Ontology & Semantic Knowledge Graphs Strategic Positioning | TotalEnergies x Forvis Mazars

Executive Deck

How should TSF and SousLeSens be positioned in the current OneTech technology landscape?

Objective: Position TSF and SousLeSens as the governed meaning layer that feeds OneTech platforms, starting with LifeX as the first production focal point.

Forvis Mazars × TotalEnergies
Contents

Executive Deck

1.Context
2.Executive summary
3.Strategic message
4.Problem
5.From data to decisions
6.AI readiness
7.Three approaches
8.KG vs SKG
9.Ontology
10.Standards
11.Where SKG applies
12.TotalEnergies foundation
13.LifeX proof point
14.Concrete LifeX model
15.Maturity path
16.Decision support questions
17.LifeX value loops
18.Target architecture
19.Operating model
20.Decision framework
21.Recommended actions
22.Conclusion
Context

Ontology & Semantic Knowledge Graphs Strategic Positioning

TotalEnergies and OneTech already have a solid semantic foundation through TSF and SousLeSens (SLS). The strategic issue is adoption: leadership needs to see why governed meaning matters, where it creates measurable value, and how it fits the OneTech landscape across platforms, operations, data products, and AI.

The work builds on a first mission conducted from November 2025 to January 2026, the Cross-FPSO Semantic Preparation and Ontology Vetting Program, where Forvis Mazars assembled the team with NCOR, the National Center for Ontological Research. That mission validated TSF's alignment with BFO (Basic Formal Ontology, ISO/IEC 21838-2:2021) through a formal audit of the TE Business Objects ontology, and delivered a machine learning prototype that matched 23,000 Pazflor functional locations to Dalia equivalents to bootstrap cross-FPSO job card transfer for life extension planning. This second mission turns that validated technical foundation into a positioning narrative for TSF and SousLeSens within OneTech.

  • 5 coordinated workstreams and 5 client-facing decks.
  • 12 internal stakeholder interview profiles and 3 Voice of Peers interview profiles.
  • 9 benchmarked tool profiles across semantic control, platform, graph, catalog, and AI layers.
  • 25 strategic intelligence organization profiles backed by 26 evidence notes.
  • Management assets include the Executive Deck, one-pager, FAQ, use-case showcase, decision framework, RFI/RFP support, benchmark matrix, and evidence registers.

This deck is the leadership synthesis. Use it as the front-door deck for COMEX and OneTech leadership, then use the supporting assets when the audience asks for proof, comparisons, decision criteria, or implementation detail.

Executive summary

Semantic capability becomes visible when it acts as a control layer.

What decision should leadership make now to turn semantic capability into operational adoption?

The next step is to position TSF and SousLeSens as the governed source of meaning that platforms can consume, then prove it in a production workflow leadership can see, question, and reuse.

Problem

Data exists, but equipment, degradation, cost, and risk are interpreted differently across systems and teams.

Foundation

TSF and SousLeSens already provide the standards-based base: Basic Formal Ontology (BFO), W3C semantic web standards (RDF, OWL, SPARQL), and TotalEnergies domain models.

Proof point

LifeX turns the topic into a management question: which FPSO evidence changes risk, schedule, OPEX, or CAPEX?

Recommendation

Push the current LifeX effort from structured pilot to governed production, and connect TSF outputs to the tools teams already use.

The executive decision

Mandate LifeX as the first operational proof point for TSF, then expose the resulting concepts and mappings to Cognite, Fabric, Collibra, Power BI, Alteryx, and future AI agents.

Strategic message

Keep meaning governed once, then expose it across platforms.

Who owns the definitions, rules, and procedures these systems use to interpret reality?

TotalEnergies can use Cognite, Fabric, Collibra, Power BI, Alteryx, and AI agents. The point is to keep the definitions, rules, and procedures outside any vendor tooling, so each platform consumes the same meaning.

Core message

  • AI needs clear definitions to produce reliable answers.
  • Vendor platforms can connect data, but TotalEnergies should keep ownership of its industrial knowledge.
  • LifeX is the right proof point: valuable FPSO assets, scattered evidence, and direct CAPEX/OPEX decisions.
  • The TotalEnergies Semantic Framework (TSF), supported by SousLeSens (open-source ontology management and graph exploration platform financed by TotalEnergies), keeps meaning stable while tools and applications change.

Positioning logic

  • Semantic layers are moving from architecture topic to AI readiness topic.
  • Platform vendors are embedding their own semantic layers.
  • Open standards matter because definitions must travel across tools.
  • TotalEnergies should keep ownership of its definitions and let platforms consume them.
Problem

Shared meaning is the bottleneck behind connected data.

What prevents connected data from answering at decision speed?

SAP, inspection reports, dashboards, and data platforms can all point to the same asset while describing it through different labels, rules, and assumptions.

What exists today

SAP

Work orders, equipment, notifications.

+

Documents

Inspection reports, engineering records, PDFs, Data Management Plans.

Dashboards

Power BI, local analyses, heatmaps.

+

Platforms

Cognite, Fabric, Collibra, OneData.

What needs to become visible

  • Stable definitions for equipment, degradation, work, cost, and risk and semantization of industry standards.
  • Ontology-driven pattern comparison across FPSOs using reusable logic.
  • A trace from field evidence to the management decision it supports.
  • Platform consumption of the shared vocabulary and logic across dashboards, applications, and vendors.

Executive message

Many signals, no shared map of reality. That is why decisions stay slow, costly, and hard to audit.

From data to decisions

Data reaches dashboards before it reaches decision logic.

Where exactly does the current data stack stop creating value for management decisions?

Reports can show what happened. Management still needs the link between evidence, business objects, rules, risk, and the action to take.

1. Data

Captured signals

Logs, sensors, records, files, reports.

2. Information

Stored data

Databases, data lakes, warehouses, dashboards.

3. Knowledge

Context and meaning

Relationships, business objects, constraints.

4. Decisions

Actionable rationale

Traceable, explainable, auditable decisions.

14%

of organizations are confident their data is secured and governed.

Source: Juan Sequeda summary of Gartner Data & Analytics Summit 2026, citing the Gartner opening keynote.

44%

implemented a semantic layer in 2025; 48% plan to by 2027.

Source: Juan Sequeda summary of Gartner Data & Analytics Summit 2026, citing the Gartner opening keynote.

40%

more AI rework and remediation by 2027 for enterprises lacking a universal semantic layer.

Source: Gartner, "Future-Proof AI Systems and AI Agents With Universal Semantic Layer Development", G00843460, 3 February 2026.

AI readiness

AI needs concepts, rules, procedures, and data together.

What must be governed before AI agents can be trusted in operational workflows?

A semantic layer gives people, applications, and AI agents a human-readable and machine-readable control layer for shared meaning.

Concepts

What things mean

Equipment, inspection, work order, degradation, risk, cost, asset, role, document.

Norms

What rules apply

Standards, constraints, access rights, validation rules, risk thresholds, governance requirements.

Procedures

What steps to take

Inspection workflows, escalation paths, maintenance planning, investment update cycles, reporting processes.

Why this matters for AI agents

Shared concepts, norms, and procedures keep agents aligned on the same task, reduce duplicated logic, and make outputs easier to govern and audit.

Three approaches

RAG, property graphs, and SKGs solve different problems.

When should TotalEnergies use RAG, a property graph, or a semantic knowledge graph?

Each technology answers a different business need and belongs at a different layer.

RAG / vector search

Retrieve likely relevant content

  • Fast exploration across documents.
  • Useful for summaries and assistance.
  • Probabilistic and context-dependent.

Best question: "What information looks relevant?"

Property graph

Connect operational entities

  • Assets, sensors, documents, time series.
  • Strong for navigation and application workflows.
  • Meaning remains local to the model.

Best question: "What is connected to what?"

Semantic KG

Govern meaning and reasoning

  • Formal concepts, rules, standards, inference.
  • Reusable across systems and assets.
  • Supports auditable decisions.

Best question: "What does it mean and what should we do?"

KG vs SKG

Semantic solutions cover different architectural layers.

Which architectural layer does each approach actually address?

Fabric, Cognite, OntoBricks, and TSF / SousLeSens address different layers of the semantic value chain.

Analytics layer

Microsoft Fabric

Strong for analytics, reporting, Power BI integration, Copilot usage, and contextualization inside the Microsoft ecosystem.

  • Best question: what should be reported?
  • Limit: RDF support sits outside the native design center.
Operational graph

Cognite CDF

Strong for industrial data contextualization, OT/IT integration, digital twins, asset-centric use cases, and operational workflows.

  • Best question: what is connected operationally?
  • Limit: formal semantic governance sits outside the native core.
Ontology + lakehouse

Databricks OntoBricks

Promising bridge between lakehouse architectures and W3C-style semantic standards such as RDF, OWL, SHACL, SPARQL, and R2RML.

  • Best question: can lakehouse data be reasoned over semantically?
  • Limit: industrial maturity remains the proof point.
Semantic control layer

TSF / SousLeSens

Strong for formalizing business meaning, standards alignment, semantic governance, interoperability, and long-term reuse across platforms.

  • Best question: what does it mean and which rule applies?
  • Role: feed platforms with governed meaning.

Executive positioning

TSF clarifies the business meaning that Fabric, Cognite, Databricks, and other platforms should consume, reuse, and expose in decisions.

Ontology

LLMs predict language. Ontologies define meaning.

Why do AI and data platforms still need ontology when they can already process language and data?

The practical value is explicit meaning that can be reused, checked, and connected to decisions.

With labels only

  • Systems store local labels instead of shared concepts.
  • AI predicts plausible answers from fragments.
  • Business rules remain buried in documents, spreadsheets, prompts, and code.
  • Teams struggle to compare decisions across FPSOs or projects.

With ontology

  • Entities, processes, roles, qualities, and impacts are defined.
  • Rules and constraints become machine-readable.
  • Facts can be checked and inferred.
  • Decision rationale becomes traceable from evidence to action.

Simple formula

LLMs help interpret and generate language. Ontologies define the stable meaning that makes their outputs trustworthy in a business process.

Standards

BFO matters because industrial decisions involve things, processes, roles, and impacts over time.

Why is a standards-based ontology materially different from another internal vocabulary?

BFO is an ISO top-level ontology used to support information integration, retrieval, and analysis across domains. It gives TotalEnergies a stable foundation for modeling reality through more than labels.

Stability

Meaning survives projects

Definitions stay stable as platforms, dashboards, and local projects change.

Governance

Meaning is owned

Definitions, roles, rules, and changes can be reviewed, versioned, and audited.

Reuse

Meaning travels

A model created for one FPSO or process can become a reusable pattern for others.

Why this is practical

The BFO 7 Buckets methodology turns operational common sense into encoded structure: Who (material entity), What (process), Where (site), When (temporal region), How-it-is (qualities), How-we-know (information), and Why (realizables). In LifeX, realizables make the impact logic explicit by connecting inspection or maintenance events to downstream north star outcomes: safety, cost, production, and risk.

Where SKG applies

Use SKG first on mission-critical processes that need memory, rules, and auditability.

Where inside the organization does shared meaning create the most value?

The opportunity is broad, but the first focus should be processes with high complexity, cross-team coordination, and direct impact on safety, cost, production, or risk.

Use SKG when

  • The process is mission-critical for operations or leadership.
  • The decision crosses systems, teams, or affiliates.
  • OPEX, CAPEX, safety, production, or emissions are at stake.
  • Definitions must hold across years and assets.
  • Standards, thresholds, or audit trails matter.
  • Reuse is worth more than another one-off dashboard.

Use a lighter approach when

  • The use case is local to one team and one system.
  • The question is simple reporting or visualization.
  • Inference or cross-asset comparison adds little value.
  • The expected reuse horizon is short.
  • Formal modeling costs outweigh the decision risk.

LifeX passes the test

LifeX brings together high-value FPSO assets, long operating history, recurring degradation patterns, cross-team coordination, and decisions with direct CAPEX/OPEX impact.

TotalEnergies foundation

TotalEnergies has the assets. LifeX gives them a place to prove value.

What has TotalEnergies already built, and what would turn it into broad business adoption?

TSF, SousLeSens, standards work, data domains, and industrial pilots are already in place. The next move is to make one use case visible in operations.

Standards and methodology

TSF developed since 2021, W3C standards, BFO alignment, semantic governance methodology.

Tooling

SousLeSens for ontology management, graph visualization, mapping, validation, and querying.

Domains and models

Facility design, facility operations, field monitoring, industrial projects, and cross-domain modeling.

Industrial use cases

LifeX programs, Dalia, Pazflor, Girassol, CLOV, OneData, Primavera, EDMS, and digital twin enablement.

Ecosystem

ISO, CFIHOS, PCA, NCOR, IOGP, DEXPI, FL3DMS, universities, and industry partners.

Next focus

Turn the semantic portfolio into a LifeX decision workflow with visible owners, users, and outputs.

LifeX proof point

LifeX is where semantic value becomes measurable.

Why should LifeX be the first production proof point for TSF and SousLeSens?

LifeX combines the conditions where semantic modeling matters most: fragmented data, high-value assets, recurring degradation patterns, and direct CAPEX/OPEX decisions.

High asset value

FPSO life extension has direct capital exposure and executive visibility.

High complexity

Equipment, inspections, documents, work orders, notifications, costs, planning, and dashboards are fragmented.

High cost of failure

Missed degradation or wrong replacement decisions affect OPEX, CAPEX, safety, and production continuity.

LifeX data landscape

SAP, SQL, CSV, Excel, JCDB, New Prodom, dashboards, inspection reports, technical documents, work orders, notifications, costs, inspections, planning.

Business need

A unified, coherent, queryable representation of FPSO systems that enables cross-source analysis, decision support, and reusable analogues from one FPSO to another.

Concrete LifeX model

For LifeX, semantics connects inspection evidence to investment decisions.

What does the LifeX semantic graph actually connect in operational terms?

The semantic chain links field observations, equipment concepts, work history, documentation, cost, risk, and decision rationale.

Semantic chain

Inspection observationPhoto, report, PDF, field note
Equipment / functional locationTSF concept aligned to BFO / IOF
Work order / notificationSAP, JCDB, operational systems
Technical documentationNew Prodom, engineering records
Cost, risk, scheduleCAPEX / OPEX / integrity impact

What TSF / SousLeSens contributes

TBOX: OWL ontology aligned to BFO, IOF, and TotalEnergies business objects. ABOX: RDF graphs built from LifeX data sources. Composite graph: SAP, inspection, documentation, dashboards, and cost/schedule context.

What management gets

A way to navigate from evidence to analogues, risks, standards, and investment rationale with stable meaning from week to week.

Decision output

The semantic chain creates a reusable path from field evidence to equipment analogues, risk assessment, cost impact, and investment rationale.

Maturity path

The LifeX semantic work has moved from validation to governed multi-FPSO deployment.

What is the practical path from pilot work to a reusable LifeX decision model?

Levels 1 and 2 have validated the approach on Dalia and Pazflor. Level 3 is the current focus: a governed, reusable pipeline across FPSOs.

Level 1 (complete)

Dalia document PoC

Manual extraction, ontology modeling, and application ontology validation.

Level 2 (complete)

Dalia to Pazflor pattern validation

Compare FPSO patterns, generate a rulebook, run QA/QC, and validate ontologies.

Level 3 (current)

Governed multi-FPSO pipeline

Create the LifeX semantic data product: extend ingestion to Pazflor, CLOV, and Girassol, then package mappings, rules, QA/QC, and outputs for reuse.

Level 5 (target)

AI agents

AI agents use semantic historian outputs to reason, plan, and assist engineering workflows.

Decision support questions

The questions LifeX management should answer by navigating the graph.

Which management questions prove that the semantic graph is connected to business decisions?

In ontology work, these are competency questions. For LifeX leadership, they are the decision support questions that define what TSF / SousLeSens governs, what remains in source platforms, and how decision queries reach the evidence.

Question 1

Which FPSO assets, equipment families, or tags are degrading faster than expected?

Navigate from inspection observations to equipment classes, tags, and functional locations. The ontology makes similarity patterns visible across Dalia, Pazflor, Girassol, and CLOV, even when tags differ.

Question 2

What is the impact on the LifeX investment plan, maintenance schedule, and risk profile?

Link degradation events to work orders, notifications, costs, planning data, and applicable standards or rules.

Question 3

What decision reduces risk while avoiding over-investment, and can we justify it?

Trace the recommendation from evidence to semantic class, rule, cost impact, operational trade-off, and Investment-ready rationale.

Boundary and query mechanism

The SKG stores governed semantics: concepts, identifiers, mappings, rules, curated graph data, and evidence links. Operational data can stay in source platforms such as Cognite, Collibra, Fabric, SAP, and document stores. Federated queries work through the semantic layer: it translates decision terms into source-specific identifiers and fields, reaches the relevant APIs or query endpoints, and returns results with provenance.

LifeX value loops

LifeX value comes from three operational loops.

What changes in day-to-day operations when LifeX processes are modeled semantically?

The business value appears when recurring work loops become explainable, reusable, and traceable.

Loop 1

Inspection → investment decision

Before: observations sit in documents and dashboards.

With SKG: observations become semantic events linked to equipment and impact.

Value: faster consolidation and traceable investment rationale.

Loop 2

Maintenance → operational impact

Before: work orders are tracked but weakly connected to cost, risk, or performance.

With SKG: maintenance is modeled as a process with conditions and outcomes.

Value: impact-driven operations beyond activity tracking.

Loop 3

Failure → reusable analogues

Before: failure analysis stays local and repeats across assets.

With SKG: causes, context, and decisions become reusable patterns.

Value: organizational learning from one FPSO to the next.

Value discipline

Value should be measured through reduced consolidation effort, faster decision preparation, reusable analogues, and stronger auditability from evidence to decision.

Target architecture

Keep the LifeX data stack running. Add a semantic control layer beside it.

How should LifeX connect operational systems and semantic governance while keeping each layer in its role?

The traditional stack moves data toward applications and reporting. TSF / SousLeSens governs the meaning, mappings, rules, and query mediation that make those outputs decision-ready.

Classic LifeX data stack
LifeX data

Sources

Work orders, inspection records, documents, costs, schedules, risks.

Operational applications

Execution

SAP, JCDB, New Prodom, engineering tools, inspection workflows.

Data platform

Storage and access

Cognite, Fabric, OneData, lakehouse services, APIs, query endpoints.

BI / agents

Consumption

Power BI, Alteryx, assistants, agents, Investment-ready reports.

Decisions

Business action

Risk, schedule, OPEX, CAPEX, integrity, production trade-offs.

Semantic control layer
LifeX data

Evidence links

Source identifiers, document references, asset and tag links.

Domain knowledge capture

Expert meaning

Inspection logic, degradation patterns, standards, rules.

Semantics

TSF / SousLeSens

Concepts, mappings, graph data, rules, provenance, KGQuery.

Decision query

Meaning to action

Federated query mediation from business terms to source evidence.

Decisions

Reusable rationale

Auditable answers with links back to evidence, rules, and owners.

The semantic control layer sits beside the traditional stack: it governs shared meaning and feeds mappings, APIs, query mediation, and provenance back into operational applications, data platforms, dashboards, agents, and decisions.

Operating model

Own meaning centrally. Let execution stay close to the teams.

Who owns the semantic map, who contributes to it, and who consumes it?

TSF should protect common definitions and rules. LifeX and platform teams should keep the practical knowledge, mappings, and user workflows alive.

Lead

Semantic authority

TSF owns canonical concepts, standards alignment, rules, versioning, and semantic governance.

Operate

Domain teams

LifeX, inspection, maintenance, asset managers, and engineers contribute and validate operational knowledge.

Consume

Platform teams

Cognite, Fabric, Collibra, Power BI, Alteryx, and AI systems consume semantic services and mappings.

Governance loop

New term, rule, workflow, or mapping → domain validation → semantic review → versioned publication → platform consumption → usage and drift monitoring → update cycle.

Decision framework

Use business questions to test semantic depth.

Which questions prove that a formal semantic layer is needed?

The practical test is whether the architecture can trace a business question from data to concept, rule, answer, and decision.

Business question What it tests Why it matters
Which similar equipment already showed this failure mode? Equipment classes, failure patterns, cross-FPSO analogues. Separates local asset navigation from reusable semantic reasoning.
What CAPEX / OPEX impact is associated with this anomaly? Connection between observed event, business impact, and financial model. Moves the graph from technical context to management decision support.
Can we explain why a recommendation was made? Traceability from observation to concept, rule, recommendation, and decision. Makes decisions auditable and reusable.
Can the model be reused across platforms? Portability, export, APIs, RDF/OWL compatibility, versioning. Reduces platform dependence and protects long-term meaning.
Can multiple métiers or affiliates align on the same definition? Concept governance, ownership, extension, and cross-domain reuse. Prevents each platform or affiliate from creating its own vocabulary.
Can agents reason on the business context? Concepts, rules, procedures, sources, justification, action. Grounds AI agents in TotalEnergies' meaning rather than platform-local context.

Rule of thumb

The more the question requires formal concepts, standards, cross-platform reuse, explainability, or AI grounding, the more TSF / SousLeSens should act as the semantic control layer.

Recommended actions

Three actions make TSF visible in operational decisions.

What should TotalEnergies do next, in what order, and how should success be measured?

The next step is a LifeX path with named owners, agreed questions, real data sources, and outputs used by business teams.

Now

Move LifeX from pilot validation to production deployment

Level 1 and Level 2 have established the semantic basis. TotalEnergies is now at the transition point: V2 should wire the LifeX graph into the weekly decision cycle, from inspection evidence to similar equipment patterns and Investment-ready reporting.

  • Metric: agreed competency questions.
  • Metric: weekly report traceable from observation to decision.
30-60 days

Wire TSF output into the platform ecosystem

Expose TSF concepts and mappings through APIs and integrate them into LifeX-relevant Cognite, Power BI, Alteryx, and Collibra workflows.

  • Metric: system integration rate.
  • Metric: semantic asset reuse rate.
30-90 days

Create the TSF portfolio registry and governance model

Document semantic assets, ownership, maturity, consumers, APIs, and decision impact for the portfolio.

  • Metric: audit completeness.
  • Metric: semantic drift resolution time.

Immediate workshop

Bring together the TSF lead, LifeX business owner, ontology lead, Cognite / DEP representative, and data product owner. Output: agreed LifeX competency questions and first semantic data product scope.

Final message

LifeX can show how TSF moves from semantic assets to daily decision support: evidence, concepts, rules, costs, risks, and rationale in one governed path.

Conclusion

The semantic layer exists; the decision on the table is making LifeX the first production proof point.

The case for TSF and SousLeSens as the governed semantic control layer for OneTech rests on three converging realities. Operational platforms multiply faster than shared meaning: Cognite, Fabric, Collibra, and future AI agents each need a stable reference for what equipment, degradation, cost, and risk mean across FPSOs. Without it, definitions are rebuilt locally in every tool and decision rationale stays fragmented. AI readiness requires explicit concepts, rules, and procedures before agents can produce governed, auditable outputs at operational speed. And TotalEnergies already has the assets: TSF, SousLeSens, domain models, standards alignment, and the LifeX use case. The work now is to make it visible.

LifeX is the right place to start. It combines high asset value, recurring decision cycles, fragmented evidence, and direct CAPEX/OPEX exposure. A governed LifeX semantic data product, connecting field observations to equipment analogues, risk, cost, and investment rationale, is the proof point that turns the semantic case from architecture into management evidence. Levels 1 and 2 have validated the approach. The transition from pilot to production is the decision on the table.

  • Own the semantic control layer: TSF governs meaning once, and platforms and AI agents consume it.
  • LifeX is the mission-critical first proof point: high value, high frequency, high stakes, and already partially modeled.
  • The operating model requires named owners across semantic authority, domain validation, and platform consumption.
  • Every major vendor platform embeds its own semantic layer: the risk of inaction is loss of portability and ownership of definitions.
  • Organizations without a universal semantic layer face 40% more AI rework and remediation by 2027.

The recommended path forward is to mandate LifeX as the first production deployment, expose TSF outputs to the platform ecosystem, and build the portfolio registry that makes semantic governance visible and measurable for OneTech leadership.