Primary-Verifying Deloitte's "Executive's Framework for AI Capability Assessment": The Page Is Real, the Description Was Wrong, and It Is the Most Agentic Framework in the Set
The question
"Primary-verify Deloitte's 'executive's framework for AI capability assessment' against Deloitte's own page, the last unverified Big-4 claim the vault still carries."
Context: [[2026-07-15-agentic-assessment-framework-competitive-landscape]] carried one Deloitte sentence in its do-not-quote Big-4 paragraph. [[2026-09-26-big4-agentic-readiness-frameworks-primary-verify]] verified the McKinsey, Accenture and KPMG claims in that paragraph and left Deloitte out of scope. This brief closes it.
Verdict: PARTIAL. The asset exists at the cited URL under the cited title. The description of what it does is not supported by Deloitte's page or its PDF, and the July brief's synthesis mischaracterizes it.
What we already know (from the vault)
- The exact claim, from the July brief's Big-4 paragraph: Deloitte publishes "an 'executive's framework for AI capability assessment' staging a progression from experimentation to strategic integration." It was tagged secondary, sourced from a search-result summary. See [[2026-07-15-agentic-assessment-framework-competitive-landscape]].
- The July brief's synthesis then used Deloitte twice more without a flag: as one of four firms "all scoring the same five-ish dimensions (strategy, data, technology, organization/talent, governance)," and implicitly inside the "Big-4 dimensions are pre-agentic" argument.
- The September brief found a 20% clean-pass rate on the other five claims in that paragraph (1 confirmed, 1 partial, 1 contradicted, 2 not found), and already narrowed the "pre-agentic" line after finding the Accenture/SEI eight-dimension model. See [[2026-09-26-big4-agentic-readiness-frameworks-primary-verify]].
- A vault-wide grep for the framework title and URL slug returns only those two briefs. The July brief is the sole carrier of the claim, so correcting it corrects the vault.
- Deloitte otherwise appears in the vault as a platform-and-module seller ("AI Advantage for CFOs," 14+ modules / 30+ agents). See [[2026-06-28-snowflake-si-cortex-positioning-caf-gap]].
What the web says
Retrieved 2026-11-11. Method note: deloitte.com served HTTP 200 to both WebFetch and plain curl with a browser user-agent, so no Playwright was needed. The article body and the PDF are both ungated - no form, no login. Term counts below were run on the raw HTML and on pdftotext output, not on a model summary.
- The title exists, but only as the browser title and URL slug. The page's
<title>is "An executive's framework for AI capability assessment | Deloitte US," which is what a search snippet displays. The on-page headline is different: "Preparing digital channels for agentic AI evolution," subtitle "An AI capability assessment for scalable, trusted customer experiences." Labeled "Perspective," dated 05 May 2026, 4-minute read, authors Siva Muthu, Anantha Ramadas, Jayanta Ghosh and Harshad Deshpande, filed under Services > AI & Engineering > Engineering. The word "executive" appears once on the page, in that title tag. (Deloitte) (primary-verified) - "A progression from experimentation to strategic integration" is NOT on the page or in the PDF. "Strategic integration" appears zero times in either. "Experimentation" appears once on the page, as a capability in a list ("multimodal interaction and experimentation capabilities"), meaning A/B-style measurement, not a maturity stage. The page contains zero instances of "maturity," "stage" or "level." The staging language was the search summarizer's invention. (primary-verified absence)
- What the framework actually is: a three-dimension agentic-readiness assessment scoped to customer-facing digital channels. Deloitte's words: "Readiness depends on three dimensions: embedded capabilities, agent operability and trustworthiness, which together form a practical AI capability assessment framework." The scope is websites, apps, APIs, devices and partner ecosystems. It is not an enterprise-wide AI maturity model. (primary-verified)
- The attached PDF adds the structure the web page omits. Title "Agentic AI capability assessment for digital channels," May 2026, 15 pages, 11MB, open download. Its Figure 4 breaks the three dimensions into 15 capability areas: Dimension 1, embedded agentic AI capabilities (6: intelligent assistants and guided journeys; discovery and decision support; personalization and adaptive content; content and knowledge delivery; integrated service workflows; measurement and experimentation). Dimension 2, digital channel operability for external agents (4: structured content and semantics; machine-readable policies and protocols; Model Context Protocol; agent navigation and actionability). Dimension 3, trustworthiness (5: governance, safety and brand control; performance and accessibility; ethical and regulatory compliance; observability, accountability and cost; data quality and provenance). (Deloitte PDF) (primary-verified)
- There is a maturity scale, but it is a different one from the claim. PDF Figure 5: "Each dimension is scored on a five-level maturity scale: Level 1: Initial ... Level 5: Optimized." Only the two endpoints are named in the extracted text; levels 2-4 are not labeled. This is CMMI-style endpoint vocabulary, not "experimentation to strategic integration."
- The process is assessment-to-roadmap, not assessment-to-build. Figure 5's flow is: align strategic goals, identify, assess (conduct evaluation, identify whitespaces, opportunity identification), adoption roadmap. The roadmap step lists "prioritized capabilities and projects, proof of concepts, MVPs, value measurement approach and governance mechanisms." The output is a prioritized plan. No tooling, automation or executable artifact is described.
- No per-agent authority or autonomy scoring. The trustworthiness dimension covers governance and brand control at the channel level. Nothing in the 15 capability areas scores how much authority an individual agent may hold.
Convergences and contradictions
- Claim-level scorecard: PARTIAL. Confirmed: Deloitte published an asset under that title at that URL. Not found: the "experimentation to strategic integration" progression. Unstated in the claim but material: the framework is channel-scoped and explicitly agentic, and the "executive's framework" title is an SEO title tag that the page's own headline does not use. The backlog prior ("this claim is also wrong") was half right - the description was wrong, the existence was not. Updated sweep total across all six Big-4 claims: 1 confirmed, 2 partial, 1 contradicted, 2 not found.
- The July synthesis sentence that lists Deloitte among four firms "scoring the same five-ish dimensions (strategy, data, technology, organization/talent, governance)" is CONTRADICTED for Deloitte. Deloitte's three dimensions are embedded capabilities, external-agent operability and trustworthiness. Only governance overlaps. After the September brief, that sentence was already false for Accenture and KPMG; it is now unsupported for every firm it names.
- "Pre-agentic" is flatly wrong for Deloitte. This asset was published 2026-05-05, ten weeks before the July brief, and agentic readiness is its entire subject. One whole dimension, operability for external agents, names MCP and machine-readable policies as assessed capabilities. That is a dimension none of the McKinsey, Accenture/SEI or KPMG frameworks carries.
- Convergence with the vault that survives: the July brief's structural finding - assessment output is a roadmap, not an executable build input - holds for Deloitte on primary evidence. So does the September brief's narrowed autonomy claim: no verified Big-4 framework scores authority at the agent level.
Synthesis for RDCO
The quotable version of the Deloitte line is now available, and it says something different from what the vault carried. Replace the July sentence with: "Deloitte's May 2026 'Agentic AI capability assessment for digital channels' scores customer-facing channels on three dimensions - embedded agentic capabilities, operability for external agents, and trustworthiness - across 15 capability areas on a five-level maturity scale, and ends in an adoption roadmap." Every element of that is on Deloitte's own page or PDF. Drop "executive's framework" as the asset's name in anything a phData reader might see; cite it by its headline or PDF title, since that is what a former Deloitte person would recognize. Drop "experimentation to strategic integration" entirely.
The more useful finding is that Deloitte scopes down where everyone else scopes up, and that leaves the enterprise-internal lane open. McKinsey Rewired, Accenture/SEI and the generic readiness quizzes all assess the whole enterprise. Deloitte assesses only the customer-facing surface: can your website, app and APIs host agents, and can someone else's agent transact on them. That is a customer-experience and digital-engineering sale, sold by an Engineering practice, not an operating-model sale. phData's Organizational Intelligence work (Org Map, IMA, OIP) looks at the inside of the organization - who decides, which agents hold which authority, how work routes. On the public evidence these two do not collide. That is a cleaner differentiation statement than "they are pre-agentic," and it has the advantage of being true.
The dimension worth borrowing is "operability for external agents." It is the only genuinely new axis in the whole Big-4 set: structured content, machine-readable policies, MCP exposure, agent navigability. The inward-facing analog is a real question for an agentic estate - are internal systems, data products and policies legible to agents, or only to humans - and it is a Snowflake-shaped question (semantic views, governed MCP endpoints, machine-readable access policy) that sits squarely in phData's delivery strength. Whether OI already covers this under another name is a judgment for Ben, not something this brief can settle from public material. Flagged as Ray's suggestion, not a gap finding.
Two things hold, one thing needs retiring. Holding: the assessment-to-build coupling as the defensible layer (Deloitte's flow stops at a roadmap), and agent-level authority as an unscored axis across all four firms. Retiring: any remaining use of "the Big-4 score the same five dimensions" or "the Big-4 are pre-agentic." With this brief, neither sentence has a single primary-verified firm behind it. Confidence note: absence of an autonomy axis and of build tooling is verified against one web page and one 15-page PDF. Deloitte's actual client assessment instrument is not public, so "does not score agent authority" means "does not publish that it does."
Why this is in the vault
This closes the last open item on the do-not-quote Big-4 paragraph in [[2026-07-15-agentic-assessment-framework-competitive-landscape]], so every competitor-framework claim that could reach a phData Organizational Intelligence competitive-landscape artifact now has a primary citation or an explicit deletion. It also supplies the replacement Deloitte sentence and retires the "same five dimensions / pre-agentic" framing before it is reused.
Open follow-ups
- Does Deloitte publish a separate enterprise-wide agentic readiness or maturity framework, distinct from this channel-scoped one? This asset comes from the Engineering practice and covers only digital channels. If the AI & Data or Deloitte AI Institute side has an enterprise-level model, that one - not this - is the real analog to Org Map.
- Who else assesses "operability for external agents," and is a standard forming? If agent-readability of an organization's surfaces is becoming a scored dimension across vendors (Deloitte names MCP explicitly), the criteria set is worth knowing before OI defines its own version.
- What are levels 2-4 of Deloitte's five-level scale, and are per-capability criteria published anywhere? Only "Initial" and "Optimized" are named. Without criteria the scale is a label, and that tells us how much rigor a Big-4 firm considers sufficient to publish.
Related
- [[2026-07-15-agentic-assessment-framework-competitive-landscape]] - the parent brief carrying the Deloitte claim; a dated correction flag was appended there on 2026-11-11
- [[2026-09-26-big4-agentic-readiness-frameworks-primary-verify]] - the sibling verification of the McKinsey, Accenture and KPMG claims; this brief completes its sweep
- [[2026-06-28-snowflake-si-cortex-positioning-caf-gap]] - Deloitte's platform-and-module positioning, the other place Deloitte appears in the vault
- [[2026-05-21-enterprise-ai-agent-deployment-paths]] - places phData a tier below the Big-4, which is why these frameworks are citation material rather than bid competition
Sources
Vault:
- [[2026-07-15-agentic-assessment-framework-competitive-landscape]]
- [[2026-09-26-big4-agentic-readiness-frameworks-primary-verify]]
- [[2026-06-28-snowflake-si-cortex-positioning-caf-gap]]
- [[2026-05-21-enterprise-ai-agent-deployment-paths]]
Web (primary-verified - Deloitte's own page and PDF, retrieved 2026-11-11):
- Deloitte US, "Preparing digital channels for agentic AI evolution" (browser title "An executive's framework for AI capability assessment"), Perspective, pub. 2026-05-05, authors Muthu / Ramadas / Ghosh / Deshpande - three dimensions; zero instances of "strategic integration," "maturity," "stage" or "level" in the page body: https://www.deloitte.com/us/en/services/consulting/articles/executive-framework-for-ai-capability-assessment.html
- Deloitte, "Agentic AI capability assessment for digital channels," PDF, May 2026, 15 pages, ungated - 15 capability areas across three dimensions (Figure 4); five-level maturity scale, Level 1 Initial to Level 5 Optimized, and the align / identify / assess / adoption-roadmap flow (Figure 5): https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/executive-framework-for-ai-capability-assessment.pdf
Claims left unquotable (do NOT cite to clients):
- Deloitte's framework "staging a progression from experimentation to strategic integration" - phrase absent from both the page and the PDF
- Deloitte as one of four firms "scoring the same five-ish dimensions (strategy, data, technology, organization/talent, governance)" - contradicted by Deloitte's own three dimensions