Skip to content
Blueprint · Engineering

Engineering decisions, grounded in evidence.

A reference workflow: connect technical evidence, coordinate specialist agents and keep engineer validation on the whole decision path. Shown on two demanding workflows: FE analysis review and non-conformance disposition.

Engineering
The question

A machined bracket failed incoming inspection. Can we use it as is, rework it or scrap it, and what does the analysis say?

Inputs

What the agents read.

Non-conformance record
Issue details, measurements and attachments
Part and configuration
Revision, drawing, material and specification
Analysis
FE models, assumptions and result fields
Test and inspection
Reports, measurements and certificates
Prior dispositions
Earlier decisions on the same part or feature
Supplier history
Quality record and open actions for the supplier
Run

What happens, step by step. Each step is logged with its sources.

  1. 01

    Retrieve the context

    Part, supplier, quality and configuration evidence for this non-conformance.

  2. 02

    Compare related evidence

    Prior dispositions, simulation, tests and reports on the same feature.

  3. 03

    Request specialist validation

    The solver stays the validation authority. Agents set up the check; engineers run and review it.

  4. 04

    Prepare the disposition

    Use as is, rework or scrap, with sources and reasoning attached.

  5. 05

    Return the decision

    After approval, a governed update goes back to the source workflow.

Agent Hive · Run 5512
Bracket NC-5512 failed inspection. Use as is, rework or scrap?
  1. ReadNC-5512 and drawing rev D0.5s
  2. Found3 prior dispositions, same feature0.9s
  3. ReviewedFE run 88, load case LC-41.6s
  4. Requestedsolver check at measured wall0.3s
  5. Draftedrework disposition, 12 sources0.6s
Answer4 sources
1.4margin
Safety margin at measured wall, LC-4

Use as is is not supported: margin falls to 1.4 against a required 1.5. Rework to blend the edge restores 1.7, as on two earlier NCs.

NC-5512drawing rev DFE run 88NC-4120
Ask a follow-up…
Approvals

Where people decide.

  • Needs you

    Specialist validation

    Solver runs and engineering checks happen under engineer control, and their results are reviewed before use.

  • Needs you

    Disposition approval

    An authorised engineer approves, revises or escalates. Agents recommend; they do not decide.

Outputs

What you get back.

Disposition pack
Recommendation, evidence and reasoning
Analysis summary
Margins, assumptions and what was re-run
Audit record
Sources, reviewers and the approved action
Data sources

Build one engineering case around the decision.

Bring source evidence together, keep its authority, and give every agent and reviewer the same technical context.

Fig. 1 · Connected sources
Sources stay in place; Agent Hive reads them through read-only connectors and builds one shared model. Logos are trademarks of their respective owners.
  • Quality and supply

    Non-conformances, inspection results and supplier records.

  • Engineering data

    Drawings, specifications, models and analysis results.

  • Decisions

    Prior dispositions, reviews and change records.

Agent workflow

Two workflows, one operating model.

FE review and non-conformance disposition show how engineering evidence and enterprise workflow meet without blurring authority.

Fig. 2 · Workflow runIllustrative data
The run as the platform draws it: tool steps, specialist agents, joins, branches and the approval it is waiting on. Drag to pan; Ctrl or ⌘ + wheel to zoom.
  • 01

    Study completes

    Models, assumptions and result fields are available.

  • 02

    Evidence is reviewed

    Regions that need attention are connected to context.

  • 03

    Alternatives are considered

    Constraints and design implications stay visible.

  • 04

    Engineer decides

    Recommendation, report and next action are approved.

Fig. 3 · FE analysis reviewIllustrative data
The second reference workflow. When the screening result is not enough, the run waits for full solver validation; it cannot skip it.
Knowledge model

Your engineering data, connected and explorable.

Tables, procedures and KPIs across the systems engineers already use, with lineage. Agents read it; engineers can browse it.

Fig. 4 · Connected engineering dataIllustrative data
Explore the catalogue: zoom from systems to tables, procedures and KPIs, with lineage between them.
Evidence and provenance

Read the drawing. Ground it in the graph.

Dimensions, tolerances and notes are read off drawings and certificates, then linked to the part, the feature and the requirement they belong to.

Fig. 5 · Reading the drawingIllustrative data
Dimensions, tolerances and notes read off the drawing and the inspection certificate.
Fig. 6 · Grounded in the graphIllustrative data
Each extracted value is linked to its part, feature and requirement.
Fig. 7 · Provenance of one claimIllustrative data

Provenance trail. Provenance of “on-time rate = 0.874”: on-time rate = 0.874 ← query run ← metric definition ← delivery policy §4 (read right to left: the fact was derived from each step in turn). Every step: entity “on-time rate = 0.874” (the fact in the answer); wasGeneratedBy query run. activity “query run” (2026-10-21 · 09:14); wasAssociatedWith analyst agent, used orders dataset, used metric definition. agent “analyst agent” (ran the query). entity “orders dataset” (snapshot 2026-10-20); wasDerivedFrom ERP export. entity “metric definition” (v3 of 3), versions v1 first draft; v2 excludes cancelled lines; v3 late = after promised date; wasDerivedFrom delivery policy §4, wasAttributedTo data steward. entity “ERP export” (nightly load). entity “delivery policy §4” (source document). agent “data steward” (approved v3).

One claim in the disposition: the activity that produced it, the data and definitions it used, and who approved them.
Evaluation

Define what good looks like before release.

Criteria reflect engineering evidence, known cases, reviewer needs and the limits of each specialist tool.

Fig. 8 · Question set by versionIllustrative data
Evaluation results by question and version
Questionv1v2v3
Which drawing revision applies to this NC?Evidence coverage✓✓✓
Has this feature been dispositioned before?Evidence coverage~✓✓
Does the measured wall thickness keep margin?Analysis✗~✓
Which load case governs the bracket?Analysis✗✓✓
Open every source behind the recommendationTraceability✓✗✓
Rework or scrap: what changes downstream?Known case~~✗
Who must approve a use-as-is on a safety part?Approval✓✓✓
v3: 6 of 7 pass · release needs all 7
Each row is a question agreed with the team; each column a version of the workflow. The candidate ships only if it holds the release bar.
  • 01

    Evidence coverage

    Configuration, revision, requirements, analysis, tests and history.

  • 02

    Source traceability

    Every recommendation opens the exact evidence used.

  • 03

    Known-case evaluation

    Reviewed historical cases and agreed acceptance criteria.

  • 04

    Human approval

    Agents recommend. Authorised engineers approve.

Outcomes

Dispositions with their reasoning, in hours.

To a disposition pack
3 h
from four to six days
Sources per decision
12
median, all openable
Engineer-approved
100%
no action without sign-off
Eval pass, v3
6 of 7
one known case still open
Illustrative data
Deployment

Keep engineering context inside the environment you control.

Private cloud, VPC and on-premises patterns keep the same context, permissions, approval gates and audit record.

Run in your own cloud tenancy.

Private cloud or VPC

Agent Hive runs inside your private network, connected to your systems through approved paths.

  • Private tenancy and network rules
  • Model access set by agent role
  • One place for traces and evaluation

Keep platform and models in your data centre.

On-premises

The platform, the models and every trace stay on infrastructure you operate.

  • Customer-controlled boundary
  • Approved read and write connectors
  • Evidence and history kept locally
Fig. 9 · Deployment options
Deployment options compared: Managed, Dedicated cloud, On-prem, Air-gapped
Runs inQpiAI managed environmentA dedicated network in your cloudYour data centre (Docker / Compose)An isolated network
Operated byQpiAIAgreed per engagementYour teamYour team
TenancyMulti-tenant, isolatedSingle-tenantSingle-tenantSingle-tenant
Data at restManaged environmentYour cloud and regionYour networkInside the perimeter
ModelsManaged modelsApproved endpointsLocal, or via a policy gatewayLocal models only
OutboundOperated by QpiAIPrivate connectionsOne controlled egress path (hybrid)None
IdentityYour identity rulesEnterprise identitySSO and RBACLocal SSO and RBAC
UpdatesRolled out by QpiAIAgreed per engagementNew image version, pulledSigned offline packages
AuditAudit historyAudit historyLocal audit historyLocal audit history

Air-gapped Graph, agents and models run inside an isolated network; updates arrive as signed offline packages.

What runs where in each option. Select a column to move the boundary.

Run this blueprint on your data. We set up the first one with you.