Messaging Audit: The Copilot Cartel

Section 1 — Executive Summary

The 10 Most Overused Words in AI Branding

Frequency measured across a 20-brand cohort of AI-native and enterprise incumbent brands:

1. Agent / Agentic — Universal. Used by ServiceNow, Salesforce, Workday, Microsoft, IBM, SAP, Oracle, Cohere, Anthropic, OpenAI, Google, Mistral. ~16 of 20 brands.
2. Copilot — Common, locked by Microsoft but borrowed by SAP (“Joule, the AI copilot”), Perplexity (“Perplexity Copilot”), and dozens of vendor partner pages. ~10 of 20.
3. Intelligence / Intelligent — Universal. “Personal intelligence” (Pi), “Now Intelligence” (ServiceNow), “Joule” (SAP), watsonx, Cohere (“aggregator of intelligence”).
4. Trusted / Trust — Universal in enterprise tier. ServiceNow (“trustworthy, human-centred”), IBM (“trustworthy AI”), Salesforce (“trust”), Workday (“responsible AI”), Cohere (“security-first”), Adobe (“commercially safe”).
5. Platform — Universal. ServiceNow, Workday, Salesforce (“digital labor platform”), IBM (“AI and data platform”).
6. Transform / Transformation — Universal in incumbent tier. ServiceNow’s old core tagline; Workday (“transform how people work”); Salesforce (“transform their business”); SAP.
7. Frontier — Common AI-native marker. Mistral (“Frontier AI. In your hands”), Cohere (“Frontier language models”), Anthropic, OpenAI, Google DeepMind.
8. Augment / Amplify — Common. Anthropic (“amplify human capability”), Salesforce, Workday (“augment human workers”), Inflection.
9. Autonomous / Autonomy — Common, accelerating. ServiceNow (“Autonomous Workforce”), Salesforce (“autonomous agents”), Workday, Oracle.
10. Workflow / Workflows — Universal in B2B. ServiceNow (workflows are the noun their brand was built on), Salesforce, SAP, Workday, IBM, Microsoft, Cohere.

The 5 Most Distinctive Linguistic Moves Still Working

1. Anthropic — “Keep thinking.” (Mother, launched September 18, 2025, multi-million dollar campaign.) Repositions AI as cognitive amplifier for problem solvers, not productivity tool.
2. Perplexity — “Where knowledge begins.” Treats AI not as answer-machine but as the start of inquiry. Owned vocabulary: “answer engine,” “Know it all.”
3. Midjourney — “New mediums of thought and expanding the imaginative powers of the human species.” No metrics, no enterprise language. Pure mission.
4. Salesforce — “Digital labor” (Benioff, Agentforce 2.0 launch event, December 17, 2024). Reframes from “AI tools” to a “limitless workforce” with $7 trillion economic upside.
5. ServiceNow — “AI control tower” + “completion problem” (Otto launch, May 5, 2026). First enterprise brand to define the category by what doesn’t work in competitors’ AI.

Vocabulary Opportunity for Infrastructure Incumbents

The category has saturated the vocabulary of capability (intelligence, frontier, agent), partnership (copilot, assistant, companion), and transformation (reshape, reinvent, revolutionise). It has barely touched the vocabulary of operational consequence: what actually happens when the AI runs, breaks, holds, governs, finishes. This is ServiceNow’s white space.

Key Recommendation

ServiceNow should stop fighting on the “agent” battlefield and reposition as the operating system of consequence — owning the language of work that completes, holds, governs, and runs. The “AI control tower for business reinvention” descriptor and the “completion problem” frame are the foundation.

Section 2 — The Category Dictionary

Semantic Field A: Intelligence / Capability

Semantic Field B: Safety / Trust

Semantic Field C: Partnership / Augmentation

Semantic Field D: Transformation / Magic

Semantic Field E: Infrastructure / Operations

Semantic Field F: Human / Emotional

Section 3 — Metaphor Map

Metaphor 1: AI as PARTNER

copilot / assistant / companion / collaborator

The dominant metaphor. Microsoft owns “Copilot” as a brand mark, but the word has been linguistically diluted — SAP calls Joule “the AI copilot,” Perplexity ships “Perplexity Copilot,” and third-party reviewers describe everything from watsonx Assistant to Cohere North as “copilots.”

Per The Information (December 28, 2025), Nadella wrote that the integrations connecting Copilot with Gmail and Outlook “for the most part don’t really work” and are “not smart.” When the originator critiques the product, the partner metaphor wobbles at the source.

Status: Dying metaphor at the linguistic level even as Microsoft still owns the trademark. Anthropic’s “thinking partner” is the most distinctive partnership reformulation.

Metaphor 2: AI as AMPLIFIER

augment / amplify / supercharge / unlock

Universally used, universally dead. Salesforce’s Benioff says Agentforce “amplifies and augments human capabilities.” Anthropic says Claude “amplifies human capability rather than replacing it.” Workday says agents “augment human workers.” When everyone says “augment,” no one differentiates on augmentation.

Status: Dead metaphor

Metaphor 3: AI as FRONTIER

frontier / cutting-edge / next-gen / breakthrough

Mistral has built almost its entire brand on “Frontier AI. In your hands” — a deliberate Europe-vs-US, accessibility framing. Cohere markets “frontier language models.” OpenAI uses “frontier” in regulatory contexts.

Status: Contested cliché in AI-native tier; legacy incumbents have wisely largely avoided it.

Metaphor 4: AI as AGENT / WORKER

agent / agentic / autonomous / workforce / digital labor

The defining metaphor of 2025–26 and the most exhausted. Gartner formally coined the term “agent washing” in its June 25, 2025 press release.

Anushree Verma, Gartner Senior Director Analyst: “The rebranding of existing products, such as AI assistants, robotic process automation (RPA) and chatbots, without substantial agentic capabilities” — warning that “most agentic AI projects right now are early stage experiments or proof of concepts that are mostly driven by hype and are often misapplied.”

Keith Kirkpatrick, Futurum Group VP & Research Director: “Buyers are increasingly fatigued by ‘claims wars’ and superlatives. In 2026, procurement teams will reward vendors that can demonstrate revenue expansion, margin improvement, or scaled productivity gains — not just task-level efficiency.”

Josh Bersin observed in May 2026 that “companies aren’t doing a good job of ‘naming’ their agents yet.”

Despite this, Salesforce’s “digital labor” remains the strongest sub-metaphor — Benioff explicitly drew the analogy to “digital payments, digital images, digital twins, digital media” and floated a $7 trillion economic upside.

Status: Contested territory, no clear winner, fatigue setting in.

Metaphor 5: AI as INFRASTRUCTURE

platform / backbone / foundation / fabric / layer / control tower

This is the metaphor that has been underused by AI-natives and is being rapidly claimed by infrastructure incumbents — most aggressively by ServiceNow. The Otto launch positions ServiceNow as “the AI control tower for business reinvention” (May 5, 2026). Workday competes with “Agent System of Record” / “central command centre.” IBM repositions watsonx as “industrial-grade AI infrastructure.”

The infrastructure metaphor is where the brand war will be won or lost in 2026–27. It’s the only metaphor family where AI-natives don’t have first-mover advantage.

Metaphor 6 (emerging): AI as COGNITIVE EXTENSION

thinking / knowledge / curiosity / mind / consciousness

Owned by AI-natives in the creative/consumer tier. Anthropic (“Keep thinking”), Perplexity (“Where knowledge begins”), Midjourney (“new mediums of thought”), Runway (“Tools for human imagination”). Almost completely absent from enterprise incumbent vocabulary.

Available Metaphor Families Nobody Is Using Well

• AI as utility (electric grid, plumbing, running water) — IBM hints at this but doesn’t own it
• AI as terrain (we run the rails, we hold the floor, we are the ground) — wide open
• AI as ritual / discipline (the practice of work, the craft) — wide open, only Adobe gestures
• AI as physics / law (gravity, force, friction) — wide open

Section 4 — The Cliché Report Card

Top 20 terms ranked by decay rate across the 20-brand cohort:

Section 5 — Distinctive Language That Works

Anthropic — “Keep thinking”

The specific language: “Keep thinking” tagline, launched September 18, 2025 in collaboration with creative agency Mother (media: Initiative); first paid campaign in Anthropic’s history, described as a “multimillion-dollar media investment.”

Andrew Stirk, Head of Brand Marketing: “‘Keep thinking’ is intended as both a rallying cry and a promise: to the industry, that we must build AI responsibly; to problem solvers everywhere, that what once seemed impossible is now within reach. Claude is for those who see AI not as a shortcut, but as a thinking partner to take on their most meaningful challenges.”

Why it works: Inversion. While the rest of the category sells AI as a shortcut (productivity, automation, faster outputs), Anthropic positions Claude as more thinking. It resolves the cultural anxiety about AI making people dumber — exactly the discomfort the rest of the category is silently avoiding.

Technique to learn: Find the cultural anxiety your category is creating and address it head-on in the brand line.

Perplexity — “Where knowledge begins”

The specific language: “The most powerful answer engine. Powering curiosity with answers backed by up-to-date sources. This is where knowledge begins.”

Aravind Srinivas at Stanford GSB: “There is no end to knowledge. That’s why at Perplexity the tagline is ‘where knowledge begins.’”

Why it works: Two moves. First, the invention of a category (“answer engine”). Second, epistemic humility — the tagline doesn’t claim to have knowledge, only to start it. In a category where every competitor is promising omniscience, Perplexity claims modesty.

Technique to learn: Invent the category noun, then position your brand at the threshold of it, not the centre.

Midjourney — “New mediums of thought”

The specific language: “An independent research lab exploring new mediums of thought and expanding the imaginative powers of the human species.”

Why it works: No metrics. No enterprise. No customers. No productivity. It is the only brand in the cohort that resists the operational language entirely and operates as a mission. The phrase “mediums of thought” is deeply original — thought has mediums, the way painting has oil and watercolour.

Technique to learn: When the entire category is talking ROI, the contrarian move is to talk meaning.

Salesforce — “Digital labor”

Benioff, Agentforce 2.0 launch event (December 17, 2024, St. Regis Hotel, San Francisco): “Agentforce 2.0 cements our position as the leader in digital labor solutions, allowing any company to build a limitless workforce.” And: “We’re not just supplying information management tools anymore. We’re now providing digital workers.”

Why it works: A noun, not a feature. “Digital labor” is a market category with implied TAM. It reframes the buyer’s question from “does this AI work?” to “should we hire human or digital labour?” — a much bigger conversation.

Technique to learn: Find the buying conversation that is two layers up from your product. Name it.

ServiceNow — “AI Control Tower” + the “completion problem”

ServiceNow, May 5, 2026 Otto press release: “Enterprise AI has a completion problem. Other major software providers ship AI inside their own applications, working in compartmentalized isolation, unable to complete work across departments or systems.”

Bhavin Shah, SVP/GM Employee Experience & AI (formerly Moveworks founder): “Moveworks understood what employees needed. ServiceNow could do the work. Together, we built ServiceNow Otto, an AI experience that completes work, across any system, department, or any workflow.”

Why it works: It names a problem nobody else has named. Every other vendor competes on capability. ServiceNow alone names the gap between capability and outcome. “Completion” is a strong, plain, operational verb — it isn’t AI vocabulary, it’s work vocabulary.

Technique to learn: Categories converge on capability claims. The breakaway move is to name the operational consequence.

Honourable Mentions

• Mistral — “Frontier AI. In your hands” — clean, accessible, sovereign-flavoured. Works for European positioning.
• Adobe — “commercially safe” — turns a regulatory headache into a value proposition.
• IBM — “watsonx your business” — verbing the brand. Underrated.
• Workday — “Agent System of Record” — strong category-naming move; less poetic but defensible.

Section 6 — The Incumbent Borrowing Analysis

ServiceNow (focal case study)

What they have adopted: “AI Agents,” “agentic AI,” “Autonomous Workforce,” “Now Assist,” “AI Control Tower,” “completion problem,” “AI platform for business transformation” → “AI control tower for business reinvention.” Notably, ServiceNow has not adopted “copilot,” “frontier,” or “augment” as core brand language — a discipline most of the cohort has lost.

How it lands: Mixed but improving. The early-2024 positioning was generic and indistinguishable from Salesforce’s, Workday’s, and SAP’s. The 2025–26 pivot to “AI control tower” + “completion problem” + “Otto handles the rest” represents the first genuinely distinctive positioning ServiceNow has held since the original “workflows” claim.

McDermott at Knowledge 2025: “We are going to bring AI agents to every corner of your business on one platform, one architecture and one data model… We are putting AI to work for people.”

McDermott on Q3 2025 earnings (October 29, 2025): elevated ServiceNow into the “Super Eight” (Mag 7 plus ServiceNow), noting “our AI Control Tower deal volume more than quadrupled quarter over quarter in Q3.”

Cost to brand equity: Limited so far. The risk is dilution: if “Otto,” “Autonomous Workforce,” “AI Control Tower,” “Now Assist,” “Workflow Data Fabric,” and “AI Agent Fabric” all coexist, the architecture becomes a museum, not a brand.

Borrowed vs. earned: Earned: “workflow,” “control tower,” “completion problem,” “Otto handles the rest.” Borrowed: “agent,” “agentic,” “transformation,” “reinvention,” “human-centred.”

Salesforce

Adopted: “Agent,” “Agentforce,” “autonomous AI,” “digital labor,” “limitless workforce,” “trust,” “data 360.”

How it lands: “Digital labor” is the most credible reframing in the incumbent tier. The Agentforce branding sprawl has triggered visible community fatigue. Benioff is fully committed — “everything needs to become about Agentforce” — which is both the greatest strength and the greatest risk.

Cost: Salesforce’s original equity (“No Software,” “the customer success platform”) has been almost entirely replaced by “Agentforce.” If “agent” becomes commodity, Salesforce’s brand will move with it.

Microsoft

Adopted: “Copilot” (originated), “AI companion,” “agents,” “Agent 365,” “Microsoft AI,” “digital workers.”

How it lands: Microsoft originated “Copilot” as a trademark in 2023 and effectively defined the partnership metaphor. But the brand has begun to weaken: Nadella’s internal critique suggests Microsoft itself doesn’t believe in its own AI companion narrative. The shift to “AI companion” (Suleyman, 2025) is a softening.

Cost: Microsoft’s incumbent equity in productivity is not threatened. But Copilot itself, as a brand, is showing wear.

SAP (Joule)

Adopted: “Copilot” (explicitly — “the AI copilot for SAP”), “Joule Assistants,” “Joule Agents,” “agentic AI,” “Joule Studio.”

How it lands: Naming the assistant “Joule” was strong — a physics unit, owned, distinctive. But framing it as “the AI copilot” immediately collapses the distinction by accepting Microsoft’s category language.

Cost: SAP gives up the linguistic high ground it created with the Joule name. The integration with Microsoft 365 Copilot reinforces dependency, not differentiation.

Workday (Illuminate / Sana)

Adopted: “Illuminate,” “Agent System of Record,” “agentic AI,” “digital workforce,” “blended workforce,” “AI Flex Credits,” “Sana.”

How it lands: “Agent System of Record” is the strongest infrastructure naming move by any HR/finance incumbent. It reuses Workday’s existing equity and extends it to a new category. However, the Sana rebrand creates architectural sprawl.

Cost: The risk is the Sana/Illuminate/Agent System of Record/Galileo overlay becoming incoherent.

IBM (watsonx)

Adopted: “Watsonx your business,” “AI built for business,” “trusted, targeted, empowering,” “responsible AI,” “agents,” “Granite,” “Orchestrate.”

How it lands: IBM has the strongest governance and trust equity in the cohort. The “watsonx your business” verbing is rare and effective. But IBM suffers from a credibility hangover from the original Watson era.

Cost: Medium. IBM’s strongest play is to lean into “AI plumbing” / “scaffolding” / “substrate” vocabulary — currently white space.

Adobe (Firefly)

Adopted: “Generative AI,” “creative generative AI,” “commercially safe,” “Content Credentials,” “Firefly.”

How it lands: Adobe owns “commercially safe” — a unique and high-value claim grounded in real legal indemnification. Adobe took the AI-native vocabulary and added one word (“safe”) that an AI-native couldn’t credibly claim.

Cost: Lowest in the cohort. Adobe has used AI to extend its existing equity rather than abandon it for borrowed glamour.

Oracle

Adopted: “AI agents,” “agentic workflows,” “AI database,” “SaaS-apocalypse” (Ellison, 2026), “thousands of live AI agents.”

How it lands: Oracle is the least visible brand in the audit because Ellison’s vocabulary is engineering rather than brand. The “SaaS-apocalypse” framing is the most distinctive 2026 line but it’s defensive.

Cost: Low. Oracle’s brand never relied on borrowed AI vocabulary; it relied on database and infrastructure equity.

Section 7 — Vocabulary White Space and the ServiceNow Architecture

Frameworks Applied

Ehrenberg-Bass / Romaniuk: The cohort has built strong category entry points but few distinctive brand assets. Romaniuk’s Distinctive Asset Grid (Fame × Uniqueness) shows most enterprise AI brands score high on Fame but low on Uniqueness. The exceptions are Anthropic (Keep thinking / Constitutional AI), Adobe (commercially safe), and Salesforce (digital labor).

Byron Sharp: Mental availability requires distinctive vocabulary. The category has saturated buyer minds with “agent,” “copilot,” “intelligence” — but these now signal the category, not any brand. The 95-5 rule applies: 95% of B2B buyers aren’t ready to buy AI infrastructure right now, but they are forming category associations.

Mark Ritson — the “zigging machine” thesis: The category zags on agents, copilots, frontier models. The brand that zigs — on completion, on consequence, on operational truth — wins disproportionately. ServiceNow’s “completion problem” is a textbook zig.

Adam Morgan — challenger: Anthropic and Perplexity have run pure lighthouse-identity challenger plays. ServiceNow has the rare option to play both incumbent and challenger.

The Available Territories Nobody Owns

1. The vocabulary of CONSEQUENCE: “works,” “runs,” “holds,” “completes,” “delivers,” “finishes,” “carries,” “stands up” — these are the words enterprise software actually does. ServiceNow’s “completes” and “completion problem” is the only foothold.
2. The vocabulary of ACCOUNTABILITY: “answer for it,” “we hold the receipts,” “audit trail,” “governed” — when AI fails (per MIT NANDA, 95% of pilots delivered no measurable P&L impact), brands that built equity in accountability survive.
3. The vocabulary of OPERATIONAL TRUST: “don’t break,” “stays up,” “holds the floor,” “won’t drop the work” — utility vocabulary. Currently used by no enterprise AI brand.
4. The vocabulary of CRAFT / DISCIPLINE: “the practice of work,” “how it gets done,” “the craft of running a company” — only Adobe gestures at this.
5. The INVERSE white space: Boring. Slow. Quiet. Backstage. Plumbing. No brand is willing to call themselves boring — but in an era of AI hype fatigue and 95% pilot failure, “we’re the boring one that works” is genuinely contrarian and ownable.

A Proposed Vocabulary Architecture for ServiceNow

The strategic move: Stop competing on “agents” and start competing on operational consequence.

Pillars of the Architecture

1. Replace “transformation” with “consequence.” Other brands transform; ServiceNow runs the consequences. Retire “transformation” entirely.
2. Refuse the “agent” battlefield. Let Salesforce and Workday fight over who has more agents. Position ServiceNow as the layer that holds them all.
3. Industrialise “completion.” The “completion problem” should appear in every deck, every earnings call, every analyst day. McDermott’s verbal habit should evolve to “we finish the work.”
4. Lean into “boring infrastructure.” In a market drowning in AI superlatives, the contrarian move is to be the brand that says: “We don’t do magic. We do work.”

Recommendations (Staged, with Thresholds)

Stage 1 (0–6 months): Discipline the Existing Vocabulary

• Retire “transformation.” Replace with “reinvention” (already in 2026 boilerplate) or, better, “consequence.”
• Audit every customer-facing document for “agent,” “agentic,” “frontier,” “augment,” “copilot.” Where the word does not add specific meaning beyond category signalling, cut it.
• Codify the “completion problem” as the canonical sales frame. Train field sales on the exact language: “Other AI brings intelligence. ServiceNow brings the work that finishes.”
• Trigger to escalate: If McDermott’s next earnings call still leads with “agents” rather than “completion,” the brand discipline isn’t holding.

Stage 2 (6–18 months): Industrialise the New Vocabulary

• Launch a brand campaign on “Work that finishes” (or equivalent). The model is Anthropic’s “Keep thinking” — a multi-million-dollar paid debut around a single contrarian word. ServiceNow’s word is finishes or completes.
• Recast McDermott’s keynote architecture around three nouns: workflow, completion, control. Avoid “agent” as a noun in the opening 5 minutes of any keynote.
• Lock the operational vocabulary (runs, holds, governs, finishes, stands up, won’t drop) into every product page, with retired words on a blacklist.
• Trigger to escalate to Stage 3: If a credible competitor publishes a “completion” or “control tower” line, accelerate the campaign.

Stage 3 (18–36 months): Inhabit the Inverse White Space

• Run a “We don’t do magic. We do work.” campaign — explicitly contrarian, explicitly anti-hype, explicitly post-Gartner-fatigue.
• Position against the 95% AI-pilot-failure narrative (MIT NANDA “The GenAI Divide,” August 2025). The pitch: “Why ServiceNow’s AI works when others don’t.”
• Threshold for kill: If 24-month brand-tracking data does not show ServiceNow’s distinctiveness scores climbing on the “completes work” / “operational consequence” attribute, the strategy needs re-evaluation.

Competitive Defensibility

Who else could do this? Workday could compete on “Agent System of Record” — but it’s HR/finance-bounded. IBM could compete on “scaffolding” / “plumbing” — but the Watson hangover limits credibility. Oracle could compete on “engineered to work” — but Ellison’s vocabulary is engineering, not brand.

What makes it hard to copy? ServiceNow is the only brand that owns workflow as its primary equity and control tower as its emerging equity. To copy this, a competitor would have to give up its own existing position.

Defensibility window: ~12 months before serious copying attempts; ~24 months before genuine erosion.

Adversarial Challenge — Stress-Testing the Top 3 Findings

Finding 1: “Agent” vocabulary has reached commodity status.

A skeptical CMO might argue: Maybe “agent” isn’t a cliché yet — maybe it’s becoming the category-defining word the way “cloud” did in 2010, and the brands that don’t use it will look out of step.

Counter-counter: “Cloud” became infrastructure-defining because it described a deployment model. “Agent” describes a fantasy of autonomy that 95% of pilots don’t deliver (MIT NANDA, August 2025). The MIT findings and Gartner’s “agent washing” warning suggest the word is decaying faster than “cloud” did.

Status: defensible.

Finding 2: ServiceNow’s “AI control tower / completion problem” is the strongest enterprise positioning.

A skeptical CMO might argue: “Control tower” is a borrowed metaphor (supply chain software) and “completion” is too operational to be brand-defining.

Counter-counter: That “control tower” is already industrial vocabulary is an advantage — buyers parse it instantly. And “completion” would be too operational for an aspirational brand, but it’s exactly right for a workflow brand.

Status: defensible, but watch for the cliché lifecycle.

Finding 3: The “vocabulary of consequence” is white space.

A skeptical CMO might argue: If it were truly white space, someone would already be there. Maybe it’s empty because customers don’t reward the boring brand.

Counter-counter: The category is still 18 months from genuine AI fatigue (CIOs are still in the FOMO phase). The brands that pre-position in the “boring works” territory will own it when the fatigue arrives. Apple’s “It just works” became a $3T brand asset precisely because Microsoft and Google were still racing on features.

Status: defensible, but UNVALIDATED at 6-month horizon — this is a 2027 play, not a 2026 play.

Reimagination — The Move Nobody Has Made

Deconstructing the category to first principles: every AI brand currently sells capability (what the AI can do). The category-redefining move would be to sell non-capability — the things the AI won’t do, can’t do, and shouldn’t do.

A brand that publicly committed to “We will not pretend the agent finished when it didn’t” would create the first credibility moat in a category currently drowning in overpromise.

For ServiceNow, the unsaid line is: “We won’t tell you it’s done until it’s done.” That is the actual brand promise of the workflow company — and no AI brand in the cohort can credibly say it.

Convert this into the next advertising platform and ServiceNow ends up where Volvo ended up with “safety” — the brand that owns the one thing the category doesn’t want to talk about.

Pattern Recognition — The Meta-Pattern

Across the entire cohort, the brands that are converging linguistically are diverging strategically. They all sound the same on the homepage, but their actual product bets are increasingly divergent:

• Anthropic on safety
• OpenAI on consumer reach
• Salesforce on labor displacement
• ServiceNow on workflow infrastructure
• Workday on HR data
• IBM on regulated industries

The implication: the language is lying about the strategy. The first brand to align its vocabulary with its actual differentiated strategy will create disproportionate brand equity.

ServiceNow’s strategy is workflow consequence; its old language (“transformation”) was generic; its new language (“completion problem”) is finally telling the truth. That is the move to industrialise.

Caveats and Limitations

• Public materials only; no proprietary research, no customer surveys, no share-of-voice analytics.
• The vocabulary counts are evidence-based but not statistically rigorous; “Universal,” “Common,” “Rare,” “Unique” are analytical categorisations from observed brand materials, not a corpus-linguistics study.
• The “completion problem” frame is fresh (May 2026) and has not yet been tested in market over multiple quarters.
• “Agent fatigue” is currently an analyst-class observation (Gartner / Verma, Futurum / Kirkpatrick, Bersin); CMO-level surveys confirming buyer fatigue lag analyst commentary.
• The MIT NANDA “95% pilot failure” figure derives from a relatively small base (52 interviews, 153 leaders, 300 public deployments) and should be treated as directional rather than dispositive.
• The OpenAI homepage in 2026 does not carry a single hero tagline; the absence of a tagline is itself a positioning choice and limits direct comparison.
• Brand assessment is necessarily subjective at the margins; the ranking of distinctive language uses qualitative judgement against named frameworks but cannot be reduced to a single defensible number.

Self-Enforcement Checklist

1. Would a CMO at a $13B company find this USEFUL? Yes — specific phrases to retire, specific phrases to claim, staged 36-month roadmap with kill thresholds.
2. Is every claim evidenced? Substantially yes; primary quotes from Anthropic, ServiceNow, Salesforce, Gartner, MIT NANDA, Futurum, Bersin, The Information are dated and sourced.
3. Named frameworks applied? Yes — Ehrenberg-Bass, Romaniuk, Sharp (95-5), Ritson (zigging), Adam Morgan (challenger lighthouse).
4. Adversarial challenge of top findings? Yes — all three stress-tested; Finding 3 flagged as UNVALIDATED at 6-month horizon.
5. At least one genuinely new insight? Yes — “sell non-capability” is a category-violating frame nobody has run.
6. Would Mark Ritson call this rigorous? He would push harder on frequency counts (flagged in caveats) and demand customer-side evidence of fatigue.
7. Better than Wolff Olins or Landor? Sharper on strategic recommendation and more honest about limitations than typical agency work.

Provenance