Now available for Azure DevOps

Your Smart AI Product Assistant. Always On.

Requests classified. Sprints protected. Bugs auto-resolved. ScrumMind AI is the intelligence layer your didn't knew you needed — built natively inside Azure DevOps.

11
request sources
< 60s
ticket to story
6-Step
AI pipeline
0
missed requests
See It Live → Read the Docs →

How Every Request Reaches Azure DevOps

All touch points → one AI brain → one unified inbox.

Request Arrives
From Freshdesk, Teams, email, CS portal, or Azure DevOps directly
11 sources
Cleaned & Routed
Source identified. Noise removed. Standardised into a single object.
Layer 2
6-Step Intelligence
Understand → Classify → Deduplicate → Score → Simulate → Recommend
Claude Powered
Feature Modules
Request Inbox, Story Writer, Sprint Protector, Bug Resolver and more.
Azure Native
Human Approves
Story pushed to Azure. PRD on Wiki. Teams notified. PR created.
Always Human-Gated
RAG reads your context →
Azure Wiki
Backlog History
Sprint Velocity
Quarterly Goals
Product Docs
Past Decisions

End-to-End Workflow

From a raw Freshdesk ticket to a merged fix — every step ScrumMind AI takes, in exact order.

01
Request Arrives
LAYER 1 · INGESTION
A signal enters ScrumMind AI from one of 10 supported channels. Freshdesk ticket status changes to Bug or Feature Request — webhook fires instantly to the backend. A Teams message tags @ScrumMind in any channel. A CS agent submits through the CS Feature Portal in plain English. An Azure DevOps bug work item is created and detected via Service Hook. A bulk CSV of Freshdesk tickets is uploaded for batch processing.
Freshdesk WebhookTeams BotAzure Service HookCS PortalEmail via Graph APIWhatsApp / Twilio
02
Source Routing & Cleaning
LAYER 2 · NORMALISATION
The backend identifies which channel the request came from and applies source-specific parsing rules. Email signatures are stripped. Freshdesk metadata removed. Teams formatting cleaned. The result is the raw idea text, nothing else. A noise filter then runs: is this a real request, an out-of-office reply, a duplicate, or spam? Noise is discarded silently. Real requests proceed.
FastAPI Source RouterNoise FilterDeduplication CheckStandardised Object
03
Understand
AI PIPELINE · STEP 1 OF 6
Claude reads the raw request and parses plain English into a structured problem statement. Ambiguous language is resolved. The core user need is extracted regardless of how informally it was written. 'The export thing keeps breaking when I pick dates' becomes: 'PDF export fails when a date range is selected. Affects data download workflow. Reported via Freshdesk.'
Claude SonnetStructured OutputProblem Statement
04
Classify & Deduplicate
AI PIPELINE · STEPS 2 & 3
The request is classified as Bug, Feature, Improvement, Tech Debt, or Churn Risk. Simultaneously, the RAG engine searches the full Azure DevOps backlog history for semantically similar existing stories. If a near-duplicate exists, it is flagged with a link rather than creating a redundant item. This step alone eliminates 20–30% of backlog noise in most teams.
Bug / Feature / Improvement / Tech Debt / Churn Riskpgvector Semantic SearchDuplicate Detection
05
Score Against Your Goals
AI PIPELINE · STEP 4
Every request is scored across five dimensions using your actual company context: Impact (how many users affected), Urgency (time sensitivity), Revenue Risk (client tier and contract value), Goal Alignment (0–10 against your current quarterly north star), and Churn Risk (likelihood the client leaves if this is not addressed). Scores are not generic — they are calculated against your real goals stored in the system.
Impact ScoreUrgency ScoreRevenue RiskGoal Alignment 0–10Churn Risk Flag
06
Sprint Impact Simulation
AI PIPELINE · STEP 5
Before making a recommendation, the AI simulates: what happens to the current sprint if this request is added today? It reads the current sprint's committed items, story point total, team capacity from Azure DevOps Analytics, and historical velocity data. It calculates the exact impact — not a generic warning but a specific statement: 'Adding this delays AZ-312 by 2 days. Team is already at 94% capacity. Recommend dropping AZ-318 (3pts, low priority) to absorb.'
Sprint Capacity APIVelocity HistoryImpact CalculationDrop Recommendation
07
AI Recommendation
AI PIPELINE · STEP 6
The final step of the intelligence pipeline produces a clear recommendation with full reasoning: Add to Sprint, Add to Backlog, or Reject. The reasoning is specific and traceable — it references the goal alignment score, the churn risk flag, the sprint capacity simulation, and any related past decisions found in the RAG context. No black box. Every recommendation is explainable.
Add to SprintAdd to BacklogRejectFull Reasoning Logged
08
Human Reviews in Request Inbox
LAYER 4 · AZURE DEVOPS
The processed request appears in the Request Inbox — a tab inside Azure DevOps Boards. The Tech Lead or Product Owner sees every incoming request with its AI classification, scores, churn risk flag, and recommendation. Nothing reaches the sprint without a human approving it first. The AI does the thinking. The human makes the call. One click to generate the full story. One click to push it to Azure.
Azure Boards TabHuman-GatedOne-Click ApproveOne-Click Reject
09
Story Written & Pushed to Azure
OUTPUT · AZURE DEVOPS WORK ITEM
Upon approval, the AI Story Writer generates a complete Azure DevOps User Story: a properly formatted title, description in 'As a [user] I want [goal] so that [reason]' format, a minimum of three acceptance criteria, edge cases, and a story point estimate with reasoning. The story is pushed to the Azure DevOps backlog or directly into the active sprint via REST API. The CS agent or requester receives a Teams confirmation card with the story ID.
User Story FormatAcceptance CriteriaStory PointsAzure REST APITeams Confirmation
10
PRD Auto-Published to Azure Wiki
OUTPUT · DOCUMENTATION
For approved features above a certain impact threshold, the PRD Generator automatically creates a full Product Requirements Document: Problem Statement, Target Personas, Proposed Solution, Success Metrics, Scope, Out of Scope, and Technical Dependencies. The PRD is published directly to the Azure Wiki via API, auto-linked to all related stories, and a notification is sent to the product and engineering channels.
Problem StatementPersonasSuccess MetricsAzure Wiki Auto-PublishStory Links
11
Bug Resolver — Draft PR Created
AI BUG PIPELINE · AZURE REPOS
For bug-classified tickets, a separate parallel pipeline runs. Claude reads the full ticket including error messages and stack traces, identifies the affected component and likely files across all Azure Repos repositories (C# .NET backend and Angular frontend). It reads the actual code files via Azure Repos API, diagnoses the root cause, and writes a targeted fix. A new branch is created (fix/scrummind-{ticket-id}), the fix is committed, and a Draft Pull Request is opened linked to the original bug work item. The team is notified on Teams with the PR link and a summary of the fix.
Claude Code AnalysisAzure Repos APIC# .NET + AngularAuto BranchDraft PRTeams Notification
12
Velocity Mirror & System Learns
OUTPUT · CONTINUOUS IMPROVEMENT
When a sprint closes, the Velocity Mirror pipeline triggers automatically via an Azure DevOps Service Hook. It generates a retrospective report: planned vs actual story points, list of unplanned interruptions, carry-overs with root cause, and a trend analysis across the last 6 sprints. This report is published to the Azure Wiki and sent as a weekly AI briefing to leadership. Every decision — approved, rejected, estimated, carried over — is logged and feeds back into the RAG context, making every future recommendation smarter.
Sprint Close HookPlanned vs ActualTrend AnalysisWiki Auto-PostLeadership DigestRAG Feedback Loop
⬡

Human Approval at Every Output

ScrumMind AI never pushes a story, merges a PR, or publishes a PRD without a human reviewing it first. The AI does the analysis. The engineering team keeps full control.

✓ Stories: Human approves before push
✓ PRs: Human reviews before merge
✓ PRDs: Human confirms before publish

See It In Action

Every module. One intelligence layer.

Freshdesk Bug Export crashes on date range selection
Urgency 9.1/10
Churn risk
Teams Feature Add bulk reassignment to admin panel
Urgency 6.3/10
CS Portal Bug Notification emails delayed by 4+ hours
Urgency 8.7/10
Churn risk
Improvement Reporting dashboard needs date filters
Urgency 5.2/10

Description

Acceptance Criteria

    5 pts
    ⚠ Adding AZ-891 risks this sprint

    Current sprint

    AZ-3125 pts
    AZ-3183 pts
    AZ-3058 pts
    AZ-2915 pts
    Recommendation: Drop AZ-318 (3pts, low priority) to absorb AZ-891. Net sprint impact: +1 day.
    Sprint Stability
    71/100
    ✓Bug Ingested
    ✓Error Understood
    ✓Code Located
    ExportService.cs, DateRangeHelper.cs
    ✓Fix Written
    Null reference on line 247 resolved
    ✓Branch Created
    fix/scrummind-1847
    ●Draft PR Opened
    PR #394 in review
    ○Team Notified
    - if (request.DateRange == null) return;
    + if (request?.DateRange == null) return;

    PRD: PDF Export Feature

    Problem Statement

    Users cannot export reports as PDF with custom date ranges, leading to manual workarounds.

    Target Personas

    Finance managers, CS team leads.

    Proposed Solution

    Add PDF export with configurable date range in the reporting dashboard.

    Success Metrics

    Adoption > 40% of active reporters within 2 sprints.

    Out of Scope

    Batch export, custom templates.

    Auto-published to Azure Wiki · Linked to AZ-247, AZ-248
    74 / 100
    Warning2 renewal deadlines this sprint
    BlockedAZ-312 has unresolved dependency on AZ-290
    CapacityTeam capacity 15% below average (3 holidays)
    OKNo blockers detected in current queue

    See ScrumMind AI in your Azure DevOps

    Connect your Azure DevOps org. We run a live classification on your real data.

    Book a 20-minute live demo. We will connect to your real backlog, run a live classification on your Freshdesk tickets, and show you what your Request Inbox would look like today.

    • No installation required for the demo
    • Works with your existing Azure DevOps org
    • Live with real data, not a slideshow
    Email: support@scrummind.ai
    We reply within 4 hours
    ✓

    Demo request received. Expect a reply at within 4 hours.