_search_repos — find the repo for the alerted service
2. github_list_commits — find commits in the 2 hours before 
   the alert fired

A deployment or config change is the most likely root cause.

3. Region Lock — Preventing Wrong-Region Investigation

Our platform spans multiple AWS regions. A naive agent querying "all 5xx errors" would mix signals from healthy and unhealthy regions, producing confused analysis.

FRIDAY's first action is always to lock a target region from the alert metadata:

🌍 Region: Description — resolved from alert hostname

Every subsequent Datadog query includes:
kube_cluster_name:region-az-* (scoped to affected region only)

This eliminated an entire class of false-positive findings where the AI would cite errors from an unrelated region.

4. Structured Output Contract

FRIDAY's output isn't freeform text. It follows a strict section contract that the Teams integration parses into visual containers:

## EXECUTIVE SUMMARY
[2-3 sentences — what happened, who's affected, what changed]

## KEY FINDINGS
[Bulleted evidence from GitHub + Datadog]

## WHAT CHANGED
[Specific commit/PR with timestamp and author]

## ERROR BREAKDOWN
[Service-by-service error counts with affected tenants]

## ROOT CAUSE
[Confirmed / Suspected / Unknown — with evidence chain]

## CUSTOMER IMPACT
[Affected tenants, operations, scope]

## RECOMMENDED ACTIONS
[Specific next steps for the on-call engineer]

The on-call engineer can glance at the Teams card and immediately know: what happened, who's affected, what likely caused it, and what to do next without reading a wall of text.

The Tool-Use Loop: How FRIDAY Reasons

FRIDAY uses Claude's tool-use capability in a multi-round loop. The AI doesn't execute a fixed script — it reasons about each alert independently, deciding which tools to call based on what it's learned so far.

for round_num in range(MAX_TOOL_ROUNDS):  # Max 25 rounds
    response = bedrock_client.converse(
        modelId="anthropic.claude-opus",
        messages=messages,
        toolConfig={"tools": TOOL_DEFINITIONS},
    )

    if stop_reason == "tool_use":
        # Execute tools, append results, continue reasoning
        for tool_call in content_blocks:
            result = execute_tool(
                tool_call["name"], 
                tool_call["input"]
            )
            tool_results.append(result)
        messages.append(tool_results)

    elif stop_reason == "end_turn":
        # AI has concluded — extract findings
        return extract_final_report(content_blocks)

Available Tools

The AI typically uses 8–15 tool calls per investigation, batching parallel calls when possible to minimize round-trip time.

The Training System: Pre-Built Knowledge

A cold investigation — where the AI knows nothing about your infrastructure — is slow and imprecise. FRIDAY includes a deterministic training mode that pre-builds architectural knowledge:

def train():
    """
    Deterministic training:
    ~13 targeted API calls, then one Bedrock synthesis call.

    Collects: cluster-service maps, HAProxy backends, 
    chronic error baselines, recent planned work.
    """
    # Phase 1: Targeted data collection (no AI — pure API calls)
    collected = {}
    for key, tool_name, tool_input in TRAINING_CALLS:
        collected[key] = execute_tool(tool_name, tool_input)

    # Phase 2: Single AI synthesis call
    knowledge_doc = synthesize_knowledge(collected)

    # Phase 3: Save to S3 — injected into system prompt
    save_to_s3(knowledge_doc)

The knowledge document contains:

• Cluster → Service map — What runs where
• Chronic error baselines — Background noise to ignore (not incidents)
• Recent planned work — Deployments and migrations that explain expected errors
• Backend inventory — Every backend serving traffic

Handling Edge Cases

Planned Work vs. Real Incidents

One of the hardest problems: distinguishing planned maintenance from real outages. During a Kubernetes cluster migration, you expect 5xx errors as traffic drains. FRIDAY handles this through:

1. Knowledge injection — Training mode captures recent PRs tagged as planned work
2. Real-time PR correlation — During investigation, it reads PR bodies for keywords like "decommission", "drain", "planned"
3. Explicit classification — If a 5xx spike coincides with a merged "failover" PR, FRIDAY reports:

"This alert coincides with planned cluster decommission. Errors are expected during traffic drain. No incident action required."

Force-Completion Under Round Limits

What happens when an investigation is complex and approaching the 25-round tool limit? FRIDAY has a graceful degradation mechanism:

if rounds_remaining <= 3:
    user_content.append({
        "text": (
            "STOP CALLING TOOLS. Write your FINAL report "
            "NOW using all data collected so far. Mark "
            "uncertain findings as 'Suspected' rather "
            "than skipping them."
        )
    })

This ensures every investigation produces a report — even if incomplete — rather than timing out silently.

Deduplication

PagerDuty retries webhooks. FRIDAY handles this at two levels:

• Webhook-level — In-memory cache of webhook IDs (survives Lambda warm starts)
• Incident-level — S3 marker files prevent re-investigating the same incident

Results

After running in production for several months:

Lessons Learned

1. Prompt Engineering IS Architecture

The system prompt is the most important file in the codebase. It's not instructions — it's the agent's operating manual. Ours is ~5,000 words covering:

• Environment topology (region mappings, cluster roles, service dependencies)
• Investigation methodology (step-by-step procedures)
• Critical rules (what NOT to do — as important as what to do)
• Output format contract

Invest in your prompt like you invest in your architecture docs.

2. "GitHub First" Was the Single Biggest Win

Before this rule, the AI would spend 10+ rounds querying Datadog, building elaborate theories about traffic patterns — then discover a config change was merged 5 minutes before the alert. Now it finds the root cause in rounds 1-2 for ~80% of change-induced incidents.

3. You Need Guardrails, Not Just Capabilities

FRIDAY is explicitly told it does NOT take remediation actions. It investigates, analyzes, and reports. A human validates and acts. This is not a limitation — it's a design choice that builds trust. When on-call engineers trust the AI's analysis, they act on it faster.

4. Separate Investigation from Notification

The two-Lambda pattern (sync for webhook receipt, async for investigation) is essential. Don't let API Gateway timeouts dictate your AI agent's investigation depth.

What's Next

We're extending this pattern to autonomous security remediation — an agent that ingests vulnerability findings, generates IaC fixes, deploys through GitOps, verifies no impact, and requests human approval before proceeding. Same tool-use architecture, different domain.

Try It Yourself

The pattern is reproducible with:

• Amazon Bedrock (Claude Opus or Sonnet for cost-sensitive use)
• Any webhook source (PagerDuty, Opsgenie, Datadog)
• Any observability platform with an API (Datadog, Grafana, New Relic)
• Any source control (GitHub, GitLab)
• Any chat platform (Teams, Slack)

The hard part isn't the code — it's the system prompt. That's where your SRE expertise lives. The AI is the execution engine; your knowledge of your infrastructure is what makes it useful.

I'm Vinothsingh Elumalai, a Platform Engineering leader building AI-native operations at enterprise scale. I lead the Platform team for a global IAM/SSO platform serving 30M+ users. Currently exploring how agentic AI transforms SRE from reactive firefighting to autonomous, closed-loop operations.

This is Part 1 of my AI-Native SRE series. Part 2 will cover JARVIS — an autonomous vulnerability remediation agent that fixes security findings through GitOps with human approval gates.

Connect on LinkedIn