Reported operating use · Energy and geoscience operations
How TGS reports using Scout for geoscience support
Investigate failed seismic-processing jobs and prepare support actions
TGS (as reported by AWS and TGS contributors)
Production rollout across TGS offices reported on 12 November 2026; AWS describes eight months of operating results and a phased global rollout.
A published third-party example, not a Nuvepro customer case or evidence of a partnership.
The workflow at a glance
How TGS reports investigating a failed seismic-processing job
Illustrative summary of the published account
- TGS source systems
Keep runbooks and incident knowledge current
Connectors sync runbooks, SOPs and issue history from Confluence, SharePoint and Jira while preserving each source’s access controls.
- NextGeoscientist
Ask Scout about a failed job
A geoscientist can ask from the Scout app or browser extension; the article also shows a failed job in LogViewer.
- NextScout retrieval
Retrieve only evidence the user may see
Scout retrieves entitled source material plus relevant HPC alerts and maintenance history for the investigation.
- NextScout
Return step-by-step guidance with citations
Scout provides a grounded diagnosis with linked citations so the geoscientist can check the source.
- NextGeoscientist and source systems
Take routine actions and gate higher-impact work
The article says Scout can open or update Jira tickets, maintain documentation and remediate routine failures; high-impact actions require human review. It does not enumerate the complete action catalogue or thresholds.
What the source reports
TGS and AWS describe Scout as a production citation-grounded agent for geoscientists working across global seismic-imaging and subsurface-processing operations. It retrieves permission-gated runbooks, SOPs and issue history from Confluence, SharePoint and Jira, plus relevant HPC alerts and maintenance history for a failed job. It returns cited, step-by-step guidance in AWS Quick or its browser extension, including from LogViewer context. The article says Scout can open or update Jira tickets, maintain documentation and remediate routine failures; high-impact actions require human review. TGS reports phased rollout across seven offices. AWS reports 35,000 responses over eight months, 99% citation grounding, 70% user-rated accuracy and time-to-solution discovery moving from two to three hours to under three minutes. These are AWS/TGS-reported results, not independent evaluation. The article says embedding in TGS's seismic-processing platform is planned, not live.
Reported technology: AWS Quick with permission-aware retrieval and Jira, Confluence and SharePoint connectors. Scope and outcomes are source-reported.
What the enterprise environment would need
The following is a proposed project design for an enterprise doing similar work. Published details are marked; the remaining systems, data and scenarios would be agreed for your project.
Proposed practice environment
Rehearse a permission-aware geoscience support handoff
Illustrative project design to scope with your enterprise.
- 1 · Set up the environment
Synthetic HPC events, runbooks and access rules
Create fictional failed-job logs, alerts, maintenance history, versioned procedures and a user-to-document permission matrix. Disconnect production compute and source systems.
- 2 · Preparation and agent contributions
Agent prepares a cited diagnosis and ticket draft
Retrieve only permitted evidence, preserve citations and uncertainty, and draft step-by-step guidance plus a simulated Jira ticket when the user requests a handoff.
- 3 · Independent checks
Independent technical and access review
A geoscience specialist checks job identity, time alignment, cited alerts and procedures, document freshness, permissions and the proposed action against a separately prepared incident reference.
- 4 · People decide acceptance
Operations owner accepts or returns the practice result
The accountable owner accepts only the exact independently checked diagnosis or simulated ticket. Changed evidence and unresolved technical decisions block acceptance.
A reviewable evidence package
A versioned diagnosis and simulated support handoff with citations, access-check results, exceptions and owner disposition.
If a check fails → withhold acceptance.
A stale runbook or permission mismatch blocks the draft. Repair the evidence or access map, repeat independent checks and return the exact revision to the owner.
Failure or owner rejection returns to review
- Hold
- → 2 · Revise work
- → 3 · Rerun independent checks
- → 4 · Owner review
Operating boundary. Designed, not executed. Read-only synthetic environment with simulated ticketing; no production connector, compute remediation or live Jira write.
- Synthetic job failures and permissioned operating knowledge
- Proposed: create fictional seismic-processing job logs, HPC alerts, maintenance history, Jira tickets, runbooks and SOPs with a frozen user-to-document ACL matrix. No production compute, employee accounts or live source connectors are enabled.
- Agent drafts cited diagnosis and support action
- Proposed: retrieve only documents allowed for the simulated user, then draft an evidence-linked diagnosis, step-by-step guidance and (when requested) a Jira ticket or documentation change proposal. Keep source citations and uncertainty attached to every recommendation.
- Independent geoscience and access checks
- Proposed: a specialist separate from the generating agent checks job identity, timestamps, alert/runbook citations, source freshness, user permissions and whether the suggested next step matches a separately prepared incident reference. No automatic compute remediation is permitted.
- Operations owner accepts or returns the draft
- Proposed: an accountable operations owner accepts only the exact independently checked practice diagnosis or simulated ticket. Changed evidence, an access mismatch or an unresolved technical decision blocks acceptance and triggers repair and rechecks.
A proposed simulation
An illustrative practice design derived from this work, not an executed assessment.
Rehearse a read-only, citation-grounded seismic job investigation and simulated support handoff with fictional incidents and permissions.
Review gate: The operations owner accepts only the exact practice diagnosis or ticket draft that passed independent geoscience, source-freshness, identity and permission checks. Acceptance enables no production compute control or live ticket write.
Challenge the result: A user requests help with a job whose alert timestamp conflicts with the runbook or whose source is outside the user's permission scope. Withhold the draft, repair the evidence or permission mapping, rerun independent checks and return the exact revision to the owner.
What needs checking
- Source boundary: the AWS post includes TGS contributors and describes TGS's reported production rollout. The operating metrics and quality assessments are customer/vendor-reported; no independent log, ACL or performance audit was inspected.
- Proposed: test a supported incident, stale runbook, missing citation, conflicting HPC timestamps, unknown error, user without access to one source and changed source permissions against an independently prepared incident reference. Designed, not executed.
- Proposed: change an incident log or revoke document access after the draft is checked. Invalidate the diagnosis, repair the source/permission map, repeat independent technical and ACL checks, then return the exact revision to the operations owner. Designed, not executed.
Review environment fidelity, human capability and agent performance separately. The enterprise’s accountable owners decide acceptance of the shared workflow and any operating use.
Shape this around your enterprise
Bring a similar workflow and the people responsible for it. Together, define the records and systems to use, the result your team will build or configure, and how its owners will check it. Nuvepro scopes the simulation, environment and preparation. Agree any existing-agent evaluation, implementation assistance and production responsibilities separately.
Sources and limitations
Curation review: . This is the date this source example was reviewed, separate from source publication dates.
- The published account is from AWS with TGS contributors and is not independent evaluation. No source-system telemetry, permission audit or external benchmark was inspected.
- The 99% citation-grounding and 70% user-rated accuracy figures cover 35,000 responses over eight months as reported in the article; the user-rating method and independent scoring are not established. The source does not publish an independent log, ACL or performance audit.
- The article says Scout remediates routine failures but does not specify the complete action catalogue or authorization thresholds. It says high-impact actions require human review. TGS's Seismic Imaging Processing Platform embedding is planned, so it is not described here as live.
- The separate practice design is fictional and unexecuted. No compute control, production connector or Nuvepro customer result is claimed.
- Always-on, grounded, and going agentic: How TGS built Scout on AWS QuickSource publication: 2026-11-12
More it operations examples
- Bright Machines' reported factory-data sharing workflowBright Machines, as reported with Databricks
- AI cloud incident investigation: Eon exampleEon
- AI disaster recovery evidence: Nubank exampleNubank
- AI app access review in Google Workspace: Ken Reiners recipeGoogle Workspace third-party AI app access recipe (organization unnamed)