PRDFeature Bet 2

Resolution Replay Loop

Verified Knowledge + Fixlets + Deflection Quality Score

Close the knowledge-to-resolution feedback loop by converting solved work into verified knowledge, "Fixlets" (one-click fix actions), and a Deflection Quality Score that optimizes good deflection---not just volume.

Target Segments

  • Primary: IT (service desk + IT ops) at 2k-10k employeesWhere deflection and automation ROI is measurable
  • Secondary: HR/Finance knowledge domainsWith strict permission boundaries (policy/benefits, procurement FAQs)

Why Now

  • "Shift-left with knowledge + automation" and "measure deflection quality" are explicit trends in the category
  • Atomicwork already does auto-generated knowledge and suggests verified answers from resolved requests
  • Competitors have pieces but not closed loops with quality measurement
  • Buyers want to prove deflection actually solved problems, not just avoided tickets

Objectives

1

Increase Good Deflection

Not just deflection volume---focus on quality

2

Convert Resolutions into Verified Knowledge

Rapidly create knowledge packages from resolved work

3

Turn Knowledge into Action (Fixlets)

Actionable fixes runnable within Slack/Teams

4

Maintain Governance

Human approval, audit trails, permissions-aware content

User Personas

Primary

Employee Requester

An answer that actually fixes the problem

Primary

Agent

Fewer repeat tickets + less KB writing overhead

Primary

Service Desk Manager

Improved deflection and lower cost per ticket

Primary

IT Admin / Knowledge Owner

Governance, approvals, versioning, permission boundaries

Functional Requirements

Core capabilities required to deliver on the stated objectives.

1P0
FR-001

Deflection outcome capture

Resolved / Still need help recorded and linked to session

2P0
FR-002

Deflection Quality Score

System labels deflection as good/bad based on defined rules (reopen, feedback, ticket creation)

3P0
FR-003

Auto-draft KB article from resolved ticket

Draft created within 5 minutes of ticket resolution

4P0
FR-004

Review + approval workflow

Draft cannot publish without approver; decision logged

5P0
FR-005

KB surfaced in Slack/Teams with citations

Answer includes source references; respects permissions

6P0
FR-006

Fixlet attachment to KB

KB entry can include a Fixlet pointer (action template ID)

7P0
FR-007

"Run Fixlet" in Slack/Teams

When user confirms, action runs and outcome is posted + logged

8P0
FR-008

Analytics dashboard for deflection + quality

Dashboard shows deflection rate + quality split by intent

9P1
FR-101

Auto-detect "unhelpful themes"

System proposes KB/Fixlet candidates based on repeat patterns

10P1
FR-102

Versioning + rollback for KB and Fixlets

Admin can roll back to prior version in 1 click

11P1
FR-103

KB freshness signals + expiry

Articles prompt review after N days or upon repeated failures

12P1
FR-104

Multi-source citation rendering

Answer shows which source each step came from (SharePoint/Confluence/etc.)

13P2
FR-201

Autonomous publish for low-risk drafts

Only allowed for whitelisted intents with high acceptance rate

14P2
FR-202

A/B test answer variants

System can route % traffic to variant and compare quality

15P2
FR-203

Multi-language KB generation

Supports N languages with review workflow

Success Metrics

MetricBaselineTargetChange
DeflectionTBD (pilot baseline)+10-20 percentage points
MTTRTBD (pilot baseline)10-25% reduction
SLA ComplianceTBD (pilot baseline)+5-10 percentage points
Cost per TicketTBD (pilot baseline)15-30% reduction
Good Deflection RatePilot baseline+10-20 percentage points
Bad Deflection RatePilot baselineNot worse than baseline (goal <=3%)

Rollout Plan

1-3

Alpha (Internal)

Run KB draft generation on internal tickets; evaluate acceptance rate and hallucination risk.

4-9

Pilot

Top 5 intents (password resets, VPN access, MFA reset, DL membership, common SaaS access). KB drafts + review workflow, Slack/Teams answers with citations, outcome capture, 1-2 Fixlets.

10-15

Beta

Expand to 15-20 intents. Add KB freshness scoring + auto-review prompts. Expand Fixlet library.

16+

GA

Stable good deflection lift without bad deflection regression. Fixlet execution success >=95%. KB freshness workflow adopted.

Want the full artifact?

Email me to request a PDF of this complete PRD with all requirements and specifications.

Request PDF