ATLAI: AI for Atlassian Jira | Admin and User Documentation

Jira administrator

Installs the app, connects the AI provider, defines global settings, controls access, and monitors provider health, routing quality, and retention posture.

Project administrator or service owner

Enables project workspaces, selects AI-managed fields, creates queues and pools, defines override rules, and tunes escalation policy for the project.

Support agent or issue owner

Uses the issue Command Center, Ask ATLAI, Case Summary Studio, and ATLAI Activities to understand the issue, apply suggestions, produce updates, and document work.

Manager, scrum master, or delivery lead

Uses the Project Command Center, Sprint Planning Assistant, Sprint Studio, release notes, and sprint inspection to manage risk and delivery flow.

1

Connect an AI provider

Open the ATLAI admin workspace, go to Settings, select OpenAI, Azure OpenAI, Anthropic Claude, or Google Gemini, enter the model settings and API key, test the connection, then save and activate the provider.

2

Set global controls

Configure access groups, allowed projects, confidence thresholds, fallback policy, privacy masking, and retention windows before enabling automation.

3

Enable a project workspace

Choose a Jira project, enable AI triage, add project routing guidance, select AI-managed fields, and save the workspace.

4

Define routing and risk policy

Create assignment pools, routing queues, deterministic override rules, and Escalation Intelligence settings for SLA, sentiment, urgency, and customer impact.

5

Use the issue centers

When ATLAI processes an issue, agents review the issue Command Center, apply or dismiss suggestions, ask questions in Ask ATLAI, and review history in ATLAI Activities.

6

Use agile and management centers

Managers use the Project Command Center, Sprint Planning Assistant, Sprint Studio, release notes, and sprint inspection to make operational and delivery decisions.

Runtime model

ATLAI runs as an Atlassian Forge app.
Provider calls run from server-side Forge functions, not from the browser.
Issue creation events can enqueue triage jobs for enabled projects.
Issue update events capture feedback such as accepted, corrected, or overridden suggestions.
Scheduled jobs handle SQL migration and retention maintenance.

Data model

Provider API keys are stored in Forge secret storage.
Installation settings, provider metadata, project settings, and issue summary state are stored in Forge KVS.
Triage history, audit events, feedback, analytics, idempotency records, and duplicate signals are stored in Forge SQL.
Only the selected customer-managed AI provider receives the minimum required issue context.

Command center design principle

ATLAI separates classification from resolution. Triage decides what Jira fields and routes should change. Resolution centers explain operational risk, missing information, recommended human actions, customer communication, and handover needs.

ATLAI admin overview showing setup steps, configuration overview, routing accuracy, AI insights, recent activity, and provider mix. — Admin Overview is the operational landing page for setup progress, health, routing quality, quick actions, recent activity, and provider usage.

Admin Panel Overview

Overview Project Workspaces Managed Projects AI Fields Assignment Pools Routing Queues Override Rules Escalation Settings

What the Overview page tells admins

Setup progress and next recommended step.
Installation snapshot: provider health, project rollout, and routing logic status.
Configuration counts for workspaces, projects, fields, pools, queues, and rules.
System health, routing accuracy, top categories, AI insights, recent activity, and provider mix.

ATLAI Settings page showing provider connection, access permissions, confidence thresholds, fallback policy, privacy, and retention. — Settings combines provider authentication, model runtime controls, access restrictions, confidence thresholds, fallback policy, privacy controls, and retention windows.

1

Open Settings

From the ATLAI Admin Panel, select Settings, then locate Provider connection.

2

Select the provider

Choose OpenAI, Azure OpenAI, Anthropic Claude, or Google Gemini. The provider list controls which model and endpoint fields appear.

3

Choose or enter a model

Select a model from the catalog or choose the custom model option when your provider has a newer, regional, or preview model ID.

4

Enter provider-specific fields

For Azure OpenAI, enter the endpoint, deployment name, and API version. For other providers, enter the model and runtime controls.

5

Enter or rotate the API key

Paste the API key. Leave the field blank when saving non-secret settings and keeping the current stored secret.

6

Test and save

Click Test connection. After a successful test, enable Activate this provider after save and click Save provider.

Runtime setting	Purpose	Typical starting value
Timeout	Maximum wait time for provider responses before failure handling starts.	`30000` ms
Retry count	Number of retry attempts for transient provider or network failures.	`1`
Temperature	Controls output variability. Triage should remain deterministic.	`0`
Max output tokens	Caps provider response size for predictable runtime and cost control.	`800`

Access and permission

Open Settings.
Under Access & Permission, select Jira user groups allowed to use ATLAI.
Select Jira projects allowed to use issue panels, issue context, project pages, and project settings.
Leave a field as ALL only when every visible user or project should be allowed.

Confidence thresholds

Auto-apply threshold: results at or above this score can be written to Jira automatically.
Suggest-only threshold: results at or above this score are shown to users for review.
Below-threshold behavior: low confidence can be logged or routed through fallback policy.

Fallback area	What it controls	Available behavior
Provider failure	The AI provider fails, times out, or cannot be reached.	Log only, suggest only, or fallback route.
Confidence failure	The provider responds, but confidence is below configured policy.	Log only, suggest only, or fallback route.
Mapping failure	AI suggests a value that cannot be safely mapped to Jira metadata.	Log only, suggest only, or fallback route.
Invalid route	Queue, assignee, or route target is not valid for the project.	Log only, suggest only, or fallback route.

Mask emails

Replaces email-like values before issue content is sent to the selected AI provider.

Mask phone numbers

Reduces exposure of phone numbers in summaries, descriptions, and comments used as context.

Mask employee IDs

Redacts employee or account-like identifiers that match the configured masking logic.

URL and IP redaction

Optionally redacts URLs and IP addresses before external AI processing.

No-retention payload mode

Avoids retaining raw payload copies when possible and keeps only operational records, normalized output, audit-safe summaries, hashes, and tokenized duplicate signals.

Retention windows

Controls how long audit events, triage history, feedback events, and duplicate signal records stay in Forge SQL.

Managed Projects screen showing project counts, search, status filter, project inventory, and open workspace actions. — Managed Projects lets admins review every visible Jira project, identify which projects have saved or enabled ATLAI workspaces, and open each project workspace.

1

Open Managed Projects

Use this page to see all Jira projects visible to the app installation.

2

Search or filter

Search by project key or project name. Use the status filter to show all projects, not configured projects, saved workspaces, or enabled projects.

3

Review rollout state

Check whether the project is enabled, how many pools, queues, and rules exist, and whether a workspace has been saved.

4

Open workspace

Click Open workspace to configure or update that project.

Project Workspaces project basics screen showing enable AI triage, allow assignee recommendations, and routing guidance fields. — Project Basics starts the guided workspace flow: enable AI triage, allow assignment recommendations, and document project-specific routing guidance.

Configuration steps

Open a project workspace.
Enable AI triage for this project.
Enable assignee recommendations and queue assignment strategies if ATLAI should recommend ownership.
Add project description and routing guidance.
Click Save workspace, then continue to the next step.

What to write in routing guidance

Project-specific terminology and product areas.
How categories, request types, components, and queues should be interpreted.
When ambiguous tickets should remain suggestion-only.
Which issues should go to support, engineering, billing, security, or incident teams.

Core fields

ATLAI can suggest or update selected core fields only when the admin allows them.

Priority
Issue type
Request type
Components
Labels
Assignee
Routing queue

Additional Jira fields

Supported additional Jira fields appear in a searchable multi-select.

Single-select and multi-select fields
User and multi-user fields
Text and long-text fields
Number fields
Date and datetime fields
Boolean fields

Best practice

Start with a small field set, observe suggestion quality, then add more fields. Keep allowed labels aligned with your operational taxonomy, such as billing, urgent, or customer-update.

Assignment Pools screen showing a CriticalPool with selected pool members and add assignee pool action. — Assignment Pools group assignable Jira users so queues and rules can reuse clean ownership groups for round-robin, least-loaded, or fixed assignment logic.

1

Create a pool

Click Add assignee pool and provide a clear pool name, such as Support Tier 1, Billing, or Critical Response.

2

Add members

Select assignable Jira users. Pools with no members are removed on save.

3

Use the pool in queues or rules

Choose the pool from a routing queue or override rule when round-robin or least-loaded assignment should use that group.

Routing Queues screen showing queue name, assignment strategy options, selected component, default labels, and add queue action. — Routing Queues bundle destination metadata, default labels and components, and assignment strategy into reusable routing targets.

Queue fields

Queue name: reusable destination name shown in ATLAI.
Description: what belongs in this queue and when to use it.
Assignment strategy: round robin, least loaded, or fixed assignee.
Assignee pool: selected when the strategy needs a reusable group.
Default components and labels: values added when this queue is selected.

How ATLAI uses queues

Native Jira Service Management queues are views, not directly writable issue fields. ATLAI therefore stores queue intent in its own state and safely expresses routing through allowed editable Jira fields such as assignee, priority, issue type, request type, labels, and components.

Rule conditions

Keywords anywhere in the ticket
Keywords only in the issue summary
Reporter email domains
Existing Jira labels

Rule outcomes

Override priority, issue type, request type, queue, pool, or assignee.
Add labels and components.
Override selected additional Jira fields.
Record a rule explanation for future admins.

When to use rules

Use rules for business logic that must be deterministic, such as all invoice and refund tickets going to Billing, VIP domains going to a customer-success route, or incident labels forcing escalation review.

Escalation Intelligence settings showing sentiment sensitivity, SLA source fields, escalation label, priority, queue, assignee, and auto-apply checkboxes. — Escalation Intelligence is configured separately from routing so SLA, sentiment, urgency, impact, and human follow-up controls remain explicit.

1

Enable the risk assessment

Turn on Enable Escalation Intelligence risk assessment for support queues, VIP workflows, incident-prone services, or SLA-driven projects.

2

Select sentiment sensitivity

Use conservative to reduce false escalations, balanced for normal support operations, and aggressive when weak risk signals must be caught early.

3

Map SLA source fields

Select Jira or JSM SLA fields. If no SLA fields are mapped, ATLAI reports SLA risk as unknown and uses textual urgency as non-SLA evidence.

4

Define escalation outputs

Optionally set an escalation label, escalation priority, escalation queue, and escalation assignee.

5

Control auto-apply behavior

Choose whether escalation labels, priority, queue, and assignee can be auto-applied. Keep customer-facing escalation replies behind human approval when needed.

ATLAI Command Center project page showing Sprint Planning Assistant, Project Resolution Command Center, Create Sprint With AI, Create Work Items From Documents, and backlog candidates. — The project-level ATLAI Command Center brings sprint planning, project risk, document work-item creation, and backlog context into one project page.

Jira issue view showing ATLAI Command Center with AI Triage and Resolution beside Ask ATLAI. — On an issue, agents work with AI Triage & Resolution on the left and Ask ATLAI on the right for questions, case summaries, and document-assisted analysis.

What users see

Triage diagnosis and explanation.
ATLAI status: pending, auto-applied, suggested, fallback, failed, or dismissed.
Provider and model used for the latest run.
Confidence progress bar.
Suggested changes and applied changes counts.
Escalation Intelligence evidence and recommended human action.
Resolution plan generator.

Primary actions

Apply suggestion: writes suggested fields to Jira when a suggestion is available.
Re-run triage: queues a fresh AI triage run for the current issue.
Dismiss: rejects the suggestion and optionally records a reason.
Generate resolution plan: creates a structured plan for diagnosis, gaps, actions, communication, handover, and evidence.

Status	Meaning	User action
Pending	ATLAI has queued or started a triage run.	Wait for completion or refresh the issue later.
Auto-applied	Confidence met the auto-apply threshold and allowed fields were written to Jira.	Review the applied fields and escalation evidence.
Suggested	ATLAI prepared field updates for human review.	Apply, dismiss, or rerun triage.
Fallback	Fallback policy was used because confidence, mapping, provider, or route safety did not meet policy.	Review the fallback reason and correct the issue manually if needed.
Dismissed	A user rejected the suggestion.	Use Ask ATLAI or rerun triage after issue context changes.

Jira issue showing Ask ATLAI conversation beside ATLAI Activities with escalation intelligence and recommended changes. — Ask ATLAI answers issue questions from comments, routing context, escalation risk, missing information, and uploaded documents while ATLAI Activities preserves the issue evidence trail.

Ask a question

Open the Jira issue.
Open the Ask ATLAI issue context panel.
Use a quick prompt or type a custom question.
Optionally upload PDF or DOCX documents as additional context.
Click Ask ATLAI or Analyze documents.

Quick prompts

Summarize Issue
Latest Comments
Missing Info
Escalation Risk
Next Steps
Customer Questions
Triage Decision when a triage summary exists

Assistant action	What it does
Re-generate	Runs the previous prompt again.
Use as follow-up	Copies the assistant answer into the next question flow.
Create subtask	Creates a Jira subtask from the assistant answer when the project supports subtasks.
Replace Summary / Description	Updates the issue summary or description when the answer is suitable for replacement.
Add as internal note / customer reply	Adds the answer as a JSM internal note or customer reply, depending on the issue context.
Add as a comment	Adds the answer as a regular Jira comment for non-JSM issue contexts.

Case Summary

Generates a full operational summary with situation, impact, status, open questions, next steps, customer-safe update, and internal handover notes.

Handover

Creates a concise internal handover for reassignment, shift change, escalation, or manager review.

Customer Update

Drafts a customer-safe status update that excludes private operational details, provider/model internals, and audit commentary.

Timeline

Builds a chronological timeline from issue context, comments, triage history, escalation intelligence, and recorded actions.

ATLAI Activities tab showing escalation intelligence, recommended changes, and Ask ATLAI conversation output. — ATLAI Activities gives agents and reviewers a persistent view of triage changes, escalation risk signals, saved artifacts, similar-ticket signals, and audit history.

Tab	Purpose	What users review
Summary	Shows current triage and escalation result.	Escalation Intelligence, recommended changes, and already applied fields.
Case summary	Lists saved artifacts from Case Summary Studio.	Mode, provider, time, and preview of each summary.
Signals	Shows duplicate and similarity indicators.	Similar issue keys, similarity scores, and explanations.
Activity	Provides historical and audit evidence.	Triage history, confidence, explanations, agent actions, feedback, and reruns.

Metrics and diagnosis

Resolution risk: low, watch, at-risk, or critical.
Latest ATLAI confidence.
Missing information count.
Triage diagnosis, issue type, status, and assignee.

Action queue

Create investigation subtasks.
Draft customer reply.
Add internal note.
Apply recommended field updates.
Escalate with risk evidence.
Generate handover.

Resolution plans

The resolution plan generator produces sections for diagnosis, missing information, recommended actions, customer communication, internal handover, and evidence. Users can add planning instructions to focus the output on billing, implementation, incident response, customer-safe actions, or any other operational need.

Project settings ATLAI Resolution Settings page showing project command center metrics and resolution queue. — Project Command Center in project settings gives managers a resolution queue with risk, ATLAI status, confidence, signals, and last-run evidence.

Where to open it

In a software project, open the project page named ATLAI Command Center. In project settings, open ATLAI Resolution Settings for the project-level risk queue.

Project metrics

Recent issues
High risk
SLA risk
Negative sentiment
Fallbacks
Release candidates

Manager workflow

Use this center after triage has run. Start with high-risk and fallback rows, review SLA and sentiment signals, then decide which issues need escalation, customer communication, release note inclusion, or manager follow-up.

ATLAI Sprint Planning Assistant showing create sprint form, project resolution metrics, document upload, and backlog candidates. — Sprint Planning Assistant uses backlog context, sprint dates, capacity, and planning instructions to generate a realistic sprint plan for review.

1

Open the project ATLAI Command Center

Open the software project and select the ATLAI Command Center project page.

2

Select sprint inputs

Choose the Scrum board, issue capacity, start date, end date, sprint name, and planning instructions.

3

Generate the sprint plan

Click Generate sprint plan. ATLAI reviews backlog candidates and returns a sprint name, goal, selected issues, rationale, risks, assumptions, and release theme.

4

Create the sprint

Review the selected issues. Click Create sprint and move issues to create the sprint and move the selected work items.

Supported input

PDF planning documents
DOCX planning documents
Up to five documents per analysis request
Each document must be under 2.5 MB

Review before creation

ATLAI generates draft stories, bugs, or tasks.
Each draft includes summary, description, issue type, source document, and confidence.
Users can edit summary and description before creating Jira items.
Users can create work items only or create work items and a sprint.

Jira sprint action menu showing ATLAI Sprint Studio with Generate release notes and Inspect sprint plan actions. — Sprint Studio is launched from the sprint action menu and exposes both release-note generation and sprint-plan inspection for the selected sprint.

1

Open the Jira backlog or sprint view

Find the sprint and open the sprint action menu.

2

Select ATLAI Sprint Studio

The menu exposes Generate release notes and Inspect sprint plan.

3

Run the selected action

ATLAI opens a modal using the selected sprint as context and lists the sprint scope visible to Jira.

ATLAI Release Notes Studio before generation showing sprint scope and release notes audience field. — Before generation, Release Notes Studio shows the selected sprint scope and lets the user define the intended audience.

ATLAI Release Notes Studio after generation showing stakeholder-ready release notes and export actions. — After generation, users can review stakeholder-ready notes and use available copy, PDF export, or DOCX export actions.

Generate release notes

Open the sprint action menu.
Select Generate release notes.
Enter the audience, such as customers, stakeholders, and release managers.
Click Generate release notes.
Review the generated title, sections, release-manager notes, and issue count.

Use the output

Copy the notes into Slack, Teams, Confluence, email, or a customer update.
Export to PDF when the action helper is available.
Export to DOCX when the action helper is available.
Validate release-manager notes before sending externally.

ATLAI Sprint Scope Inspector before inspection showing selected sprint and sprint scope. — Sprint Scope Inspector starts with the selected sprint and visible sprint scope so delivery leads can run an inspection from current Jira data.

ATLAI Sprint Scope Inspector after inspection showing readiness, scope observations, risks, and recommended manager actions. — The inspection result gives a readiness posture, scope observations, risks, and concrete recommended manager actions.

Output	Meaning
Readiness	Overall sprint posture: ready, watch, or at-risk.
Summary	Plain-language delivery assessment for the sprint.
Scope observations	Notes about size, clarity, status mix, and coherence.
Risks	Delivery, quality, missing goal, unclear scope, blocked, or incomplete work risks.
Recommended manager actions	Concrete next steps for scrum masters, project managers, or delivery leads.

What ATLAI records

Triage jobs and triage history.
Audit events for apply, dismiss, rerun, assistant actions, and case summaries.
Feedback events for accepted, corrected, overridden, helpful, and not-helpful outcomes.
Duplicate signal records and analytics facts.

Dashboard metrics

Tickets processed and auto-triaged.
Auto-applied, suggestion-only, low-confidence, and failure counts.
Average confidence and latency.
Override, correction, and accepted-suggestion rates.
Estimated minutes and hours saved.

Customer-managed providers

ATLAI uses only the provider credentials configured by the customer. Supported providers are OpenAI, Azure OpenAI, Anthropic Claude, and Google Gemini.

Server-side calls only

Provider calls are made from Forge backend functions. API keys are not exposed to client-side UI code.

Strict output validation

AI triage output is parsed through a strict schema, normalized, and validated against Jira project metadata before mutation.

Fail-closed routing

Invalid options, invalid assignees, invalid queues, low confidence, and provider failures follow admin-configured fallback policy.

Ask ATLAI says no provider is connected

Open Settings, configure a provider, test the connection, save it, and make sure it is active.

The issue panel says no triage result exists

Confirm the project is enabled, the project is allowed by access settings, an active provider exists, and the issue was created after configuration. You can also click Queue triage run or Re-run triage.

Suggestions are not auto-applied

Check the confidence score, auto-apply threshold, selected AI fields, mapping validity, fallback policy, and Jira permissions for field updates and assignment.

Document analysis fails

Upload only PDF or DOCX files, keep each file under 2.5 MB, and confirm an active provider is available.

Sprint creation is unavailable

Use a Scrum board that supports sprints. Some boards are visible but do not expose sprint creation through the Jira Agile API.

Escalation risk looks incomplete

Map SLA fields in Escalation Intelligence. Without mapped SLA fields, ATLAI can still assess sentiment and textual urgency, but SLA risk may show as unknown.

JSM queues are not directly writable

ATLAI preserves queue intent and routes through editable Jira fields such as assignee, priority, issue type, request type, labels, and components.

Subtasks depend on project capability

Ask ATLAI can create subtasks only when the current project has a creatable subtask issue type.

Document limits apply

Issue and sprint document analysis supports PDF and DOCX files under 2.5 MB each.

Provider quality affects output

All AI output depends on the selected provider, model, runtime settings, project guidance, and the quality of issue context.

Fallback does not replace human review

Fallback routes are safety behavior, not proof that the issue is fully resolved or correctly classified.

Access settings can hide modules

If a user or project is excluded by ATLAI access settings, issue panels and project modules may be unavailable.