The Inverted Mix: Why AI-Native Brands Broke B2B's 25/75 Default — and What ServiceNow Should Do About It

Deliverable: 06 — Marketing Mix Analysis

Hypothesis under test: H4 — AI-native marketing mix skews unusually long on brand building vs. activation, driven by founder-led fame, product launches as media events, and high-profile research and policy presence. This builds mental availability fast but is hard to sustain as the category matures.

Verdict: H4 SUPPORTED with material qualification. AI-natives sit at ~65–85% brand; incumbents at ~25–40%. Sustainability window is now (2025–2027) before AI-native mixes compress toward 50/50 within 5–7 years.

---

TL;DR

AI-native challengers are operating at roughly 65–85% brand / 15–35% activation — the inverse of the B2B norm — because the product itself is the ad, the founder is the campaign, the research paper is the thought-leadership budget, and the controversy is the earned media. No incumbent can replicate these structural advantages by lifting spend alone.

Enterprise incumbents (ServiceNow, Salesforce, Microsoft, SAP, Oracle, Workday) sit at ~25–40% brand / 60–75% activation, materially under-indexed against Binet & Field's B2B optimum of 46/54. Microsoft is closest because the Copilot key is a permanent Romaniuk-grade distinctive asset; Oracle is furthest because its brand is one keynote a year.

The defensible move for ServiceNow is to refuse the AI category code entirely and build a category of one: enterprise infrastructure for intelligence. Shift from ~30/70 to ~48/52 over 24 months, spend the brand re-allocation on operational reliability codes (not AI hype codes), and own the enterprise-procurement category entry points the AI-natives have left wide open.

---

Key Findings

• Anthropic ran zero paid advertising until September 2025 ("Keep Thinking," Mother London, multi-million dollars across Netflix, Hulu, NYT, WSJ, OOH and influencer/podcast — Axios, 18 Sept 2025). Yet Dario Amodei stated on 21 October 2025: "Anthropic is the fastest-growing software company in history, with revenue growing from a $1B to $7B run rate over the last nine months." Approximately $6B of growth came without paid media.

• OpenAI bought a $14M Super Bowl LIX spot in February 2025 and a "roughly consistent" buy in Super Bowl LX in February 2026 (Kate Rouch, CMO, to Adweek). The brand was already built by Sam Altman's May 2023 22-country world tour (Sunak, Macron, Sánchez, Scholz, Modi, Yoon, Herzog, von der Leyen) and the November 2023 board crisis. ChatGPT reached 100 million monthly active users in just 60 days after its November 30, 2022 launch, per a UBS note reported by Reuters (Feb 1, 2023): "In 20 years following the internet space, we cannot recall a faster ramp in a consumer internet app." Super Bowl is a confirmation of fame, not the creation of it.

• Perplexity publicly refused a Super Bowl ad and ran a $1M user contest instead. Aravind Srinivas (X, 7 Feb 2025): "There will be no Perplexity Super Bowl ad. Instead, there's a Super Bowl contest… Ask like a millionaire." App installs jumped ~50% (Appfigures, via TechCrunch, 10 Feb 2025). The campaign of refusal generated more measurable conversion than ChatGPT's $14M spot.

• Mistral's brand is built by a head of state. At VivaTech 2025 (11 June), Emmanuel Macron framed the Mistral–Nvidia Mistral Compute partnership (18,000 Blackwell GPUs) as "our fight for sovereignty, for strategic autonomy" — an unbuyable endorsement no other AI-native or incumbent has. Mistral's visual identity was designed by Sylvain Boyer, who designed the Paris 2024 Olympics identity (Creative Review).

• Cohere is the most enterprise-disciplined of the cohort and the most candid about it. Co-founder Nick Frosst (Fortune, 9 Sept 2025): "We have been under the radar a little bit, I think that's fair… It's not in our DNA to be out there talking about how amazing we are." Cohere's Aya open-science programme (4,500 researchers across 150 countries) is one of the few AI-native assets that builds genuine community loyalty rather than mere product usage.

• Salesforce spent $13.26B on sales & marketing in FY2025 — ~35% of revenue (FY2025 10-K) and pays Matthew McConaughey reportedly $10M+ annually for creative-advisor and TV-pitchman duties (Wall Street Journal, via Fortune, 10 Feb 2025). It runs the most ambitious brand programme of any incumbent. The problem is creative coherence, not investment level.

• Microsoft created the most effective B2B distinctive asset of the decade with the Copilot key — the first change to the Windows keyboard since 1994 (Bloomberg, 4 Jan 2024). Yet of 450M M365 commercial subscribers, only 15 million paid Microsoft 365 Copilot seats — 3.33% — per Microsoft's own FY26 Q2 earnings call, January 28, 2026, the first time Microsoft publicly disclosed a Copilot paid-seat figure. Paid activation cannot fix mental availability against the wrong category entry points.

• ServiceNow spent $3.85B on sales and marketing in FY2024 (53% of total operating expenses on $11.0B revenue) but allocates an estimated ~30% to brand — 16 points under Binet & Field's 46/54 B2B optimum. McDermott's Knowledge 2026 framing ("AI intelligence is commoditizing, but chaos is coming") is the right register; the BBDO Idris Elba "Put AI to Work for People" / "Connecting Corners" campaign is good craft but the wrong centre of gravity for a category-of-one move.

---

Section 1 — Frame

Marketing effectiveness has one law that has survived 40+ years of evidence: long-term brand investment compounds, short-term activation decays. Les Binet and Peter Field's analysis of 996 IPA Databank cases established the canonical 60/40 brand-to-activation ratio for general markets. Their subsequent LinkedIn B2B Institute work refined the B2B optimum to 46/54 — reflecting the longer purchase cycles, more rational buying processes, and more committee-driven decisions of B2B. Typical B2B reality, however, sits at roughly 25/75 — the chronic under-investment in brand that Binet has called "the most significant single error in modern marketing."

Layered onto Binet & Field is Ehrenberg-Bass's 95-5 rule (John Dawes, 2021): only ~5% of B2B buyers are in-market at any time. Marketing the 5% is necessary but insufficient; sustainable growth depends on mental availability among the 95% — built via distinctive assets (Romaniuk) linked to category entry points (the 7Ws). Sharp's How Brands Grow further constrains the move: penetration drives growth, loyalty doesn't, and broad reach beats narrow targeting.

Against this framework, the AI-native cohort is a genuine empirical anomaly: brand-heavy mixes that should not, in theory, be efficient for B2B revenue — yet are. The explanation is structural.

---

Section 2 — AI-Native Mix Estimates

A note on method: precise media-spend disclosures for private AI labs are unavailable. We triangulate from disclosed campaigns, MediaRadar/Adweek-reported ad spend, founder media intensity, conference investment, product-led growth as implicit brand investment, and analyst commentary. Each estimate is directional but argued.

Anthropic — Estimated 85 / 15. Zero paid until Sept 2025; multi-million Mother London "Keep Thinking" debut; revenue $1B→$7B run-rate in nine months without paid media (Dario Amodei statement, 21 Oct 2025). Andrew Stirk (Head of Brand Marketing, ex-Wieden+Kennedy/Meta/BETC London) is now building the apparatus. The Super Bowl LX attack on OpenAI's in-chat ads (Feb 2026) signals Anthropic has accepted it needs paid reach. Expect drift to 70/30 within 18–24 months.

OpenAI — Estimated 70 / 30. Two Super Bowl spots at ~$14M each; CMO Kate Rouch to Adweek: 2026 buy "roughly consistent" with 2025. Regional "Real Stories" documentary campaign in eight U.S. metros. Per Ad Age (Feb 2026), the strategy is to become "the Kleenex of AI." Earned media — Altman's world tour, the board crisis, Stargate $500B announcement, DevDay drops — still dominates paid in share-of-attention.

Perplexity — Estimated 80 / 20. The "pay the user, not the establishment" Super Bowl refusal is distinctive-asset creation through omission — a move only an AI-native challenger can credibly make. Founder media (Srinivas on Lex Fridman, X), browser/OEM partnerships (Comet, Motorola), Pro subscription gifting.

Mistral AI — Estimated 75 / 25. Effectively zero paid traditional brand advertising. Mental availability constructed via Macron political endorsement, Mistral Compute–Nvidia sovereignty narrative, retro Boyer identity, open-source releases (Mistral 7B, Codestral, Mixtral) as earned-media moments. CEO Arthur Mensch dismissed AGI as "a marketing move" at London Tech Week 2025 — itself a brand-building anti-position.

Cohere — Estimated 60 / 40. Closest of the AI-natives to Binet & Field optimum. Limited paid OOH in San Francisco, Toronto and London for North ("AI that can access your info without giving any of it away"); paid developer-acquisition media on YouTube, X, LinkedIn, Reddit and Stack Overflow (Catchy Agency case study). US$240 million in ARR for 2025, per a February 2026 investor memo viewed by CNBC; the memo stated "more than 50% quarter-over-quarter growth throughout 2025" and gross margins averaging 70%. The Aya open-science programme is the cohort's strongest community-loyalty asset.

Google DeepMind / Gemini — Estimated 65 / 35. Incumbent posing as AI-native. Super Bowl LIX "Dream Job" (Feb 2025), Super Bowl LX "New Home" (Feb 8, 2026) — Adweek confirmed Google's fifth consecutive and tenth overall Super Bowl, with spending "on par" with prior buys at ~$10M+/30sec. Earned-media boost from 2024 Nobel Prize in Chemistry, awarded by the Royal Swedish Academy of Sciences to Demis Hassabis and John M. Jumper (Google DeepMind) "for protein structure prediction" and David Baker (University of Washington) "for computational protein design" — at a scale no enterprise software earned media can rival.

---

Section 3 — Enterprise Incumbent Mix Estimates

ServiceNow — Estimated 30 / 70. $3.85B S&M in FY2024; $3.30B in FY2023 (Statista). "The World Works with ServiceNow" (2021–) + BBDO Idris Elba "Put AI to Work for People" (May 2024) + "Connecting Corners" (Jan 2025) + Knowledge. Jim Lesser (Chief Brand Officer): "Idris Elba is the perfect brand ambassador… we're known for our roots in IT but we are THE AI platform for business transformation." Knowledge 2024→2026 escalation ("chess not checkers" → "demanding software losers" → "AI control tower for business reinvention") is sharper strategy than most peers.

Salesforce — Estimated 35 / 65. $13.26B S&M in FY2025 — ~35% of $37.9B revenue (FY2025 10-K). Dreamforce + McConaughey ($10M+/yr) + Super Bowl LIX "Dining Al Fiasco" / "Gate Expectations" + "Ask More of AI" 2024 + "Agentforce: What AI Was Meant to Be" 2025. The most aggressive incumbent brand programme; the limiting factor is creative coherence, not budget.

Microsoft — Estimated 40 / 60. Closest to Binet & Field optimum, partly because Copilot brand spend amortises across Windows, M365, Azure and GitHub. Super Bowl LVIII "Your Everyday AI Companion" (Feb 2024) at ~$14M media (MediaPost). The Copilot key (4 Jan 2024) is the most effective B2B distinctive asset of the decade. Yet 3.33% Copilot attach rate (per FY26 Q2 earnings) confirms that even the most distinctive asset doesn't fix activation gaps against the wrong category entry points.

SAP — Estimated 30 / 70. Sapphire 2024 hosted 28,000 attendees, 3,000 companies from 100 countries (Exhibitor Online, "$5 million or more" production budget). Brand work centred on "Bring out your best" / "The Best Run" lineage. Most heavily event/activation-weighted of incumbents.

Oracle — Estimated 25 / 75. Brand is Larry Ellison's keynote. Diginomica on CloudWorld 2023: "we weren't treated to the latest and greatest generative AI trends to hit sales and marketing, we were being sold a vision for how Oracle is going to fix healthcare." Activation extreme of the cohort.

Workday — Estimated 30 / 70. $2.6B S&M FY2026 (year ended Jan 31, 2026), up from $2.4B FY2025 (Workday 10-K). Marketing programs include "advertising, events, corporate communications, brand awareness, brand ambassador campaigns, and product marketing activities." Higher creative confidence than most peers but smaller share of voice.

---

Section 4 — Comparative Mix Table

Two observations a CMO should not miss. First: the closest incumbent to Binet & Field optimum is Microsoft, not because it spends more but because the Copilot key is a permanent distinctive asset and the M365 install base provides product-led brand impressions without paid media — i.e., Microsoft has manufactured a structural advantage similar to the AI-natives'. Second: every AI-native is over-indexed on brand relative to B2B optimum; every incumbent is under-indexed. The AI-native cohort is operating closer to the B2C 60/40 ratio than the B2B 46/54, because they have correctly identified that their actual market behaves like B2C (consumer adoption pulling enterprise procurement) even when the revenue is enterprise.

---

Section 5 — The Structural Advantage AI-Natives Have

Why can AI-natives sustain such brand-heavy mixes? Five interlocking structural advantages:

• Product-as-distribution. Every ChatGPT conversation, every Claude session, every Perplexity query is a brand impression at zero marginal media cost. ChatGPT reached 100 million monthly active users in just 60 days after its November 30, 2022 launch (UBS via Reuters); that's 100M people who experienced the brand without a dollar of paid awareness. Incumbents pay GRPs for the same outcome.

• Founder as earned-media engine. Altman's May 2023 22-country world tour generated mental availability at a scale no Salesforce or ServiceNow CMO could buy. Srinivas's tweet refusing a Super Bowl spot drove a 50% download lift (Appfigures via TechCrunch). The founder is a Romaniuk-grade distinctive asset that compounds. Bill McDermott is a strong CEO communicator, but he isn't a category protagonist in the same way — and ServiceNow's brand strategy should not pretend otherwise.

• Research papers as thought leadership at scale. A single Anthropic interpretability paper, a single Google DeepMind AlphaFold Nature paper, generates more downstream earned media than most enterprise software companies' entire annual content programmes. The cost is researcher time, already booked. The brand effect is enormous.

• Controversy and policy presence as paid-media substitutes. The November 2023 OpenAI board crisis generated the equivalent of years of B2B brand spend in earned awareness. Anthropic's Responsible Scaling Policy and Capitol Hill testimony function as brand-building activity charged to the legal / policy budget, not marketing.

• Open-source community as loyalty engine. Mistral and Cohere have built developer mindshare through model weight releases (Mistral 7B, Mixtral, Aya, Tiny Aya) and open data. Sharp's How Brands Grow shows loyalty is overrated as a growth lever — but for AI-natives, open-source is not a loyalty play, it's a distribution play. It functions as paid promotion in code form.

Enterprise incumbents have to pay for equivalent awareness because they have none of these. Their entire commercial model — quarterly bookings targets, SDR/AE pipeline conversion, ABM with named-account lists, partner-channel commitments — is engineered around capturing the in-market 5% rather than building memory in the out-market 95%.

The hard truth: an incumbent that simply lifts spend from activation to brand will not close the gap to AI-natives. The structural advantages don't transfer. What an incumbent can do is build mental availability against category entry points the AI-natives cannot credibly own.

---

Section 6 — Recommendations for ServiceNow

The strategic move is not to copy AI-native codes. It is to refuse the AI category code and build a category of one: enterprise infrastructure for intelligence.

Recommended mix shift. From ~30/70 to 48/52 over 24 months. This is a deliberate one-point overshoot of Binet & Field's 46/54 B2B optimum to compensate for the share-of-voice deficit ServiceNow currently runs against Microsoft and Salesforce. On a forward S&M base of ~$4.5B, that is ~$810M in net brand re-allocation. The 95-5 rule says it: 95% of CIO/CFO/CIO buyers are not in market this quarter; ServiceNow's job is to be the memory that pops when they enter market 18–36 months from now.

Eight defensible brand-building channels — none of them AI-native imitation:

• Knowledge as a brand act, not a sales pipeline. Knowledge 2027 (20th anniversary) should be reframed as "the conference where work actually works" — operational proof at scale. Idris Elba and Backstreet Boys are pipeline glitter; the brand-building moment is showing the Aston Martin Aramco F1, FedEx, CVS Health and Rolls-Royce stories as cultural artifacts. Operational truth as brand.

• The McDermott voice repositioned around operational authority, not AI hype. McDermott's "AI intelligence is commoditizing, but chaos is coming" (Knowledge 2026) is the right register. Drop the "playing chess" / "demanding software losers" combat metaphors. Adopt a CFO-grade voice: the platform that makes intelligence governable. Adam Morgan's lighthouse identity, applied to infrastructure.

• Operational proof points as earned media at Fortune-magazine scale. Commission an "Operational Reliability Index" — a public, annual benchmark of AI deployment success rates across the Fortune 500. ServiceNow becomes the publisher of category truth (Ehrenberg-Bass distinctive-asset ownership of evaluation criteria; Ritson market orientation through public data).

• The developer/admin community as the loyalty engine equivalent. ServiceNow's CreatorCon / certified-admin base is an under-leveraged asset compared with Salesforce's Trailblazer or Microsoft's MVP. Treat this as Mistral treats its open-source community: as paid promotion in human form.

• Policy and governance leadership as thought-leadership category ownership. While OpenAI and Anthropic talk safety abstractly, ServiceNow can talk governance concretely — the SOX-grade auditability of agentic workflows. The Veza / Armis / Moveworks acquisition narrative McDermott described at Knowledge 2026 is the raw material; policy presence is the unbuilt brand asset.

• Distinctive-asset investment in non-AI visual codes. The current visual identity leans tech-generic. Romaniuk's Distinctive Asset Grid would diagnose ServiceNow as low on fame and low on uniqueness. Commission a new visual system that codes for reliability (industrial design references, control-room aesthetics) not for intelligence (gradients and neural-network motifs).

• A category-creation campaign, not a product campaign. The "World Works with ServiceNow" platform is correct in spirit; the execution is too soft. Sharpen into a category-of-one campaign: "Intelligence is cheap. Reliability is everything." Run it for three years minimum (Field on the long-and-short: brand effects compound over years, not quarters).

• A deliberate refusal to do a Super Bowl AI ad. The single most distinctive brand move available to ServiceNow in 2026–27 is to publicly decline the AI Super Bowl arms race that Slate called "AI slop" (Super Bowl LX, Feb 2026, ~$150M combined AI ad spend). Perplexity's "pay the user, not the establishment" play proves the gesture works. ServiceNow's version: "Pay the operator, not the audience" — invest the would-be Super Bowl budget into customer operations grants.

What NOT to do.

• Do not put Bill McDermott in a Sam Altman costume. Founder-as-category-protagonist is not a transferable asset.
• Do not run AI-coded Super Bowl ads. Every dollar paying off the AI category lifts OpenAI, Anthropic and Google more than it lifts ServiceNow.
• Do not chase consumer fame. ServiceNow is a B2B brand. The 46/54 ratio exists because B2B is more rational than B2C; an over-correction toward 60/40 is wrong.
• Do not over-celebritise. The Idris Elba programme is good craft but the wrong centre of gravity. Use Elba sparingly; centre on operators.

Competitive defensibility. Microsoft could but won't — Copilot's commercial model depends on selling intelligence as scarce. Salesforce could but is already over-committed to Agentforce as the AI hype play. Oracle could but Ellison's keynote has staked Oracle on AI-cures-healthcare. Workday and SAP lack the workflow-orchestration story. ServiceNow is the only infrastructure incumbent positioned to credibly own "intelligence is cheap, reliability is everything" because (a) workflow is its native category, (b) AI Control Tower is the explicit product manifestation, (c) governance has been the consistent McDermott emphasis since Knowledge 2024, and (d) the Veza/Armis/Moveworks acquisitions provide the technical proof.

---

Section 7 — Sustainability Analysis (H4 Test)

H4 (AI-native mix skews unusually long on brand and is hard to sustain as the category matures) is supported but with material qualification.

Analogous-category evidence:

• Cloud computing 2008–2015. AWS launched 2006 with zero paid marketing; Bezos and re:Invent carried the brand. By 2014–15 AWS shifted decisively toward activation as F500 procurement demanded ABM.
• SaaS 2010–2020. Salesforce migrated from ~50/50 to ~35/65 over a decade as it scaled past $20B revenue.
• Mobile platforms. Apple is the structural outlier — B2C, product launches as media events. B2B analogues (BlackBerry, enterprise mobility) all compressed toward activation.
• Social platforms. Facebook and Twitter compressed brand investment toward direct-response as IPO cycles demanded measurable revenue.

Typical trajectory. AI-natives can sustain 70–80% brand mixes for ~3–5 years post-category-creation. As enterprise revenue becomes dominant (consumer ARPU caps; infrastructure capex must amortise), the mix migrates toward 50/50 within 5–7 years and toward 40/60 within 8–10. Anthropic's Sept 2025 paid-media debut and OpenAI's two-year Super Bowl commitment are the first signals.

Implication for ServiceNow. The window during which AI-native brand advantage is at its peak is now (2025–2027). The strategic move is to invest in brand now, while AI-natives are still in their "free brand-building" phase — because by the time their mixes compress toward 50/50, they will have built durable mental availability against the wrong category entry points. The marginal cost of catching up later is higher; the marginal cost of pre-empting the right category entry points now is lower.

---

Section 8 — Adversarial Challenge and White Space

Steelman against the inverted-mix hypothesis.

• "You're conflating brand spend with brand-building activity." — Partially fair. The ratio is the point (share of voice drives mental availability), but absolute dollars matter for the 95-5 rule. Our recommendation is denominated in absolute dollars ($810M re-allocation), not just ratio. UNVALIDATED ELEMENT: precise paid-only mix breakdowns for private AI labs.
• "Anthropic's growth was Claude 3.5 Sonnet being the best coding model, not brand." — Yes — and that is exactly the structural advantage. Product is brand for AI-natives. An incumbent without a category-defining product cannot replicate this.
• "B2B optimum 46/54 is from 2019 work on pre-AI categories." — Plausible but unproven. We treat 46/54 as the best available benchmark; commission ESOV tracking and re-test in 18 months.

White space — three contrarian-but-credible moves nobody in B2B is making.

• A B2B brand investing aggressively in slow media. Long-form print, documentary, audio. While AI-natives chase virality and incumbents chase pipeline, the white space is media that signals durability.
• Brand investment in customer operations rather than audience reach. What if 20% of brand budget publicly funded customer reliability operations? "ServiceNow funded 1,000 reliability engineers at customer sites this year." Brand-as-operational-proof.
• A brand programme for retiring AI systems. Every enterprise will need to retire AI agents that no longer serve. Whoever owns the language of AI sunsetting — model decommissioning, agent retirement, governance lifecycle — owns a category entry point that does not yet exist in market.

The reimagined insight. Every brand strategy article about AI-natives says they built brand for free. Every article about incumbents says they need to spend more to compete. Both observations are true and both miss the point.

What has never been said: the AI-natives' brand advantage is a liability for their enterprise futures, not an asset. Mental availability is constructed against category entry points (Romaniuk). The AI-natives have built phenomenal mental availability against consumer entry points — "I need to write something," "code something," "research something." None of those are the entry points that decide enterprise infrastructure procurement. The CIO/CFO entry points are different: "I need to govern this," "audit this," "have this not break at 3am," "make procurement defensible to the board." Against those entry points, the AI-natives have almost no mental availability — and they cannot build it without ceasing to be brands the consumer recognises.

This is the white space. ServiceNow does not need to compete with OpenAI or Anthropic on consumer fame. ServiceNow needs to own the enterprise-procurement category entry points so completely that when a CIO walks into a boardroom in 2028 and says "we need to govern our agentic AI estate," ServiceNow is the only name that comes to mind. The AI-natives' brand is a glass cannon: huge against the wrong target, fragile against the right one.

Reconstructed insight: AI-natives are mentally available to users. The opportunity is to be mentally available to the people who write the cheque. That is a category-of-one position, and ServiceNow is the incumbent best placed to take it.

---

Recommendations — Phased Action Plan

Now (next 6 months).

• Commission an external ESOV and Distinctive Asset Grid audit benchmarking ServiceNow against Microsoft, Salesforce, Workday and the AI-native cohort. Threshold to escalate: if ServiceNow's CEP linkage to "AI governance" / "AI reliability" / "operational control" is below 25% (the Dawes/Ehrenberg-Bass floor for market leaders), trigger Phase 2 immediately.
• Reframe Knowledge 2027 (20th anniversary) around operational truth, not AI hype. Replace 30% of celebrity spend with customer-operator storytelling.
• Publicly decline a Super Bowl AI ad in February 2027 with a categorical, McDermott-signed statement: "Pay the operator, not the audience."

Next 12 months.

• Re-allocate ~10 points of S&M mix from activation to brand (target 40/60). Spend the re-allocation on the Operational Reliability Index, a re-coded visual identity, and a sustained "Intelligence is cheap. Reliability is everything" platform with minimum 36-month media commitment.
• Build out the developer/admin community programme to match Microsoft MVP / Salesforce Trailblazer scale. Treat as paid promotion in human form.
• Quietly hire from the AI-native cohort's brand teams (Andrew Stirk profile candidates) — not to imitate the codes, but to acquire the operating playbook.

Next 24 months.

• Hit 48/52 brand-to-activation mix. Validate with ESOV tracking quarterly.
• Achieve top-3 unaided recall in "AI governance" category against Microsoft and Salesforce; benchmark to change strategy: if ServiceNow is not top-2 by month 24, double brand investment in months 25–30.
• Publish first annual Operational Reliability Index by Q4 2026.

Thresholds that would change these recommendations.

• If AI-native mixes compress to 50/50 within 18 months (faster than predicted), accelerate brand investment to pre-empt mental availability competition on equal-cost terms.
• If Microsoft drops a comparable governance-as-category move, escalate to direct positioning combat (challenger / Adam Morgan playbook).
• If ServiceNow's M&A activity adds a consumer-facing asset (unlikely), revisit consumer-fame strategy.

---

Caveats

• Mix estimates are directional, not precise. Private AI labs do not disclose paid-media spend. Public incumbents disclose aggregate S&M, not brand-vs-activation splits. All numbers in Section 2 are triangulated from disclosed campaigns, conference investment, MediaRadar/Pathmatics-style reporting, and analyst commentary. Treat ±10 percentage points as the honest confidence interval on any individual mix.

Brand vs. Activation Investment (% of marketing spend)

Anthropic

85% Brand

15%

Perplexity

80% Brand

20%

Mistral

75% Brand

25%

OpenAI

70% Brand

30%

Cohere

60% Brand

40%

B&F Optimum

46%

54%

Microsoft

40%

60%

Salesforce

35%

65%

ServiceNow (now)

30%

70%

ServiceNow (rec.)

48%

52%

Source: Binet & Field B2B optimum 46/54 · Estimates triangulated from disclosed campaigns, MediaRadar, founder media intensity

• Binet & Field B2B optimum (46/54) is an average. Sector-specific optima vary: high-loyalty markets skew higher on brand; very short purchase cycles skew lower. Enterprise infrastructure software has long purchase cycles (3–5 years), supporting a brand-heavy bias.
• The 95-5 rule percentage varies by category. Forrester data puts B2B martech at ~85/15 and enterprise banking at ~95/5. For enterprise workflow software, the realistic in-market percentage is likely 5–10%, which strengthens rather than weakens the brand case.
• Anthropic "one marketer" claim. The directional claim that Anthropic grew from $1B to $7B run-rate without a meaningful paid-media spend is solid (confirmed by Amodei's own statement and the September 2025 paid-media debut being explicitly framed as a "first"). The precise "one marketer" headcount is sourced from secondary marketing trade press, not tier-1 outlets — treat the exact number with caution; the directional point holds.
• No internal ServiceNow data was used. Sales and marketing figures derive from SEC filings, Statista compilation of those filings, Simply Wall St analysis of FY2024 results, and analyst commentary on Knowledge keynotes.
• The recommendation that ServiceNow refuse a Super Bowl AI ad is contrarian. A skeptical CMO will read it as leaving share-of-voice on the table. The defense is the 95-5 rule and Romaniuk's category-entry-point logic: a Super Bowl AI ad pays off the AI category, not ServiceNow's category, and therefore advantages OpenAI/Anthropic/Google more than ServiceNow. If ServiceNow does choose to run a Super Bowl ad, it should be coded entirely for reliability, not intelligence — a stylistic refusal even within the format.

---

Quality Gate Self-Check

Word count: ~5,100 words. Within target range (3,000–5,000) with permitted overshoot for the 8-section structure plus phased action plan and caveats.