William Toll: Blog

Where B2B marketing stands on AI in 2026, and five questions to score your team

2026-10-26T12:00:00Z

Bottom line: Nearly every marketing organization now uses AI, and few can prove it pays. In seven 2026 studies, headline numbers for AI use, spending or value range from 76% to 98%. Only 6.5% to 32% have scaled it, proven a return or reached an advanced stage. Organizations that report returns name an owner, standardize, redesign the work and track a business number.

Every marketing leader I talk with says their team is “doing AI.” Far fewer can tell me what AI added to pipeline last quarter.

Use is universal, depth is rare

This fall I reviewed more than 30 studies on AI in marketing for a CMO briefing, most of them published in 2026. Seven measured both a headline reading (use, spend or reported value) and how far teams have taken it, and all seven found the same shape. CMI and MarketingProfs found 95% of enterprise marketers use AI applications, yet only 7% call their implementation advanced and 2% call it leading. Basis found 6.5% evaluate AI investments against defined metrics and benchmarks. KPMG’s Q2 Global AI Pulse found only 7% of leaders across business functions report establishing ROI.

Spend has moved faster than results. CMOs now put an average of 15.3% of marketing budgets into AI, and only 30% report mature or fully developed AI readiness, according to Gartner’s 2026 CMO Spend Survey. In BCG’s 2026 CMO survey, 20% of B2B CMOs say their agentic marketing transformation already has significant, measurable revenue impact, against 31% of B2C CMOs.

Headline use, spend or value against depth in seven 2026 studies. Each pair uses one study's own measures; studies are not combined. Sources: Basis, KPMG Q2 Global AI Pulse, CMI and MarketingProfs, HubSpot, Comviva, Wharton-GBK, BCG. Open full size

Text version of this figure

Study	Headline measure	Headline	Depth measure	Depth
Basis	Use AI	98.4%	Measure ROI against benchmarks	6.5%
KPMG Q2 Global AI Pulse	See value	76%	Established ROI	7%
CMI and MarketingProfs	Use AI applications	95%	Advanced or leading	9%
HubSpot	Use AI	98%	Transformational (share of AI users)	10%
Comviva	Raised AI spend	90%	Can prove it	12%
Wharton-GBK	Any positive ROI	80%	Strongly positive ROI	28%
BCG	Claim transformation	96%	Leaders	32%

Hours saved is the wrong scoreboard

Most teams treat maturity as adoption: more seats, more tools, more prompts, more hours saved. That view has data behind it. Leaders in The CMO Survey credit AI with a 14.1% lift in sales productivity and a 14.6% cut in marketing overhead. CMI found 84% of enterprise marketers using AI for content say productivity has improved.

Read one line further, though. In the same CMI study, 38% report better content performance. Wharton and GBK Collective found hours saved is the KPI departments most often use to measure AI ROI, at 40%, and net-new revenue the least often, at 27%. Meanwhile about 70% of CEOs judge marketing partly on revenue growth and margin, and only about 35% of CMOs track those measures. McKinsey cites both figures from its 2025 article “The CMO’s comeback.”

Among the teams I talk with, the most mature also update their key metrics. They adjust attribution to credit influence that never shows up as a click, track prompt coverage on the buyer questions that matter most, and count citations and mentions in LLM answers next to pipeline. Measurement still trails adoption. In The CMO Survey, 41.5% of companies already use AI for generative engine optimization, and in Muck Rack’s State of PR 2026, 39% of PR professionals aren’t currently measuring GEO.

Hours saved is a fine way to start a business case. I’d never end one there, because the CFO’s next question is always what those hours bought.

What the teams getting returns do differently

Strong returns cluster around a handful of operating habits. Only one involves tools at all, and it’s about having fewer of them.

They aim AI at growth. Departments that judge AI on a growth metric report strongly positive returns 35% of the time, against 13% for efficiency-only departments (Wharton-GBK).
They standardize. Organizations on one approved platform report strongly positive ROI at 50%; those still experimenting, 9% (Wharton-GBK).
They can see the cost. Organizations with full visibility of AI costs report established ROI at 15%, against 3% for the rest (KPMG).
They put someone senior on the hook. Where the CEO is clearly accountable for AI, 14% report established ROI, against 4% elsewhere (KPMG).
They finish the redesign. Only 20% of departments meet all three of Wharton-GBK’s workflow-transformation conditions, and 67% of those report strongly positive ROI.
They own it. Among Bain’s B2B growth winners, 1% report shared or undefined AI ownership. Among laggards, 15% do.

One finding surprised me more than the rest. Wharton-GBK compared four ways of supervising AI. Where a person approves every AI action before it takes effect, departments achieved workflow transformation 16% of the time. That’s the lowest of the four, and it’s also the most common model, used by 48%. Departments running AI under real-time monitoring and guardrails reached 31%, though only 7% of departments work that way, and Wharton-GBK reads that small group as directional.

People belong at the decisions that matter. Put a gate with a qualified person on every key step and you’ve built in the workflow.

Score your team: five questions

Answer yes only for what runs in production today.

Does every AI workflow in production have one named owner and an outcome metric?
Have you rebuilt at least one workflow end to end, with handoffs between agents and people defined?
Do you report a business number, such as pipeline, conversion or win rate, next to every productivity number?
Are your positioning, proof and voice encoded once, in a source every AI tool loads?
Does a named person approve everything AI-assisted that reaches a customer?

Count your yeses. Zero or one means individual use: people work with AI on their own, and the value shows up as hours. Two means assisted workflows. Three or four means a governed system. Five means you’re building a growth engine, and your next job is funding the people who run it.

My read of the studies is that most teams will land on two or fewer, and that’s worth knowing early.

Where to start

Start with your first “no.” CMOs own questions 1 and 3, because ownership and measurement need budget authority. Product marketing leaders own question 4, the messaging source every agent repeats. They share 2 and 5.

Then pick one workflow per team where a 10% improvement would change this quarter’s results. That’s Bain’s test, applied team by team. Record a baseline before you set a target, because a target announced before the baseline becomes the only number anyone remembers.

If I were starting this quarter, I’d begin with question 4. Every agent repeats whatever messaging you hand it, so a sharp position gets scaled and a vague one does too.

What this can’t tell you

My bet for 2027 is that the edge goes to teams that encode how their best marketers think, then measure it against growth. More content won’t be the edge.

Request a copy of my CMO briefing. This post draws on my briefing, AI in B2B Marketing 2026, which gives the source and page for most figures here. Request your copy. If you ran the five questions and want a second read on where to start, send me your answers too. For the team-by-team view, see my companion post.

FAQ

How many marketing teams can prove AI pays?

Few. Basis found 6.5% evaluate AI investments against defined metrics and benchmarks. KPMG found 7% of leaders report establishing ROI, and BCG found only 20% of B2B CMOs report significant revenue impact from their agentic marketing transformation.

Is hours saved a good way to measure AI ROI?

On its own, no. Wharton and GBK Collective found hours saved is the KPI departments most often use to measure AI ROI, at 40%, while net-new revenue is used least often, at 27%. About 70% of CEOs judge marketing partly on revenue growth and margin, according to McKinsey’s 2025 research, so pair every hours number with a pipeline, conversion or win-rate number.

Sources

CMI and MarketingProfs, Enterprise Content and Marketing Trends: Insights for 2026
Basis, AI and the Future of Marketing: 2026 Edition
KPMG, Global AI Pulse Q2 2026 press release and brochure
Gartner, 2026 CMO Spend Survey press release
BCG, Moving the Agentic Marketing Transformation from Illusion to Reality
The CMO Survey 2026, Highlights and Insights, pp.22, 25
Wharton Human-AI Research and GBK Collective, The Adaptive Advantage
McKinsey, From anxiety to advantage, footnote 4, citing “The CMO’s comeback” (2025)
Bain, AI Won’t Transform Your Go-to-Market Until You Transform the Work
Muck Rack, State of PR 2026, 971 usable responses, May 14 to June 12, 2026
HubSpot, The State of AI in Marketing 2026 (vendor survey of 1,754 marketers, June 2026)
Comviva, Global CMO Survey 2026 press release (200+ retail, ecommerce and telecom executives)

AI use cases by team at a Series A or B cybersecurity software company

2026-10-26T12:00:00Z

Bottom line: At a Series A or B cybersecurity software company selling to enterprises, AI pays first where it does the research and people make the calls. Win/loss mining, account research briefs and event data cleanup are proven. Uses that act on the buyer directly, such as AI SDRs sending unreviewed email, are still early. Give each team one workflow, one human gate and one business number, and start with the evidence work.

Security buyers read vendor claims the way they read logs, line by line, looking for the entry that doesn’t add up. That’s why AI in cybersecurity marketing has to start with evidence.

Where each team starts and what it holds; the full map below adds what scales next, the human gate and the number to report. Ratings: my judgment, drawn from 2025 and 2026 studies and conversations with 11 product marketing leaders in Q3 and Q4 2026. Open full size

Why a Series A or B cybersecurity software company is a special case

Picture an enterprise cybersecurity software company a year or two past its Series A or B. Marketing fits around one table, the pipeline target doesn’t, and every buyer has an engineer who will test each claim against the product.

That buyer does a lot of homework before talking with anyone. In a Ponemon Institute and NOLA Marketing study of 320 enterprise cybersecurity decision-makers, more than a third used AI somewhere in product selection. Peer recommendations ranked highest in vendor discovery, at 55%, and research and survey reports had the strongest direct impact of any asset type, at 46%. Gartner found 67% of B2B buyers prefer a rep-free experience. In a Gartner survey of 645 buyers fielded over the same period, 69% prefer to validate AI-generated insights with a sales rep.

Read together, buyers want to skip the rep until they need a fact checked. That moment decides a lot of security deals. AI only helps if the rep’s answer matches what the buyer already read in an LLM answer.

So a Series A or B team works under four constraints at once: few people, a skeptical technical buyer, long cycles and analyst evaluations that shape enterprise shortlists. Those constraints decide which AI work pays.

How I rate readiness

I use three ratings. Proven means the use case runs in production at a meaningful number of companies, with published results or practitioner accounts behind it. Scaling means it’s past pilots at a growing share of companies. Early means mostly pilots, or vendor claims without independent results.

The map

Team	Start here (proven)	Next (scaling)	Hold (early)	Gate a person owns	Report
Product marketing	Win/loss and call mining for message evidence; answer-first content for SEO, AEO and GEO drafted from approved claims	Competitive intel agents that refresh battlecards	Synthetic personas in place of buyer research	PMM approves positioning and every claim; an SE or product manager signs off on any product truth not already in the claims ledger or a support documentation MCP	Win rate; message adoption in calls; citations on tracked prompts
PR	Reporter and outlet research; first drafts from an approved fact base	Executive commentary on your research placed in high-authority outlets, tracked for LLM citations	Unreviewed replies to reporters	Whoever owns press signs every external word	Pickups that name your research; citations on tracked prompts
Analyst relations	Briefing prep and inquiry notes	Evaluation and questionnaire drafts from approved claims	Evaluation answers sent without product sign-off	AR lead and product sign each answer	Analyst references; evaluation inclusion
Demand generation	Account and lead scoring on stacked intent signals; ad and email variant tests	An AI search sprint on the top 50 buyer questions	Nurture agents for leads below MQL	Marketing ops approves scoring changes	Pipeline; cost per qualified lead
Field marketing	List ingestion, cleanup and enrichment	Attendee scoring and routing on signals; event-to-opportunity reporting	Generic “personal” follow-up at volume	Reps send the follow-up	Opportunities created in 90 days; days to close
Customer marketing	Case-study drafts from interview transcripts	Several versions of each story (case study, page, YouTube video and Short, PDF) aimed at buyer prompts you don’t rank for yet	Customer stories written without an interview	Customers approve every quote attributed to them	References delivered; influenced pipeline
Sales enablement	Deal and call prep briefs	Live battlecards, role-play and post-call scoring	Custom decks sent to buyers without review	Enablement owns what’s approved	Seller use; win rate where content is used
BDR and SDR	Account research briefs before every call	Signal-ranked call lists and send timing; short, role-specific first emails; live objection prompts; post-call scoring	AI SDRs sending without review	PMM approves sequence structure; the BDR picks the play	Positive replies; meetings held; accepted opportunities

Team by team

Product marketing

Product marketing shapes what every other team says and has the least room for error, so put AI on the evidence and keep people on the positioning. Win/loss interviews, call recordings and support tickets hold the proof a security buyer will accept, and AI can now turn that raw material into usable plans quickly. Prashant Sridharan, whom PMM Hive identifies as head of product marketing at Supabase, described turning customer calls, specs, roadmap and research into a launch plan written into Notion (your team might use Monday, Asana, Jira or another platform). Tactics become Linear issues. He put the time at “about twenty minutes, including the time I spend guiding the reasoning” (PMM Hive).

Content is where most teams start. In The CMO Survey, 73.9% of companies use AI for content creation, and 41.5% now use it for generative engine optimization. For a cybersecurity software company the bar is higher than speed. Every post, page and brief should be written answer-first for SEO, AEO and GEO, drafted only from approved claims, and checked against product truth before anyone outside sees it. Any claim not already cataloged in the claims ledger or a support documentation MCP needs sign-off from an SE or a product manager. If the most marketing-skeptical solutions engineer on your team can read it and stand behind it, it’s ready.

Positioning is where I’d hold the line. Agents can gather the evidence and test the message, but choosing the market and the claim is the job, and it stays with a person until there is confidence in the data sources and the review gates are strong enough to catch errors. Only 6% of B2B marketing leaders trust AI to contribute meaningfully to positioning, in a Move Forward Strategies survey of 277 leaders in the UK and Ireland. In security, I’d put a claims ledger first. Every number, customer name and integration claim carries a source, an owner and an expiry date, and agents draft only from the ledger. In the agent system I helped build and run, a refuter strips any statistic or quote it can’t trace to a published source before a person reviews the draft.

I’d hold synthetic personas. They’re quick, they tend to echo your own assumptions back, and security buyers are too specific for that.

PR

Earned media now feeds the answer engines. Muck Rack’s Generative Pulse team analyzed more than 25 million links. Earned media accounted for 84% of AI citations in its May 2026 edition, and paid and advertorial content for 0.3% (Muck Rack). For a Series A or B company without a brand budget, that changes what a trade-press story is worth. My read is that a good one can keep surfacing long after launch, whenever buyers ask an LLM about your category.

PR teams have noticed. In Muck Rack’s State of PR 2026, a survey of 971 PR professionals fielded in May and June, 73% call generative engine optimization at least somewhat important. Securing coverage in high-authority publications is their most common tactic for AI visibility, chosen by 55%. Half of them, 51%, say thought leadership now makes up at least a quarter of the job. Measurement lags, with 39% not currently measuring GEO and 29% saying no one at their organization owns it.

Results data is thinner and mostly comes from vendors. Stacker, which sells content distribution, reported a 2.4x lift in how often those brands were cited after it distributed 87 of their stories (Stacker). I haven’t found an independent study that isolates what a placed executive quote does for LLM citations, so I’d treat agency case results as promising and unaudited.

AI handles the legwork well: reporter and outlet research, coverage summaries and first drafts of releases built from an approved fact base. Executive commentary pegged to your research, placed in high-authority outlets and tracked for LLM citations, is scaling. Two things stay human. Replies to reporters go out as options with a short risk memo, and whoever owns press relations picks one. At a Series A or B company, that’s often the CMO, working with an outside agency. Executive quotes come from the executive, in their own words, and every claim inside one traces to the ledger.

Analyst relations

In enterprise security, analyst evaluations often shape shortlists, and on a team that fits around one table, AR is usually somebody’s second job. Terezia Stancik, Senior Analyst Relations Manager at Shopware, told PMM Hive that major analyst evaluations once meant weeks of drafting hundreds of questionnaire answers. With an AI tool she built for those RFPs and questionnaires, she said, “what used to take weeks now takes under five working days with just a few stakeholders” (PMM Hive).

Draft evaluation and questionnaire answers only from approved claims, and have product sign every technical answer. Stancik named agentic digital twins, AI agents that know every report and remember every briefing, as the biggest potential disruptor ahead. I’d file that under early and keep the relationship itself human. Report analyst references, evaluation inclusion and the follow-ups you close after each briefing.

I’d rather walk into an analyst briefing with three proof points I can defend than ten I can’t.

Demand generation

AI use for targeting and segmentation keeps climbing. Since fall 2023, AI for targeting rose from 31.7% to 45.2% of companies, and for segmentation from 21.5% to 35.6% (The CMO Survey). Depth lags, though. Among enterprise marketers who personalize, 58% call their personalization basic (CMI).

Scoring on stacked intent signals and variant testing are proven. An AI search sprint on your top 50 buyer questions is scaling, and in my view it’s the cheapest discoverability work a Series A or B team can run. Nurture agents for leads below MQL are early.

Report on three clocks: signals in hours, conversion in weeks and pipeline in months. B2B cycles run 6 to 18 months or longer, per CIMM, so a pipeline number alone will lag every change you make. Agree on definitions with sales before the quarter starts.

Field marketing

Events are expensive for a Series A or B company. They’re also hard to defend. CMI found 88% of enterprise marketers allocated budget to experiential in 2025, yet 25% aren’t sure what they spend and 39% measure revenue or ROI. At the same time, 44% say deals close faster when experiential is involved.

Agents can now handle most of the event data loop. Badge scans and registrant lists sync into the CRM, and agents dedupe records, match people to accounts and enrich them. Scoring and routing attendees on signals is scaling. Reps send the follow-up themselves. Report opportunities created from attending accounts within 90 days, and days to close on deals with an event touch against matched deals without one.

Customer marketing

I’d argue customer proof carries more weight in security than almost anywhere. In the Ponemon and NOLA study, peer recommendations led vendor discovery for enterprise security buyers. Retention also matters more than budgets suggest. In The CMO Survey, customer retention is the strongest performance metric, growing 12.8% against 7.4% for acquisition. Acquisition budgets still run 26.0% larger than retention budgets.

Proven work here turns customer interview and call transcripts into case-study drafts the customer approves. Anthropic says its own customer marketing team drafts case studies in 30 minutes instead of 2.5 hours, a vendor-reported figure (Anthropic). A newer use is scaling, and it’s one of my favorites. Build several versions of each customer story, such as a case study, a web page, a YouTube video, a YouTube Short and a PDF, and optimize each one for SEO, AEO and GEO around the buyer prompts you don’t rank for yet, so the same story gets found by more of the right people. Never let AI invent or improve a customer quote.

Sales enablement

AI saves sellers time. Most sales organizations don’t reinvest much of it. Gartner says AI saves sellers an average of 4.8 hours a week, yet 72% of sales organizations report low reinvestment of that time in high-value selling. In its survey of 210 sales leaders, organizations that save moderate to large amounts of time and reinvest it are 2.2x more likely to exceed customer growth goals (Gartner). In Salesforce’s State of Sales, 46% of reps say they rarely get feedback on their conversations.

Prep is the proven use: deal and call briefs built from approved content. Live battlecards, AI role-play and post-call scoring are scaling. Hold anything that sends a custom deck to a buyer without review. Set a reinvestment rule for the hours saved, and report seller use and win rate where the content was used.

BDR and SDR

BDR teams have gone close to universal on AI, and the extra volume hasn’t predicted better results. In 6sense’s State of the BDR 2026, 99% of BDRs use AI, up from 53% in 2024. Touches per contact climbed to 34, from 17 in 2024. Volume wasn’t a reliable predictor of performance. Among BDRs whose companies have account-prioritization tools, 90% of the sample, only 19% use them to decide when outreach begins. Reviewing conversations through role-play and call simulation was the AI use reliably tied to higher quota attainment.

Account research briefs before every call are proven. Signal-ranked call lists and send timing are scaling, fed by signals from intent and enrichment platforms such as 6sense or Clay. So are short, role-specific first emails, live objection prompts and post-call scoring. AI SDRs that send without a person reviewing are early. Before anything loads into the sales engagement platform, PMM approves the sequence structure and a refuter strips any claim or customer name without a source. Report positive replies, meetings held and accepted opportunities, and leave touches off the dashboard.

Account research and cold email practices that hold up:

Research before the first touch. Build a brief on the account, its buying committee and its triggers, with a hypothesis the rep forms and owns.
Let signals set the timing. Where in-market accounts get their own targets, “wrong timing” rejections fall to 44%, against 65% with prioritization tools alone (6sense).
Keep the first email short. Elite performers on Instantly’s platform average fewer than 80 words in the first touch, and 58% of all replies come from step one (Instantly, platform data for 2025).
Write for one buying role, and point to one asset. Each email links to the post, page or finding that fits that role, so outbound and content say the same thing.
Stop at four to seven touches. Instantly puts the sweet spot there, and its average reply rate across the platform is 3.43%.

In the age where narrow ICPs and ABM enable precision targeting, I’d rather a BDR send ten emails built on real account research than a hundred that only swap in a first name.

One research asset, eight teams

I’d build toward shared inputs. One flagship research report or survey, governed once, can feed all eight teams. PR pitches the findings, AR briefs analysts on them, PMM builds the narrative around them, demand gets findings pages and webinars to run, field turns them into a session, customer marketing adds proof from customers who match the data, enablement builds the deck section, and BDRs open their first email with a number.

Enterprise security buyers in the Ponemon and NOLA study rated research and survey reports the asset type with the strongest direct impact. One evidence packet and one claims ledger keep every derivative quoting the same number.

Tag every derivative by funnel stage, persona and vertical before anything ships, so each team knows which version it owns and which number it quotes.

Team	Derivative	Funnel stage	Persona
PR	Findings release, data pitches, executive commentary	Awareness	Trade and business press, then the CISOs who read it
Analyst relations	Briefing deck on the findings	Awareness to consideration	Analysts who cover the category
Demand generation	Findings pages, webinars, full report, paid social	Awareness to consideration	CISO, security architect
Field marketing	Event session, regional roundtable	Consideration	Security leaders in each region
Product marketing	Narrative, proof points, claims in the ledger	Every stage	Every persona, through every other team
Customer marketing	Customer stories that match the data	Decision	Champion and economic buyer
Sales enablement	Deck section, talk track, objection handling	Decision	Buying committee, through AEs and SEs
BDR and SDR	First-touch emails that open with one finding	Awareness to consideration	Security and IT directors

Cut the findings by vertical too, such as healthcare, manufacturing and financial services, so a hospital CISO sees the hospital number. That only works if the report or survey collects enough responses per industry, so plan the cuts before fielding it.

The first 90 days

Days 1 to 30. A messaging source and claims ledger owned by PMM, win/loss and call mining, BDR account briefs and event list cleanup. Record baselines before anything else changes.
Days 31 to 60. Battlecard refresh, call scoring and role-play, a tracked set of 50 buyer prompts, and evidence packets for PR and AR.
Days 61 to 90. One flagship research asset with derivatives for every team, each tagged by buyer funnel stage, vertical and persona, plus a reinvestment rule for the hours sellers save.

Where this map can be wrong

If I had one AI budget at a Series A or B cybersecurity software company, I’d spend it first on the evidence work in product marketing, because every other team repeats what PMM approves.

Request a copy of my CMO briefing. This post draws on my briefing, AI in B2B Marketing 2026, which gives the source and page for most figures here. Request your copy and tell me which two teams you’d start with. To score your team first, use the five questions in my companion post.

FAQ

How do enterprise cybersecurity buyers use AI when they evaluate vendors?

More than a third of the 320 enterprise security decision-makers in the Ponemon Institute and NOLA Marketing study used AI somewhere in product selection. Peer recommendations still led vendor discovery, at 55%, and research and survey reports had the strongest direct impact of any asset type, at 46%. Gartner found 69% of B2B buyers turn to sales reps to validate AI-generated insights.

Does PR still matter when buyers ask LLMs instead of searching?

Yes. Muck Rack’s Generative Pulse found earned media accounted for 84% of AI citations in its May 2026 edition, against 0.3% for paid and advertorial content. Yet in Muck Rack’s State of PR 2026, 39% of PR professionals aren’t currently measuring generative engine optimization.

Sources

Ponemon Institute and NOLA Marketing, The State of Cybersecurity Marketing Influence 2026 (full report, behind a form)
Gartner, 67% of B2B buyers prefer a rep-free experience and 69% turn to sales reps to validate AI-generated insights, two releases from surveys fielded Aug to Sep 2025 (n=646 and 645)
PMM Hive, How I Use Claude To Build Launch Plans From Chaos and issue 36
Move Forward Strategies, 2026 State of AI and B2B Marketing, p.12
The CMO Survey 2026, pp.22, 59, 75
Muck Rack Generative Pulse, May 2026 release
Muck Rack, State of PR 2026, 971 usable responses, May 14 to June 12, 2026
Stacker, More Surface Area, More Citations (vendor-reported, 87 brand stories)
PMM Hive, Analyst Relations in the Age of AI
CMI and MarketingProfs, Enterprise Content and Marketing Trends: Insights for 2026
CIMM, LinkedIn and LiveRamp, The Measurement is the Message: Effective Practices for B2B Marketing Measurement, p.9
Anthropic, How Anthropic uses Claude in marketing (vendor-reported)
Gartner seller survey via BusinessWire
Salesforce, State of Sales, p.25
6sense, State of the BDR 2026
Instantly, Cold Email Benchmark Report 2026 (platform data, Jan 1 to Dec 18, 2025)