Product shape
Three layers sold as one platform, wrapped in packaged services.
Packaging: Quick Start (3 data sources, 2 of 5 agents, "under 8 weeks to first value"), then solution areas sold as core plus add-ons plus services, then OneSupport and Success Tracks. Cognite Embedded resells the platform under partner brands, such as Rockwell's DataMosaix and SLB's OptiFlow.
Use case map
Five solution areas in the nav, four older ones still live, nine industry pages, 18 named agents and a 34-entry solutions library. Every value below is Cognite's own claim.
Solution areas
| Area | Problem | What Cognite does | Agents and add-ons | Claimed value | Proof |
|---|---|---|---|---|---|
| Advanced Troubleshooting | RCA takes weeks; SMEs hunt for data | One workspace with P&IDs, 3D, documents, work orders and live data; no-code RCA in Canvas and Charts | Equipment templates, Observations, bad-actor dashboard, RCA agent per equipment class | Downtime −15-30%; EBIT +1-2% | NOVA Chemicals: decoking model scaled to the whole site in under 48 h |
| Field Operations | Field can't find information; paper rounds | Prepare, plan, run rounds with readings, capture photos, troubleshoot against live data, monitor progress | Auto checklists, 3D tag access, SAP S/4HANA notification writeback | Throughput +10-30% | Aker BP: maintenance execution time −30-80%, visual inspection −50% |
| Turnaround Planning | Plans built by hand fail in the field | Work orders and risk on P&IDs; digital work packs 6-12 months ahead; SimOps risk view; progress tracking with rescoping | TAR planning agent, plan optimiser agent, P&ID parsing; extra extractor $10-20k | Planned downtime −15-30% | Celanese: over $5M cost avoidance; another TAR $4.6M est. |
| Integrated Supply Chain New | Plant optimises OEE, logistics optimises OTIF | SCOR data model joined to production and maintenance; cascade impact; AI root cause; resolution plans | Risk and Deviation Watchtower agent, Supply Chain Data Scout | OEE +20%, OTIF +15%; protect 5% of revenue | No customer named (Jun 2026) |
| AI Solutions Library | Each use case built from scratch | 34 ready apps and agents: 7 by Cognite, 27 by partners | See below | Two entries carry a number | IQumulus client: over $50M a year |
| Production Mgmt & Optimization | Unused capacity; ageing brownfield | 11 use cases: KPI tracking, RCA, well monitoring, capacity, gas lift, virtual flow metering, energy drivers, load forecast, energy optimisation, GHG reporting | Physics simulators in the loop | Throughput +10-30%; emissions −30-50% | OMV, Aker BP Skarv |
| Asset Performance Mgmt | APM programmes stall in silos | Connect, map, discover, low-code insight, field data, health monitoring with equipment templates, planning, field execution | Delivered with Accenture, Rockwell, Pinnacle | Forrester 2022: 400% ROI | Celanese |
| Unified Operations Insight | Every site reports differently | Cross-site KPIs (OEE, MTBF, emissions per unit) on a no-code calculation engine; benchmarking; dashboards | Grafana and Power BI connectors | None stated | Aarbakke, Moelven, Aker BioMarine |
| Power Market Operations | Intraday markets outpace decisions | Price forecasts, bid generators, virtual power plants, imbalance management, ETRM | Market connectors | None stated | Hafslund Eco, Lyse, Equinor |
Industries
Six industries in the nav on an older template; Energy, Petrochemicals and Life Sciences on a newer one that leads with numbers.
Energy
- Upstream operations: visibility across the asset network, well intervention planning EBIT +1-2%
- Midstream: model the pipeline network, prioritise interventions
- Continuous HAZOP: ongoing risk discovery with an auditable record
Upstream
- Production optimisation: capacity, energy use, early warning
- Well monitoring: real-time operating envelopes
- Virtual flow metering and allocation: hybrid physics and data $3.5M/yr est. at Aker BP
- Vessel speed optimisation: PGS over $2M/yr
- APM for rotating equipment and turbines
Downstream
- Operations and throughput: simulators, optimisers, guided advisors
- Maintenance and reliability: digital twin, predictive analytics, turnarounds
- Field safety and productivity: mobile data, 3D, computer vision reads equipment tags
- Autonomous operations: robot rounds, remote inspection
- No-code calculations reused across sites
Petrochemicals
- Troubleshooting downtime −15-30%
- Turnaround planning planned downtime −15-30%
- Field operations throughput +10-25%
- AI agents for common operator tasks
- Repeatable data products across sites
Life sciences
- Golden batch: compare inputs and yield across lines
- Reliability and maintenance: faster RCA cycles
- Technology transfer: lab to production, reuse recipes
- RCA and batch disposition under GxP (library)
Continuous process
- Unified operations insight across acquired sites
- Brownfield APM and digital twins
- Cross-site KPIs: OEE, MTBF, emissions per unit, energy load
- Trusted data: quality rules and profiling
- Predictive maintenance: Elkem off-spec costs over €100k per incident
Power generation
- Brownfield APM with prescriptive insight
- Power trading across markets
- Portfolio reporting for acquired plants
- Offshore wind predictive maintenance (Ørsted)
- Turbine start-up optimisation (Hafslund Eco hydro)
Power grid
- Grid connection analysis on CIM digital twins 60% faster
- Asset risk: transformer health index $2M/yr est.
- Dynamic line ratings from weather and sensors
- Cross-site KPIs and APM
Renewables
- Remote-asset APM with drones and robots
- IRR improvement on new and existing projects
- Portfolio visibility and trusted data
- Computer vision for bird and bat detection
Atlas AI agents named
| Agent | What it does | Sold as |
|---|---|---|
| Operations Summarizer | Pre-shift summaries of what changed; role-specific handover reports | Quick Start option |
| Work Package Generator | Assembles maintenance work packages on a Canvas; 30 min → under 5 min | Quick Start option |
| Asset RCA | Root cause and failure assessment for one asset | Quick Start option |
| Operations Dashboard | Leadership view of unit performance, reliability, safety, quality | Quick Start option |
| Blueprint | Turns P&IDs and PDFs into live monitoring dashboards | Quick Start option |
| Work Order Classifier | Classifies work orders to ISO 14224 failure modes and writes back | Library |
| Time Series Trender | Trends, correlations and anomalies without scripting; "+20% analysis speed" | Library |
| Data Scout / Data Exploration | Finds assets, files and series in plain language and drops them on a Canvas | Demo, templates |
| Troubleshooting agent | Lists causes with sources, builds a cause map, checks data against design specs, drafts work orders | Troubleshooting add-on |
| TAR planning and plan optimiser | Summarise turnaround documents, answer questions, suggest scope | Turnaround add-ons |
| Supply chain Watchtower and Data Scout | Monitor production, demand, inventory, logistics; map the cascade; query the SCOR model | Supply chain add-ons |
| Production engineer, energy optimisation, sustainability | Correlate process, energy and equipment data; recommend operating changes and alert thresholds | Demo agents |
| Simulation agent (Aker BP) | Picks the simulation model, sets inputs, runs queued what-if batches; hours to minutes | Customer-built |
| WellView agent (national oil company) | Plain-language search over ~750,000 drilling log entries for 1,000+ wells | Customer-built |
AI Solutions Library
| Provider | Entries | Focus |
|---|---|---|
| Cognite (7) | Operations Dashboard, Asset RCA, Blueprint, Work Package Generator, Work Order Classifier, Operations Summarizer, Time Series Trender | Generic agents for any process plant |
| SLB | OptiFlow production assurance, OptiSite facility and pipeline | Upstream: throughput up to +20%, opex −15-35% |
| Rockwell | FactoryTalk DataMosaix | Cognite inside, Rockwell brand |
| Pinnacle | Newton Intelligence, RDMS, Integrations | Reliability programmes |
| Tridiagonal.ai | What-If Simulator, OpsEdge, EmitOps | Decoking, work orders, emissions |
| Radix | JO.ai, Operational Overview Control Tower, OEE | Document copilot, site overview |
| Tulip | Quality Management, Traceability & Compliance, Digital Guidance, Production Management | Shop-floor apps, GxP records |
| Ignos | Pulse, Engage, Inspect, Keep, Move | Machining shops; Framo recovered 150 h of uptime a day on 62 machines |
| IQumulus, Omny | CorpFinML executive decision intelligence; OT security | Finance over the graph (over $50M a year at one client) |
| Unlabelled | Energy Optimization, DC Ops Intelligence, Golden Batch, RCA & Batch Disposition | Energy, data centres, life sciences |
Quick Start demo: search, Canvas, agents
The guided demo you sent: four chapters, about 30 steps. It opens and closes on "30% faster time to impact, 25% more efficiency, 465% ROI with 6-month payback".
Search across every source
- One box over P&IDs, work orders, manuals, data sheets and time series.
- Facets show the size of the graph: 1,105 assets, 363 time series, 6,493 maintenance activities, 2,399 checklists.
- AI search inside documents; switch between list and 3D view.
3D with context
Pick the first-stage compressor in the 3D model; the panel shows properties, tag tree, 26 linked assets, time series and files.
Same data on a phone
The 3D model streams to mobile; tap a valve to get its tag and data sheet. Search follows the worker into the field.
Industrial Canvas
- One board for a problem: smart P&IDs with clickable tags, trend charts, 3D, the compressor manual, an RCA report and work-order cards.
- Comments with @mentions bring in an expert; versions and governance built in.
- Agents work on the board: they add data, summarise documents and draw cause maps.
Agent finds the asset
"Find the 1st stage compressor asset" returns name, space and external ID with a source chip and "Show reasoning". Next prompt: add its P&ID, manual and images to the Canvas.
Agent reviews anomalies
Per-tag findings (oil viscosity trending up), sparklines as sources, and next steps: correlate with work orders around 7 Jan or narrow the window.
Charts: no-code calculations
Node graph (source, subtraction, output) with a function library: data quality, detect, equipment, filter, fluid dynamics. "Save and schedule" turns it into a monitor with thresholds and alerts. Claim: RCA time −50%.
Agents inside the tools
An agent picker docked in Canvas: Time Series Trender, Atlas Data Scout, Work Package Generator, Operations Summarizer, Work Order Classifier, plus customer agents. Quick Start includes 2 of the 5 stock agents.
Operations Summarizer
Generates a PDF for 1 Jun to 14 Aug: 54 completed work orders, the critical ones, trends (30 of 54 on one asset), durations. Pitch: shorter handovers, fewer meetings.
Work Package Generator
Lays out a work package on a Canvas: work order, asset, P&ID, time series, inspection checklist, related activities. Claim: 30 minutes to under 5.
Low-code agent builder
- Form fields: sample prompts, model (azure/gpt-4o-mini here), goal, instructions, tools such as get_asset.
- Live preview shows the thought process and each tool call: list_maintenance_orders with isOpen and priority 1 returns 2,306 critical open work orders.
- Draft, auto-save, Publish.
Agents in action: from symptom to SAP
A refinery's first-stage compressor shows excess oil discharge. About 23 steps take it from a blank Canvas to a confirmed work order, a purchase order and an SAP update.
Causes with sources
- The agent gathers the asset, P&ID, manual, one month of trends and recent work orders onto the Canvas.
- It lists causes and corrective actions, citing the asset and page 7 of the compressor manual.
- Recent work orders on the board come from SAP.
A living cause map
- "Create a cause map" draws a fishbone on the Canvas: excess oil discharge from improper viscosity, overfilled crankcase, blocked intake filter, worn valves, worn rings, humidity.
- The engineer marks each cause confirmed, uncertain or rejected.
Checks against design spec
The agent compares operating data with the design spec and finds the lab-sampled oil viscosity far above range. It recommends 30wt non-detergent oil and offers to create a work order.
Drafts the work order
WO-2025-001 with description, dates and cost, then "Does this work order look good to you?" The engineer confirms.
Orders the missing part
- The agent checks stock, finds no oil, raises a PO for 20 gallons, reads the usual one-week lead time from history and moves the work order out ten days.
- Once approved, the schedule is set by resource availability and sent to SAP.
Production engineer agent
In Charts, the agent finds the compressor without tag names, plots flow, temperatures and vibration, then concludes: throughput consistently below design, vibration healthy, operating safely but under-used. When the request is too big it offers to run on aggregated data.
Report with next steps
Quantify the gain of moving flow toward the design point, consider intermittent high-load running, re-run quarterly, and set alerts: flow under 2,500 m³/h for 30 min, vibration over 7 mm/s for 5 min.
Flows: role-based homes and custom apps
Launched 12 May 2026 as "the new standard for industrial workers". Claims: 100x faster to build, 50+ customers enabled, 1,200+ production apps, 70% less manual work, 25% less downtime.
A home per role
- Widgets the user asks for in chat: throughput against target, a trend, an AI-generated widget.
- Shift handoff notes from operators ("P1 circuit break, already fixed"; "pipe leak repair scheduled").
- Continue where you left off: a steam leak chart, a turnaround work package, an RCA session.
Chat builds the view
"Add bearing temp of pump 402, weekly" adds a widget. A template gallery offers alarm rationalisation, battery limit balance, instrument calibration, production allocation.
Custom apps
Built with agentic coding tools (Claude and Cursor named) on Cognite's SOC 2 hosting, grounded in the knowledge graph: an ISO 14227 workflow, pump monitoring, cause analysis.
Agentic troubleshooting app
- Five-step stepper: equipment overview, data retrieval, troubleshooting, notification, handoff.
- The agent assesses ten pump symptoms as present, absent or inconclusive, each with data, linked work orders and a confidence (vibration 8.2 mm/s vs 7.0 alert, 88%).
- The engineer can override any verdict before finalising.
Turnaround planner
Executive daily brief for a 2027 refinery turnaround: at risk, day 7 of 32, schedule +1.5 days, cost +8.2%, scope +12.4%, with a situation summary.
Refinery dashboard
Safety view on an aerial plant map: overdue inspections 8.5% vs 8% target, 1,966 safety observations a week, 15% invalid permits, ISO 45001 incidents, 55 work orders.
Supply chain: every alert priced in euros
The first solution area outside the plant (Jun 2026), built on Flows with a SCOR data model.
Impact analysis
- A blocked material behind a late shipment: €87,000 revenue at risk, €13,050 late penalty, €2,400 to expedite, €10,650 saved if actioned now.
- An impact chain from work order to asset, production schedules and customer orders.
- Three recommended actions, each with a Start button.
Manager home
178 active shipments, 7 delayed, 8 alerts; inbound shipments for the next 72 hours with delay and impact; a logistics map; Ask in Chat.
Resolution wizard
Analysis, execution plan, risks and monitoring. Tasks per role for the next two hours (inventory analyst in ERP, logistics operator in WMS) and a 72-hour checklist.
Field: InField and Maintain
The older field apps (2023 demo UI), now inside Flows. Claims: maintenance execution −30-80%, routine inspection −75%, scheduled work orders 30-50% faster.
Plan and assign
Operator rounds replace paper with checklist templates (a temperature reading between 10 and 20 °C). Maintain turns work-order schedules into field tasks: a heat exchanger cleaning with isolate, clean and de-isolate steps and an interactive isolation plan.
Copilot in the field
"Find anomalies for heat exchanger 9115 two to three days ago" returns the exceedance with time and value, plus buttons to show the data, open work orders, or pin it. No call to the control room.
Value and proof
How Cognite turns use cases into money, and what backs it.
$1B "verified" value, by source
Unplanned downtime and production cost; frontline productivity; production increase; faster time to value. A customer value team tracks KPIs the customer picks and "co-validates" them with the operator; there is no audit. Target: $100B cumulative customer value by 2035.
Headline claims by pillar
| Pillar | Claims used across the site |
|---|---|
| Uptime | Unplanned downtime −15-30% (troubleshooting); planned downtime −15-30% (turnarounds, Maintain) |
| Throughput | +10-30% (field ops, production, heritage page); EBIT +1-2% (troubleshooting) or +2-4% (scaling page) |
| Frontline | Maintenance execution −30-80%; workforce productivity +5-15%; routine inspections −75%; RCA time −50% |
| Time to value | Under 8 weeks to first value; 7 weeks to measurable impact; 9 facilities in 4 months |
| Energy and emissions | Emissions −30-50% (production page); GHG reporting templates |
| Supply chain | OEE up to +20%; OTIF up to +15%; protect up to 5% of revenue |
How it is packaged
| Offer | What is in it | Claim |
|---|---|---|
| Quick Start | Search, Canvas, Charts, operations map, data model, agent workbench; 2 of 5 agents; up to 3 data sources; discovery workshop, contextualization, InField-lite, Value Launchpad workshop; on Azure and Fabric | Under 8 weeks to first value |
| Solution areas | Core plus add-ons plus services that extend Quick Start to a site, e.g. "configure and deploy one AI agent" | Per-area KPIs above |
| OneSupport | 24/7 support across native apps, custom code and third-party logic | 15-minute critical response; 99.9% uptime |
| Success Tracks | Solution integration, value acceleration, GenAI enablement | — |
| Embedded | White-label platform for OEMs and ISVs (Rockwell, SLB, MaMaTa) | 25-50x faster to market |
| Value Review | Free 30-minute call, five questions, a benefits report within a week | Replaces an ROI calculator |
Two commissioned Forrester studies are live at once: 2022, 400% ROI, $21.56M NPV, 14-month payback; 2026, 465% ROI, $29.4M NPV, under 6 months, from 7 decision-makers at 4 organisations.
Customer proof with numbers
| Customer | Use case | Result (vendor-claimed) |
|---|---|---|
| Offshore operator, 30 platforms | Maintenance planner across 29 assets | 30% fewer planned shutdowns; +700,000 bbl/yr; $38M est. a year |
| Fortune 500 energy (via IQumulus) | Executive decision intelligence | Over $50M a year; cash forecast +20-30% |
| Equinor | Robotic inspection and repair | $20M saved by avoiding a shutdown; 1,000+ autonomous Spot rounds |
| National oil company | Oil-quality prediction, separation set-points | $11.5M est. on one train |
| Aker BP + Siemens | Condition monitoring | $6.5M est.; 1,300 h of integration saved |
| Aker BP | Field operations, 3D tags | Maintenance execution −30-80%; routine inspection −75% |
| Aker BP + Framo | Operator-to-OEM data sharing | Pump shutdowns −70%; availability +40% |
| Aarbakke | Machining data foundation, 15 use cases | $6M a year, 6% of revenue, after 11 months |
| Celanese | Seal monitoring inside a shutdown window | Over $5M cost avoidance |
| Polyethylene producer | Turnaround planning | $4.6M est.; scaled to another asset in two days |
| OMV | Surface data hub | Inspection planning −50%; deferment −6% |
| Statnett | Grid connection, transformer health | $1.2M and $2M est. a year; analysis 60% faster |
| Framo (via Ignos) | Machine uptime apps | 150 h uptime a day across 62 machines |
| Koch Ag & Energy | Multi-site foundation, InField | 6 sites in under 6 months; 7th in under 2 |
| Nippon Shokubai | Plant data discovery | ~9,000 h a year saved; 80% of contextualization by AI |
| Tokuyama | Plant-wide integration, InField | ¥200M est. a year |
| Top-10 pharma | AI workflows on Flows | 4 days vs 3-5 months; 70% less manual data gathering |
86 stories in total, 61 with a customer number. Aker BP appears in about 13. Most dollar figures are labelled estimated; newer stories (2023-26) are mostly Japanese chemicals and refining, measured in hours saved.
Analysis
Fact = from Cognite's site, demos or press. View = our assessment.
Platform Fact
- Data Fusion: 90+ extractors, AI contextualization of P&IDs and entities, Core Data Model, 107T+ data points in context.
- Security: ISO 27001, SOC 2 Type 2, IEC 62443, GxP; Private Link deployment.
- Runs on Azure, AWS and Google Cloud; Quick Start sold on Azure Marketplace.
AI Fact
- Atlas AI in over 70% of 2025 bookings; 59% of customers by ARR use it.
- Agents cite sources, show reasoning and tool calls, and write back to SAP after human approval.
- LLM choice with an in-house LLM/SLM benchmark; NVIDIA anomaly models integrated.
Go-to-market Fact
- Land with Quick Start, expand by solution area SKU, retain with OneSupport.
- Resold under partner brands through Embedded (Rockwell, SLB).
- A value team tracks customer KPIs into the "$1B verified" figure.
Moat View
- Real: the contextualized graph per site is hard to rebuild; years of energy and chemical templates; packaged offers with numbers; Schneider and AVEVA distribution after close.
- Thin: proof leans on one shareholder customer (Aker BP); most money figures are estimates; agents run on licensed models anyone can buy.
Weak spots View
- Everything starts from data that already exists: historians, SAP, P&IDs, 3D. Plants without them get little.
- Frontline capture is a checklist app and free-text handoff notes; structured worker events are absent.
- Site inconsistencies: two Forrester studies live; one $4.6M case credited to two customers; 50,000 vs 3,000 hours for Idemitsu.
- The AVEVA takeover weakens the "any cloud, any data, no lock-in" pitch.
Market Fact
- Energy first: Aker BP, TotalEnergies, Equinor, OMV, Aramco, ADNOC; then chemicals (Celanese, Koch, NOVA), Japan (Idemitsu, Cosmo, Tokuyama), power (Statnett, Ørsted), life sciences (B. Braun).
- ~80% of revenue from outside Aker; ~40% from outside oil and gas.
- IDC MarketScape and Verdantix leader; no Gartner placement found.
Cognite vs Minder
Where the two products start, and where they meet.
| Cognite | Minder | |
|---|---|---|
| Starting point | Existing historians, SAP, P&IDs and 3D models, contextualized into a knowledge graph | First-party event capture from workers and signals; legacy systems as adapters |
| Primary user | Engineers, SMEs, planners, data teams at large sites | Frontline workers and managers on one event log |
| Work surface | Industrial Canvas, Charts, role-based Flows homes | Chat with artifacts, dashboards, floor apps |
| AI stack | Atlas AI agents over the graph; low-code builder; agents act through SAP with approval | Agent loop over the operational graph via read-only Trino, cited answers; world model in training |
| App creation | Flows custom apps with Claude or Cursor; 1,200+ claimed | AI app creation for forms, checklists and handovers (planned) |
| Value story | Downtime, throughput, frontline time, time to value; priced alerts; $1B "verified" | Output, Quality, Energy, Workforce; pilot KPIs on the customer's baseline |
| Deployment | Quick Start under 8 weeks; 9 facilities in 4 months claimed | First site six weeks; repeat site two to three engineer weeks target |
| Geography | Norway, US, Japan, Middle East; soon AVEVA's channel | Vietnam proving ground, Europe next |
Sources
- Product screens: captured from Cognite's public Navattic demos in Chrome, 28-29 Oct 2026: Data Fusion Quick Start, Atlas AI for RCA, Atlas AI for Production Engineers, Cognite Flows, Integrated Supply Chain, Field Productivity for Manufacturing. Demo data is sample data.
- Product pages: Data Fusion, Atlas AI, Flows, Quick Start, AI Solutions Library, extractors, OneSupport
- Solution areas: troubleshooting, field operations, turnaround planning, supply chain, production, APM
- Industries: energy, upstream, downstream, petrochemicals, life sciences, continuous process, power generation, power grid, renewables
- Value: $1B verified value release, Forrester TEI 2026, Forrester TEI 2022, Value Review, 86 customer stories
- Company and deal: Dow Jones via Morningstar, 30 Jun 2026, Techzine, Aker release via Inderes, Aker Q4 2025 slides (revenue, ARR), About Cognite
- Analysts: IDC MarketScape 2026, Verdantix
Conflicts left visible: founded 2016 (Cognite) vs 2017 (deal coverage); revenue $170M (Cognite, unaudited) vs $164M (Aker); two Forrester studies; the $4.6M turnaround case credited to two different customers. The deal had not closed as of 29 Oct 2026. Pricing, the LLM benchmark results and most FAQ answers are not public.