Everyone Is Talking About Context for AI. Here's What Most Companies Still Miss.

Celonis (sponsor content published on hbr.org, byline credited to "Harvard Business Review") · article · 2026-07-21 source ↗ Reader 🖥️

Context Engineering Enterprise AI Adoption & Readiness Retrieval & RAG Vendor Research Conflict of Interest Vendor Self-Reported Metrics

Enterprise AI initiatives are failing to deliver returns not because models lack capability but because they lack operational context — real-time, proprietary knowledge of how a specific business actually runs — and only once organizations supply that context (which the sponsor's platform is positioned to provide) can AI agents be trusted to act reliably at enterprise scale.

practical · argumentative — This is SPONSOR/native-advertising content from Celonis (a process-mining / enterprise-AI vendor) published on hbr.org under the HBR masthead — not a staff-written HBR piece. Surface form argues a general-sounding thesis ('AI needs operational context, not just generic context'), but the underlying practical ask is to adopt Celonis's 'Context Model' / process-mining platform (explicit CTA link at the end). Two customer anecdotes (Fujitsu, Novo Nordisk) are offered as evidence but are Celonis's own case studies, sourced from celonis.com with UTM-tagged referral links — self-reported, methodology-free metrics, not independently verified or benchmarked results.

Structure

Part	Role
Opening: AI hype cycle recap — from utopian speculation to 'where's the RoAI' skepticism	context / framing
'Eliminating Operational Blind Spots' — McKinsey stat (90% using AI, 10% scaling agents successfully) plus the generalist-vs-specialist framing (AI gives generic answers because enterprise data is proprietary/fragmented)	evidence + core argument
'Why Operational Context Matters' — before/after dialogue example contrasting a vague AI answer with a data-rich, specific one	illustrative narrative / argument by example
'Enterprise AI and Operational Clarity' — Fujitsu and Novo Nordisk case studies with a quote from a Novo Nordisk executive	evidence (vendor-sourced)
Closing restatement + explicit CTA to 'Learn more about Celonis and the Context Model'	conclusion / call to action

Key terms

Operational context
The combination of process data, business knowledge, and decision intelligence that gives AI a real-time, proprietary understanding of how a specific enterprise actually functions (as distinct from generic, public-internet-derived context).
RoAI (Return on AI)
The as-yet-unrealized business value/ROI enterprises expect from AI investment, framed as the open question the whole industry is now asking after the pilot-hype phase.
Generalist vs. specialist AI
A framing device: out-of-the-box AI models are 'generalists' (fluent but generic); AI grounded in a business's operational context becomes a 'specialist' capable of enterprise-specific judgment calls.

Propositions

explicitcorroboratedMost enterprises are using AI (nearly 90%, per McKinsey) but very few (10%) are successfully scaling AI agents to enterprise-wide value.conf: high
extendsPresentation: AI-First Software Delivery: Balancing Innovation with Proven Practices · p10 — This doc's McKinsey stat (90% adopt, 10% scale) is a concrete data point for the broader claim that most AI projects fail to deliver ROI because organizational approach isn't matched to actual context.
supports ←Thomas Nys' Post · p4 — The post's 'pilots stall, POCs can't scale' claim is a mechanism-level account of the same enterprise-scaling gap McKinsey's adoption-vs-scaling statistic (90% using AI, only 10% scaling it) measures at the industry level.
explicitGeneral-purpose AI models give generic, unhelpful answers to enterprise-specific questions because the relevant data (e.g. specific invoices, specific distributors) is proprietary, private, and fragmented across internal systems.conf: high
explicitcorroboratedWithout operational context (process data + business knowledge + decision intelligence), no AI agent can be trusted to make reliable real-time decisions or take effective actions that drive business outcomes.conf: high
supportsBuilding AI Agents For The Enterprise · p15 — Both claim enterprise-specific/organizational context (not generic AI) is the precondition for AI to be useful/trustworthy — this doc frames it as 'operational context', the other as giving Claude organizational context from the start to avoid generic first interactions.
supportsPodcast: Tiger Teams, Evals and Agents: The New AI Engineering Playbook · p10 — Both argue that proprietary data/domain specificity, not raw model capability, is what makes AI valuable/defensible in an enterprise setting.
supports ←Why AI agents need a Context Lake · p4 — Same claim from a different vendor's piece in the same genre: without operational/organizational context, agents cannot be trusted to make reliable decisions or take effective action.
extends ←How Netflix Built GenPage: a Single GenAI Model to Build Personalized Homepages · p2 — Both propositions argue that operational/contextual richness dominates raw model capability for reliable AI outcomes; GenPage extends the claim with quantified production evidence (6.9% vs 1.3% loss reduction) in a non-agentic, high-scale personalization setting rather than the enterprise-agent setting of the other document.
explicitProviding AI with operational context is what allows organizations to responsibly scale AI agents across the enterprise and realize ROI.conf: high
explicitExecutive-reported AI-adoption metrics overstate real progress because leaders see rosier returns than the people closer to day-to-day integration work.conf: medium
implicitA digital twin of the operation, complete with business context, is a near-necessary precondition for knowing where to deploy AI and automation.conf: medium
supportsGitLab 19.2 Puts AI Agents to Work on the Security Backlog · p8 — Parallel pattern: both articles present a headline vendor-sourced ROI/impact figure (this doc's Fujitsu/Novo Nordisk metrics; GitLab's self-commissioned 400% ROI study) without flagging the conflict of interest — same critique applies to both.

Arguments

Enterprises should adopt an operational-context layer (i.e., a platform like Celonis's) before or alongside AI deployment in order to realize AI ROI at scale.

Explicit premises:

Generic AI models produce generic, low-value answers to enterprise-specific questions.
Enterprise data needed for specificity is proprietary, private, and fragmented.
Operational context (process data + business knowledge + decision intelligence) closes that gap and produces specific, actionable, trustworthy answers (illustrated by the supply-chain reorder example).
Companies that have adopted operational-context tooling (Fujitsu, Novo Nordisk) report strong results.

Implicit premises:

A vendor's own selected customer case studies, without disclosed methodology or independent verification, constitute sufficient evidence that a general causal claim ('operational context → ROI') holds broadly, not just for these two accounts.
The specific, richly-detailed 'good' AI answer in the demonstration dialogue is achievable in practice at the fidelity shown, rather than being an idealized illustration constructed for persuasive effect.
The article's own vendor (Celonis) is representative of, or equivalent to, 'operational context' in general — i.e., the argument for the general concept doubles as an argument for the specific product.

Inferences:

If operational context is genuinely the binding constraint, then AI vendors and enterprises should be able to produce independently verified, apples-to-apples RoAI comparisons (with vs. without operational-context tooling) rather than relying on anecdote — the article does not attempt this.

Narratives

A hypothetical exchange where a manager asks an out-of-the-box AI model about a demand spike and reordering; without operational context the AI gives a vague hedge, but with operational context it names a specific region, SKU, stockout risk percentage, lead times, and a recommended action (authorize air freight).
→ Makes the abstract 'generalist vs. specialist' proposition vivid and concrete — shows rather than argues for the value of operational context.
Fujitsu case: using operational context for inventory management, achieving a reported 20% reduction in excess inventory and multimillion-dollar savings across 10 processes.
→ Supplies a quantified, named-company proof point for the article's central claim.
Novo Nordisk case: using operational context and 100+ AI agents to speed clinical development, with an executive quote about needing a 'digital twin' of the operation, expecting a 12-month reduction in time to market.
→ Supplies a second, higher-stakes (pharma) proof point and a customer voice endorsing the underlying premise.

Questions addressed

✅ Why are enterprises not seeing the AI returns (RoAI) that early hype promised, despite near-universal AI adoption?
Because AI models are generalists lacking the enterprise's proprietary operational context, so they can't make reliable, specific, trustworthy decisions at scale.
⭕ How would one independently verify that 'operational context' (as opposed to other factors — organizational readiness, change management, data quality generally, model capability improvements over time) is the binding constraint on AI ROI, rather than just a plausible-sounding, commercially convenient explanation?
Not addressed — the article treats its causal claim as established via anecdote and framing rather than testing it against alternative explanations.

Gaps

incompleteThe article presents exactly two customer stories (Fujitsu, Novo Nordisk), both sourced from Celonis's own case-study pages with UTM referral tracking, as evidence for a general causal claim about operational context driving AI ROI. No independent benchmark, third-party audit, control group, or disclosed methodology accompanies either figure (20% inventory reduction, 12-month timeline reduction, 100+ agents). This is a textbook vendor self-reported metrics pattern, and per the reader's stated preference for benchmark data and practitioner experience over vendor claims, this piece should be read as marketing framing with a real (if unoriginal) causal intuition, not as rigorous analysis.
Read next: Look for independent, third-party research on enterprise AI ROI drivers (e.g. academic studies, analyst firm reports without vendor sponsorship, or practitioner postmortems) that isolate 'context/data quality' from other commonly cited blockers like organizational readiness, change management, and governance.
incompleteThe article never engages with competing or complementary explanations for why '90% use AI but only 10% scale it successfully' — e.g. organizational readiness, unclear ownership, change-management failure, poor evals/feedback loops, or unrealistic expectations set by leadership — all of which appear elsewhere in this corpus as major explanatory factors for enterprise AI adoption stalling. By attributing the gap almost entirely to a missing operational-context layer (conveniently the sponsor's product category), the piece is incomplete about the multi-causal nature of the 90%-vs-10% gap.
Read next: Compare against extract-building-ai-agents-for-the-enterprise and extract-ai-first-software-delivery, which name proprietary data/domain specificity as one factor among several (trust relationships, domain expertise, matching approach to context) rather than the singular cause.