Skip to content
Blueprint · Automotive

One field symptom, a decision engineers can defend.

A worked example: a symptom appears after a software update. The evidence sits across vehicle events, ECU state, software history, workshop records and engineering changes. Agents rebuild the case, engineers challenge it, and a named owner approves the next step.

Automotive
The question

A camera function drops out after the latest software update. What caused it, and which vehicles share the conditions?

Inputs

What the agents read.

Telemetry and fault codes
Vehicle events and DTCs around each occurrence
ECU configuration
Hardware revision, calibration and option codes
Software and OTA history
Builds, packages and rollout waves per vehicle
Service and warranty
Workshop findings, parts replaced and claims
Engineering records
Requirements, tests and change requests
Supplier evidence
Component validation and bench results
Run

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

  1. 01

    Rebuild the timeline

    Order the events per vehicle, each linked to the record it came from.

  2. 02

    Build the shared case

    Line up vehicle, build, configuration and service history in one view.

  3. 03

    Test competing explanations

    Keep supporting, contradictory and missing evidence with every hypothesis.

  4. 04

    Define the exposure

    Find vehicles that share the build and configuration conditions.

  5. 05

    Package the decision

    Evidence, reasoning, alternatives, exposure and validation still needed.

Agent Hive · Run 0472
A camera function drops out after the latest update. What caused it?
  1. Rebuilttimelines for 41 vehicles3.1s
  2. Joinedbuild, ECU revision and calibration1.4s
  3. Read27 workshop reports2.0s
  4. Foundchange request CR-0770.7s
  5. Kept3 hypotheses with evidence0.5s
Answer4 sources
38 of 41vehicles
Share build R7 and hardware rev. B

Leading hypothesis: a software-build dependency introduced by CR-077. Three vehicles on R7 show no symptom; replay on rev. B is still needed.

telemetryota_historyservice recordsCR-077
Ask a follow-up…
Approvals

Where people decide.

  • Needs you

    Specialist review

    Vehicle software, systems engineering, validation and supplier engineering each review the case in their own role.

  • Needs you

    Containment approval

    The quality approver accepts or rejects the containment package. Agents never trigger a campaign or an OTA rollback.

Outputs

What you get back.

Decision package
Evidence, causal reasoning, alternatives and approval
Exposure list
Vehicles sharing the build and configuration conditions
Validation plan
The evidence still needed to reproduce the condition
Audit record
Every source, review and decision, in order
Timeline

Reconstruct what happened before deciding why.

The order matters. Each event stays connected to the record it came from, so an engineer can open the evidence behind the sequence.

Fig. 1 · Vehicle 0193, in orderIllustrative data
  1. 01Software installed

    Build R7 reaches the vehicle in rollout wave 3

    OTA history
  2. 02Camera service restarts

    Two restarts within ten minutes of start-up

    Telemetry
  3. 03ECU state captured

    Hardware rev. B, calibration set 12 active

    ECU snapshot
  4. 04Fault code stored

    F-212 recorded after the second restart

    DTC log
  5. 05Workshop visit

    Blank view reproduced after a restart

    Service record
  6. 06Change found

    CR-077 changed a dependency of the camera service

    Change record
Each event links to the record it came from. The change request is the step the leading hypothesis rests on.
Fig. 2 · Events across the fleetIllustrative data
Occurrences cascade after the rollout. The critical path ties the symptom to one build and one hardware revision.
Data sources

Vehicle, field and engineering evidence, in one case.

Agent Hive keeps the relationships between vehicle identity, field records and engineering evidence visible, while every record keeps its source.

Fig. 3 · 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.
  • Vehicle and fleet

    Telemetry, DTCs, ECU configuration and software history per vehicle.

  • Field

    Service records, warranty claims and workshop notes.

  • Engineering and supplier

    Requirements, tests, change requests and supplier validation.

Competing explanations

Do not settle on an explanation too early.

The leading explanation stays open to challenge. Supporting, contradictory and missing evidence travel with every hypothesis, and engineers can reject any of them.

Software-build dependency

Leading
Supporting
Starts after build R7 and lines up with CR-077.
Contradictory
Three vehicles on R7 show no symptom.
Missing
Dependency replay on hardware rev. B.

ECU and configuration

In review
Supporting
Affected vehicles share hardware rev. B.
Contradictory
Rev. B alone does not reproduce the issue.
Missing
Full calibration and option-code comparison.

Component or environment

Open
Supporting
Workshop notes record intermittent recovery.
Contradictory
Region and climate do not match the pattern.
Missing
Supplier bench result at the same software state.
Agent workflow

Specialists build the case. The approver owns the gate.

Timeline, configuration and service agents work in parallel. Hypotheses go to specialist review, and containment stops at the quality approver.

Fig. 4 · 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.
Review roles

Field quality and engineering review the same case.

  • Field quality

    Owns the case

    Checks the recurring symptom and the exposure.

  • Vehicle software

    Reviews the build

    Build, package and rollout evidence.

  • Systems engineering

    Checks interfaces

    Configuration and dependency evidence.

  • Validation

    Defines what is missing

    The evidence needed to reproduce the condition.

  • Supplier engineering

    Reviews the component

    Bench tests and contradictory evidence.

  • Quality approver

    Owns the gate

    Accepts or rejects the containment package.

Evaluation

Graded on closed cases. Before it touches an open one.

Past investigations with a known root cause become the test set. A version that names the wrong cause, or drops contradictory evidence, does not ship.

Fig. 5 · Question set by versionIllustrative data
Evaluation results by question and version
Questionv1v2v3v4
Order the events for vehicle 0193 around the dropoutTimeline✓✓✓✓
Which build and calibration were active at the fault?Configuration~✓✓✓
Which vehicles share build R7 and hardware rev. B?Exposure✗✗✓✓
What contradicts the software-dependency hypothesis?Challenge✗~✗✓
Which change request touched the camera service?Engineering✓~✓✓
Did workshops reproduce the symptom after restart?Field~✓✓✓
What evidence is still missing to confirm the cause?Validation✗✗~✓
v4: 7 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

    The right vehicle, build and change records are present.

  • 02

    Alternatives kept

    Contradictory evidence is shown, not dropped.

  • 03

    Known cases pass

    Closed investigations reproduce their root cause.

  • 04

    Approval stays explicit

    No action leaves the workflow without the approver.

Outcomes

A shared case, sooner. Containment that holds up in review.

To a reviewed case
2 days
from two to three weeks
Vehicles in scope
1,240
instead of the whole build wave
Hypotheses kept open
3
with contradictory evidence
Eval pass, v4
7 of 7
closed-case set
Illustrative data
Deployment

Keep the investigation close to the evidence.

Identity controls, source permissions, evidence traceability and deployment boundaries stay explicit in every option.

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. 6 · 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.