NuePrism Documentation
NuePrism is an Execution Verdict System for software delivery. It connects to Jira and GitHub, analyzes execution signals, and converts them into evidence-backed verdicts, focus areas, and recommended recovery actions — with a confidence score tied to how complete the underlying evidence actually is, not to how the diagnosis sounds.
Popular
Verdicts & Reports
What NuePrism can tell you about a sprint or release.
Connect Jira & GitHub
Set up your evidence sources in minutes.
Reading a confidence badge
What High, Medium, and Low actually mean.
Executive Report
Generate a shareable sprint summary.
Roles & permissions
Manage who sees what.
Jira-only vs Jira+GitHub
Know exactly what each mode covers.
Related reading
Background on how NuePrism's execution-verdict approach relates to adjacent categories.
Execution Intelligence vs Execution Verdicts
How NuePrism relates to the execution intelligence category.
NuePrism for Jira Sprint Risk Detection
Sprint health, spillover, and predictability from Jira alone.
NuePrism for Jira + GitHub Delivery Intelligence
Adding PR, review, and CI evidence to planning data.
Is NuePrism just another dashboard?
How verdicts differ from metrics and trend charts.
NuePrism and process mining
A related idea, applied where no single process model exists.
Who uses NuePrism?
NuePrism is designed for software teams and delivery leaders who need earlier visibility into execution risk — not a generic project-management tool.
- Engineering managers tracking sprint and release execution
- Product owners monitoring delivery confidence
- Scrum masters / delivery leads managing flow and blockers
- Engineering leaders reviewing team-level execution health
- Portfolio leaders looking for propagated delivery risk across teams and products
Common use cases
| Use case | What NuePrism helps with |
|---|---|
| Mid-sprint risk review | Identify where delivery risk is building before sprint close |
| Release readiness review | See whether planning status matches engineering evidence |
| Review bottleneck diagnosis | Understand whether review flow is slowing delivery |
| Jira hygiene review | Identify when stale records reduce confidence |
| Dependency review | Surface blocked or unowned work affecting delivery |
| Action follow-through | Track whether recommended actions were actually completed |
| Executive readout | Generate a concise health, verdict, and action summary |
What NuePrism does not do
NuePrism is not a replacement for Jira, GitHub, or your existing planning process. NuePrism does not:
- Replace your system of record
- Read or store source code content
- Enforce one specific software methodology
- Treat every metric movement as a problem
- Claim AI involvement where the connected evidence doesn't support it
- Replace human delivery judgment
Connect Jira & GitHub
NuePrism reads directly from your existing tools — nothing to migrate, no new system of record.
Jira-only mode (built on Atlassian Forge, internally "Forge v1")
If you're not ready to connect GitHub, NuePrism still surfaces meaningful signal from Jira alone — flow, intent stability, dependency ownership, and Jira hygiene. PR-level, merge-readiness, and CI-based verdicts require the GitHub connector and become available once it's added.
Jira-only vs Jira + GitHub — what each mode covers
| Capability | Jira-only mode | Jira + GitHub |
|---|---|---|
| Flow and ageing signals | Yes | Yes |
| WIP and owner-load signals | Yes | Yes |
| Scope and intent stability | Yes | Yes |
| Dependency / blocker visibility | Yes | Yes |
| Jira hygiene / stale record detection | Yes | Yes |
| PR review flow | No | Yes |
| Merge readiness | No | Yes |
| CI / quality-gate evidence | No | Yes |
| Jira–GitHub traceability | No | Yes |
| AI-assisted PR risk patterns | Limited | Conditional on available attribution evidence |
Overview
The Overview screen shows your team's current health at a glance: a headline status, supporting KPIs, and the primary verdict driving that status. Click View Evidence on any verdict to see exactly what it's based on.
Illustrative example — layout and figures shown for orientation, not a live screenshot.
What is a verdict?
A verdict is a structured diagnosis of an execution risk pattern. It combines multiple signals into a clear explanation of what may be affecting delivery, how confident NuePrism is in that explanation, and what the team can do next — not a raw metric or a status flag.
Verdict catalog
Each verdict is scoped to a specific condition and comes with a confidence level — see Reading a confidence badge for what that means.
| Verdict | What it tells you | Requires | Status |
|---|---|---|---|
| Execution Reality Gap | Planning-tool progress looks ahead of what your engineering data actually shows. | Jira + GitHub | Available |
| Review Capacity Bottleneck | Code review is the reason work is waiting, not effort or complexity. | Jira + GitHub | Available |
| CI Failure Drag | Failing or flaky checks are the reason merge-ready work isn't shipping. | Jira + GitHub | Available |
| Traceability Gap | Jira and GitHub aren't linked well enough to trust the reported status. | Jira + GitHub | Available |
| AI Rework Inflation | AI-assisted changes are creating more review and correction work than they're saving. | Jira + GitHub | Conditional |
| Flow Continuity Breakdown | Active work is ageing, bouncing, or stalling faster than it's completing. | Jira-only mode | Available |
| Focus Fragmentation | Individuals are carrying more concurrent work than they can realistically progress. | Jira-only mode | Available |
| Dependency Drag | An unresolved, unowned dependency is the reason work isn't moving. | Jira-only mode | Available |
| System of Record Staleness | Jira itself isn't being kept current enough to support a reliable diagnosis. | Jira-only mode | Private preview |
Available — generally available for connected teams. Conditional — available only when supporting evidence exists (e.g. AI attribution data). Private preview — not yet generally available.
Example verdict workflow
Here's what a verdict looks like end to end, using Review Capacity Bottleneck as an example:
Verdict: Review Capacity Bottleneck Confidence: Medium-High Focus area: Review & Feedback Health
What NuePrism is seeing:
- Multiple work items are waiting for review
- Review-ready work is ageing beyond this team's normal turnaround
- Reviewer load is concentrated on one or two people
- Release-critical work is affected
Recommended actions:
- Assign backup reviewers for release-critical items
- Create a daily review queue checkpoint
- Split oversized review items where possible
- Recheck merge readiness after review movement improves
Focus Areas & Actions
Every verdict maps to one or more focus areas — themed groupings of recommended actions, each with an owner, a due date, and an expected impact. Multiple verdicts can share a focus area when they call for the same kind of fix from the same kind of person.
| Focus area | What it helps improve | Example actions |
|---|---|---|
| Work Definition Health | Unclear or changing work | Assign acceptance owner, clarify active work, re-estimate changed items |
| Scope-Capacity Balance | Overcommitment and scope pressure | Freeze new starts, de-scope lower-value work, rebalance capacity |
| Flow Stability | Stalled or ageing work | Assign active owner, split stale work, unblock priority items |
| Workload Balance | Too much parallel work | Cap WIP, finish before starting, redistribute overloaded work |
| Review & Feedback Health | Delayed review, QA, or clarification | Add reviewer rotation, assign backup reviewers, create feedback checkpoint |
| Release Quality | Work not ready to merge or release | Triage failed checks, require validation evidence, clear release blockers |
| Record Accuracy | Weak or stale evidence | Link work items, update stale records, fix missing evidence |
| Dependency & Integration Health | Blockers and cross-team waits | Assign dependency owner, escalate blocker, re-sequence dependent work |
Conditional focus area — AI Delivery Governance: used when AI-assisted work is detected and the connected evidence supports AI-specific review.
Action lifecycle
Recommended actions move through a simple lifecycle:
If an action isn't completed, NuePrism can keep the related risk visible until the team resolves or dismisses it.
Why Rechecked matters: this is the step that closes the loop between a verdict and reality. Once an action is marked complete, NuePrism doesn't just trust that it worked — it re-evaluates the same verdict against the next evidence window (e.g. the following sprint) and records whether the risk pattern actually improved. Over time this is what lets a verdict be confirmed correct against actual outcomes, rather than staying a one-time guess. See Outcomes for how this shows up after a sprint closes.
Outcomes
Track projected delivery for the current sprint, or review actual results and spillover after it closes. Once a sprint completes, NuePrism shows whether its mid-sprint verdicts held up against what actually happened.
Executive Report
Click Export from any Overview screen to generate a one-page, printable summary — health, KPIs, top verdicts, and recommended actions — suitable for a leadership readout. Choose Save as PDF from the preview to download it.
Illustrative example — layout and figures shown for orientation, not a live screenshot.
Reading a confidence badge
Every verdict shows a confidence level: High, Medium-High, Medium, Low, or Signal Only. This reflects how strong and complete the underlying evidence is — not how severe the problem looks. A Low-confidence verdict is still worth reading; it means treat it as a lead worth checking, not a settled conclusion.
Illustrative example — actual thresholds vary by verdict and are being refined through the validation process described under Validation status.
Evidence coverage
Evidence coverage reflects how completely your connected tools support a reliable diagnosis — for example, how many work items have a linked, merged pull request. Low coverage caps confidence across every verdict for that team, not just the one about traceability, since every diagnosis depends on the same underlying data being trustworthy.
Worked example — Review Capacity Bottleneck for a 12-item sprint:
- 10 of 12 items have a linked, merged PR (83% traceability coverage)
- Reviewer-assignment data is present for all 10 linked items
- Review-wait timestamps are complete and consistent for 9 of the 10
With coverage in this range and no conflicting signals, the verdict is capped at Medium-High rather than High — the two untraceable items mean NuePrism can't fully rule out that some of the apparent review delay is actually a Jira hygiene problem instead. If coverage dropped below roughly half of items, the same underlying pattern would surface only as Signal Only: worth a look, not yet a diagnosis.
Roles & permissions
NuePrism supports role-based access aligned to how delivery teams are actually structured:
- Team member — views their team's Overview, Execution, Actions, and Outcomes.
- Team lead — additionally assigns actions and adjusts team-level configuration.
- Admin — manages data connectors, org-wide settings, and access.
Setup prerequisites
Before setup, you'll need:
- Jira Cloud admin or project-admin access
- Access to the Jira projects you want NuePrism to analyze
- GitHub organization/repository access for Jira + GitHub mode
- Permission to approve the requested app scopes
- A mapped team/project structure
Recommended before your first review:
- Confirm team membership
- Confirm sprint/release field mapping
- Confirm Done/closed status mapping
- Confirm blocker/dependency fields
- Confirm Jira–GitHub linking convention
Security & data handling
NuePrism reads execution metadata from your connected tools — it does not require source code access and does not store credentials. The table below lists exactly what is and isn't accessed.
| Data type | Accessed? | Notes |
|---|---|---|
| Jira issue metadata | Yes | Status, assignee, timestamps, labels, blockers, sprint/release fields |
| Jira comments | Optional / configurable | Only if enabled for evidence context |
| GitHub PR metadata | Yes | PR status, review state, checks, timestamps, linked issue references |
| Source code content | No | NuePrism does not require code-content access |
| CI check status | Yes, via GitHub metadata | Used for quality-gate visibility |
| User identity metadata | Yes | Used for ownership and permissions |
| Secrets / credentials | No | Not required for verdict generation |
Platform & hosting
NuePrism's Jira-only mode runs as an Atlassian Forge app. Forge apps execute on Atlassian-managed infrastructure, inherit Atlassian's data-residency options, and cannot make arbitrary outbound network calls — meaning your Jira data doesn't leave Atlassian's platform boundary for the Jira-only mode. Jira + GitHub mode additionally reads GitHub data through scoped, read-only API access that you grant explicitly per organization.
What we don't have published yet
NuePrism is pre-GA, and some standard enterprise security answers aren't formalized yet: SOC 2 status, a subprocessor list, and self-serve retention/deletion controls are on the roadmap rather than live today. If your evaluation needs any of these — including SSO/SAML support, specific data-residency commitments, or a security questionnaire — contact support and we'll work through it directly rather than pointing you at a console that doesn't exist yet.
AI-assisted work
When AI-assisted work is identified in a connected repository, NuePrism can help teams review validation coverage, rework rate, and quality-risk patterns specific to that work — surfaced at a confidence level appropriate to how clearly it can be attributed. NuePrism does not claim AI attribution where the underlying data doesn't support it.
Future integrations
Available now: Jira Cloud, GitHub. On the roadmap: Azure DevOps, GitLab, Bitbucket, survey/feedback tools, and documentation systems.
Validation status
NuePrism's verdict model is being validated through historical delivery data, customer pilots, and evidence-based review. The system is designed to improve over time as more outcomes are observed, rather than presenting a fixed, unchanging rule set.
NuePrism is designed to be evidence-calibrated, not purely AI-generated: the system separates evidence analysis from AI-assisted explanation, so the diagnosis itself doesn't depend on a language model's judgment.
Frequently asked questions
Does NuePrism replace Jira or GitHub?
No. NuePrism reads from your existing tools and adds a diagnosis layer on top — it isn't a system of record.
Do we need GitHub connected to use NuePrism?
No. Jira-only mode works with Jira alone. Connecting GitHub unlocks additional verdicts — see the capability matrix.
Does NuePrism read our source code?
No. NuePrism reads execution metadata, not code content.
How is a verdict different from a dashboard metric?
A metric shows a number. A verdict combines multiple signals into an explanation, with a confidence level and recommended next steps — see What is a verdict?
What happens if evidence is incomplete?
Confidence is capped accordingly, rather than presenting a verdict with more certainty than the evidence supports.
Can we use NuePrism for AI-assisted teams?
Yes, if the connected evidence supports AI attribution. See AI-assisted work.
Is there an API?
Not yet — see API overview for current status.
Glossary
- Verdict — a structured diagnosis of an execution risk pattern, with a confidence level and recommended actions.
- Focus area — a themed grouping of recommended actions that share a common remediation owner and approach.
- Confidence — how strong and complete the evidence behind a verdict is, shown as High, Medium-High, Medium, Low, or Signal Only.
- Evidence pack — which data sources are connected for a team (e.g. Jira-only, or Jira + GitHub). This determines which verdicts can be produced at all.
- Evidence coverage — how complete and trustworthy the data is within whatever evidence pack is connected (e.g. what fraction of items have a linked, merged PR). This determines the confidence ceiling for verdicts that are already possible. A team can have the full Jira + GitHub evidence pack and still get Low-confidence verdicts if coverage within that pack is poor.
Known limitations
- Verdicts requiring GitHub data are unavailable in Jira-only mode — see Jira-only mode.
- NuePrism does not currently evaluate deployment or production events — its evidence sources cover planning through code review, CI, and testing.
- The API is planned but not yet available — see API overview.
API Planned
A NuePrism API is planned, to let you pull verdicts, actions, and evidence-coverage data into your own dashboards or workflow tools. Not yet available — reach out to your NuePrism admin or contact support to discuss early access.
Verdicts endpoint Planned
Endpoint reference, authentication, and webhook events will be published here once the API reaches private preview.
Last updated: July 2026