Marketing invested in AI. It still cannot prove the return.
The market scores 45 out of 100, down two points from last year. It slipped for one reason: AI adoption is widespread, but measurable return is not. 88% of teams use AI somewhere, and only 39% can point to bottom-line impact.5
One score, one scale, and 14 years of the same report card
The Revenue Marketing Index grades how well a marketing organization converts spend into revenue it can prove. We built the framework in 2012 and have graded companies against it ever since, which is why a score in 2026 is comparable to a score in any earlier edition.1
Every company lands on a 100-point scale that maps to four stages of the Revenue Marketing journey. The stage is the finding. The number just says where inside the stage a company sits.
600 companies surveyed
1,000 companies analyzed
Both scores sit in Lead Generation. The full arithmetic behind each is in How we built this.
Marketing runs campaigns and creates materials. Success is measured by activity, not revenue.
Marketing is measured on volume: leads, clicks, downloads. It hands leads to sales and loses visibility after the handoff. Most of the market is here.
Marketing owns a pipeline target and can connect programs to opportunities and deals. Sales and marketing work from the same funnel and definitions.
Marketing commits to revenue, measures return by program, and reallocates investment based on what performs.
Every score in this report, whether for the market, a revenue band, an industry, or one of the six parts of the engine, is placed on this same scale.
The whole report on one page
For companies with $50M–$5B in revenue, the same range measured last year. The score fell from 47 to 45 as we raised the bar from leading indicators to provable revenue outcomes.
Across all 1,000 companies, including those under $50M, the score falls to 42. Smaller companies tend to have lower maturity, which pulls the broader average down.
Six parts, each scored out of 100. Average them and you get 45. Notice the shape: the 2 parts you can buy score highest. The 4 you have to build all trail.
Better indicators. Still very little revenue proof.
Marketing is getting better at measuring opportunities instead of leads, but both are still leading indicators. Only 39% can point to AI reaching the bottom line, and 95% report no return on AI spend.
Read the full story on Results →22% are Traditional, 34% are in Lead Generation, 28% have reached Demand Generation, and only 16% operate as a Revenue Marketing organization.1
88% use AI somewhere, but only 39% can point to bottom-line impact.5,6
What to do next: document the workflows AI will touch, measure revenue instead of lead volume, and understand how AI describes your company to buyers.27
The bar moved.
Most companies did not.
Across the 1,000 companies we analyzed, and in last year’s survey before it, only about 16 in every 100 run marketing as a real revenue engine.1 That group is now shrinking, not because anyone got worse, but because the definition got harder. Counting opportunities instead of leads used to be enough to look competent. Now the question is what marketing delivered and what AI returned, and that is where the market fell 2 points this year.
What is really going on
Everybody bought AI. Almost nobody can bank it, and that is what moved the score. 88 out of 100 companies use AI somewhere in the business. Only 39 can point to a single dollar of profit from it, and about 6 are genuinely good at it.5 The software is not the problem. The problem is everything around the software: how work flows, whether the data is clean, whether anyone knows how to use it, and who is watching.
Teams are getting smaller, and nobody is training them. Marketing hiring slowed by half in a year while AI use more than doubled.31 Leaders expect machines to handle 36% of marketing work by 2028, more than double today.30 Meanwhile training spending fell to 3.8% of the budget.31 Companies are cutting the exact muscle they are about to need.
Getting good at AI is how you win the budget fight. The average marketing team gets 7.8% of company revenue to spend, and 56% of marketing chiefs say it is not enough to do what they promised. Teams that can prove they are ready for AI get 8.9%, and they put more than a fifth of it straight into AI.3 Readiness is not just about doing better work. It is how you get the money to do it.
Your buyer picks a favorite before you know they exist. Nearly 9 in 10 B2B purchases now involve AI somewhere. 94 out of 100 buying teams have already ranked their top choice before they talk to a single salesperson, and that early favorite wins about 8 times out of 10.18 The race is decided in a chat window you never see.
AI helpers separate the good from the loud. The best marketing teams are 60% more likely to have AI agents doing real work.7 But 45% of tech leaders say the agents their vendors sold them do not live up to the pitch, and half admit their own data is too messy for an agent to touch.15 Buying agents is easy. Being ready for them is not.
Stop asking whether you are using AI. Ask whether your revenue engine can actually run it. The winners this year treat AI like a new engine for the whole car, not a gadget bolted onto one wheel.
45 out of 100 puts the market in Lead Generation, and Results is what pulled it down
Last year the market measured 47. This year it scores 45. The decline does not mean companies suddenly got worse. The bar moved. Counting opportunities and running attribution used to look like revenue measurement. It is not. Real revenue measurement is pipeline delivered against target, deals closed, cost per deal, and return on AI spend, and on that test almost everyone fails. Results went from the strongest score to one of the weakest.
The important point: the market did not just lose two points. It moved farther from Demand Generation.
Want the math? See How the index works →
The shape matters more than the number. Strategy and Technology score highest at 50, the capabilities companies can most readily fund or acquire. The harder capabilities to build and sustain all trail: Customer at 47, People at 43, Results at 42, and Process at 40. You can buy technology and fund a strategy. You cannot buy proof, and proof is what this year’s index graded hardest.
Where companies sit
| Group | Last year, measured1 | This year, 1,000-company analysis | What the evidence suggests |
|---|---|---|---|
| Traditional | 22% | 21–26% | More than 1 in 4 marketing chiefs still report little or no AI in their campaigns, and budgets at this level cover basic automation. |
| Lead Generation | 34% | 33–38% | Lead-count reporting is losing credibility, but reporting changes tend to lag a budget cycle or two, and this remains the largest group. |
| Demand Generation | 28% | 25–30% | Widespread experimentation, 39% of it with AI helpers, alongside reported gaps in process and data quality. |
| Revenue Marketing | 16% | 12–16% | The smallest group in this year’s analysis. The standard now includes proving AI value in revenue terms, which only about 6 in 100 companies, the genuinely AI-mature subset, report doing. |
Why we trust the direction: our 1,000-company analysis was cross-checked against more than 35 external reports, and the model reproduces last year’s measured score to within roughly one point.3,5,7 Both point to the same pattern: Results weakened and the gap between leaders and laggards widened. The full method is in How we built this.
about it
Grade yourself on the same 1 to 4 scale before you compare to 45. A number you cannot reproduce is not a baseline.
Size shapes where companies start, and most begin in Lead Generation
We analyzed 1,000 companies across 10 industries and five revenue bands, then cross-checked the findings against 35+ industry reports.55 The goal is not to give you one market average. It is to show you where companies your size, in your industry, actually stand.
The method is in How we built this.
Here is what the average hides. A $1B+ technology company scores about 64, while a sub-$10M higher education organization scores about 24, nearly a 40-point spread. That is why the overall market average is only a starting point. Use your size-and-industry cell as the more relevant benchmark, and treat each score as a range of roughly 3 to 5 points rather than a precise ranking. *We had no survey data for these industries last year, so their numbers are our most cautious guess.
Every step up the size ladder adds 6 to 8 points on average
Once smaller companies are included, the overall score drops from 45 to 42; previous survey editions started at $50M in revenue. Across this analysis, company size is the strongest predictor of maturity. Each revenue band scores roughly 6 to 8 points higher than the one below it, and the share of companies operating as a true revenue engine rises from 2% in the smallest band to 22% in the largest.
Only the largest companies have crossed into Demand Generation. $100M–1B sits at 49, one point short of the line. Everything below $100M is still a lead factory, and the smallest band is a rounding error away from Traditional.
| Company size | Expected score | Share running a real revenue engine | What that looks like day to day |
|---|---|---|---|
| $0–10M | 28 (25–31) | 2% | Marketing is busy and nobody is counting. Almost nobody here runs a real revenue engine. |
| $10–50M | 35 (32–38) | 5% | The lead factory years. Software arrives before anybody writes down how the work happens. |
| $50–100M | 41 (39–43) | 8% | Where our survey editions have always started. This rung tracks the measured baseline closely. |
| $100M–1B | 49 (46–52) | 15% | Marketing drives demand and has budget to experiment. This is where AI pilots pile up and stall. |
| $1B+ | 55 (52–58) | 22% | Money and people compound. Even so, 3 out of 4 big companies still fall short of a real revenue engine. |
| Every size together | 42 (39–45) | 10% | Across every size: 27 in 100 old-school, 36 chasing leads, 27 driving demand, 10 running a revenue engine. |
How the industries stack up
One industry of ten has crossed into Demand Generation. Technology and Software clears the line at 51. The other nine, including financial services at 46, are all inside Lead Generation, which means the industry gap people talk about is a gap within one stage, not between stages.
| Industry | Expected score, with range | Share running a real revenue engine |
|---|---|---|
| Technology & Software | 51.0 (48–54) | 19% |
| Financial Services | 46.3 (43–49) | 14% |
| Healthcare & Life Sciences | 44.3 (41–47) | 12% |
| Manufacturing & Industrial | 42.4 (39–45) | 10% |
| Business Services | 41.8 (39–45) | 10% |
| Automotive & Transportation* | 41.2 (36–46) | 9% |
| Retail & E-Commerce | 40.9 (38–44) | 9% |
| Hospitality & Travel* | 39.7 (35–45) | 8% |
| Media & Communications | 39.1 (36–42) | 8% |
| Higher Education* | 38.0 (35–41) | 7% |
*We had no survey data for these industries last year, so we scored them cautiously.
What we looked at
We scored each company on 10 externally observable signals.
Two patterns stood out. First, technology depth drops quickly beyond the basics: many mid-market companies have CRM and marketing automation, but few add much beyond that. Second, even stronger companies show clear technology gaps. In this analysis, the stack is often the constraint.
Use your own cell, not the overall average. A $30M manufacturer should benchmark against companies with similar resources, not against the full market. Start with your size-and-industry cell, then aim for the next rung up.
What each stage looks like now
The framework has not changed. The cost of staying put has.
The four stages are the same. What has changed is what it costs to remain in each one, especially as AI reshapes how buyers research, how teams work, and how marketing proves revenue.
Marketing runs campaigns. Nobody links them to revenue. The new cost: these companies are effectively invisible in AI answers, which is where half of buyers now start looking.17
The lead factory. It is falling apart in public: fewer than 1 in 100 leads becomes a customer,48 and most teams have already switched to counting real deals instead.49
Marketing moves pipeline and sales mostly agrees. This is where AI projects come to die: plenty of pilots, no clean data or clear process to scale them on.6
Marketing commits to revenue and hits it. This year the bar went up: AI does real work under human supervision, the team knows what AI says about them, and the finance chief hears about AI in dollars.
Where the traffic jam is: the middle. Lead Generation teams are being pushed beyond a model that no longer works,48,49 while Demand Generation teams are struggling to scale AI without clean process, data, and measurement.3 The organizations that break through tend to have those foundations in place before they scale AI.5,38
The percentages come from last year's survey. We did not re-run it this year, so any statement about companies moving is our reading of the evidence, not a new measurement. See How we built this.
about it
Write one sentence naming your stage and the specific thing that disqualifies you from the next one. That sentence is your roadmap.
Every plan names AI.
Almost none of them funds it.
7 in 10 marketing chiefs say becoming an AI leader is critical this year. Only 3 in 10 say they are actually ready.3 That gap between wanting it and being able to do it is the defining strategy problem of the year, and it shows up on the invoice.
Follow the money. Marketing budgets are flat at 7.8% of company revenue, and 56% of CMOs say that is not enough to deliver what they promised.3 AI-ready organizations average 8.9%, and they allocate 21.3% of that budget to AI versus 15.3% overall.3 Readiness may not cause the larger budget, but the two consistently show up together.
Your go-to-market plan now has to answer a question that barely registered a year ago: what is your plan for the AI channel? 4 in 10 companies are already writing content to be picked up by AI engines,31 and 88% of marketers say they are optimizing for AI answers.7 Brand has a new job too. 85% of buyers think better of a company when an AI tool mentions it.17 Your brand is now something a machine reads, not just something a person feels.
| What we grade | What the evidence says | What good looks like now |
|---|---|---|
| How you go to market | 51% of software buyers now start in an AI chatbot. 67% would rather buy without talking to a rep, up from 61%.17,50,51 | AI answers and self-serve paths are real channels with owners and targets, not side projects. |
| How AI connects to revenue | Only 39 companies in 100 that use AI can point to any profit from it.5 | Every AI investment has a revenue promise attached, and someone checks whether it landed. |
| How you invest in AI | AI-ready organizations allocate 21.3% of the marketing budget to AI, against 15.3% overall.3 | AI has a funded line in the revenue plan, with an owner and a target, not an experiment fund. |
| What your brand stands for | 85% of buyers think better of a company when AI mentions it. 69% switched vendors because of what AI told them.17 | You check what AI says about you every quarter and treat it as a pipeline number. |
Put AI in the revenue plan, not the innovation fund. Less mature teams treat AI as experimental spend. More mature teams define the revenue outcomes AI is expected to influence and report against them every quarter.
about it
Put the AI channel in the go-to-market plan as a line item with an owner and a budget. Ambition without a line item is not a strategy.
Hiring slowed by half, and the training budget went with it.
Marketing hiring slowed by half in a year while AI use more than doubled.31 Leaders expect machines to do 36% of marketing work by 2028, more than double today.30 The org chart is being rewritten right now, and most companies are writing it with no plan.
The workforce is being reshaped: fewer overlapping roles, more AI-specialist skills, and more teams treating AI as added capacity. 39% of marketing chiefs are already cutting labor costs by merging overlapping roles, and 22% say AI has reduced how much they lean on outside agencies.4 At the same time, 78% are thinking about hiring people whose whole job is AI, and 45% plan to keep headcount flat and treat AI as extra hands.35 Gartner expects that by 2029, at least half of all knowledge workers will need to know how to work with, supervise, or build AI helpers.10
Here is the part that does not add up. The skills bar is rising fast. AI know-how is the fastest-growing skill on LinkedIn, showing up in job posts 6 times more often than a year ago.36 And training budgets fell to 3.8% of marketing spend.31 80% of marketing chiefs say their own people are too anxious about AI to experiment with it.30 Culture and training are not the soft stuff this year. They are now a constraint on execution.
Leaders have their own blind spot. 65% of marketing chiefs expect AI to change their job dramatically within two years, but only 32% think the role requires a real change in skills.33 That gap matters: AI literacy is quickly becoming a leadership requirement, not just a team capability.
| What we grade | What the evidence says | What good looks like now |
|---|---|---|
| Whether leaders can lead this | 65% of chiefs expect their job to change dramatically. Only 32% think they need new skills.33 | Leaders use the tools themselves and can explain the payoff to the finance chief. |
| Whether people get trained | Training is down to 3.8% of budget while AI know-how is the fastest-growing skill on LinkedIn.31,36 | Training is a funded line item, with a standard each role has to meet. |
| How you plan the team | Half of companies saw AI-related headcount drops in at least one function. 30% expect more next year.5 | Your staffing plan counts people and machines together, 2 years out. |
| Whether the team will adopt it | 80% of chiefs say staff anxiety about AI is blocking experimentation.30 | People are rewarded for working well with AI, and know where they go next. |
Use the capacity AI creates deliberately. If machines absorb a larger share of the work,30 decide now where those hours go: strategy, customer work, training, and AI oversight. Organizations that capture efficiency without building new capability risk ending up with lower headcount but no stronger workforce.31,36
about it
Train the team you have. Pick two use cases and get every marketer competent at them before you hire anyone new.
Nobody wrote the workflow down, and now a machine has to follow it.
Every major study this year points at the same culprit, and it is not the software. 7 in 10 marketing chiefs admit their processes are not ready to run AI at scale.3 Nearly half of revenue leaders say the work is still done by hand in ways that cannot grow, and 82% agree you have to fix data and process first.38 AI does not fix a broken process. It just runs it faster.
The basics are shakier than most executives think. Only 26 in 100 companies hold themselves to a deadline for following up with a new lead, and 45 say follow-up is slow or gets missed. Just 17 have AI doing real work in more than one part of the business.38 Gartner sees the same pattern from the other side: at least 30% of AI projects are abandoned after the demo, most often because of messy data, weak controls, cost, or no clear payoff.8
An AI helper can only run work that somebody has defined. If a process lives in one person’s head, no software can reliably take it over. Undocumented work used to create inefficiency. Now it creates a ceiling on automation and scale.
The team that should fix this is revenue operations, and it is thinly staffed: 1 in 4 of those professionals is a team of one, and 43% of those teams control no budget at all.39
| What we grade | What the evidence says | What good looks like now |
|---|---|---|
| How the work actually flows | Nearly half still do the work by hand in ways that cannot grow. Half cite tangled systems as the top barrier.38 | Your core revenue workflows are written down, measured, and a machine could run them. |
| How far AI gets past the pilot | Only 17 companies in 100 have AI doing real work in more than one area. 2 call it central.38 | Money and effort go where the predicted revenue is, not where the loudest request is. |
| How fast you follow up | Only 26 in 100 hold a deadline for calling a lead back. 45 admit follow-up is slow or missed.38 | The callback clock runs itself, gets measured, and sales owns it with you. |
| How you keep customers | Customer service teams using AI helpers jumped from 39% to 66% in a year.53 | Customer programs run continuously in the background, not as a quarterly push. |
Write it down before you automate it. Take the 5 workflows that matter most to revenue, write each one out, measure it, and put a clock on it. That unglamorous week of work is the whole difference between the 82% who know process comes first38 and the 3 in 10 whose AI projects quietly die.8
about it
Document the three workflows AI will touch first. Automation on top of an undocumented process multiplies the mess.
The stack shrank.
Now a machine has to run what is left.
The marketing software market has finally stopped growing. It sits at 15,505 products, barely up from last year, and 40% fewer new tools showed up.43,44 Inside companies the shift is sharper: the average large B2B company went from 62 tools down to 37 in a single year.38 The cleanup we flagged last year is well underway. The next question is harder: can an AI agent actually operate what remains?
Mostly not. 81% of tech leaders are testing or rolling out AI agents, but 45% of the ones who tried say the agents their vendors sold them do not do what was promised, and half admit their own systems and data are not ready for agents anyway.15 The older utilization problem has not gone away either: teams still use only about a third of the software they pay for, down from more than half in 2020.45
Two shifts are worth watching. First, software bills are rising even as tool counts fall, as AI-capable platforms command more spend.38 Second, the center of gravity is moving from the application layer to the data underneath it. Marketers with connected customer data are 60% more likely to have AI agents doing real work, and top performers are nearly 2.5 times more likely to have unified data.7
Underneath all of it sits the thing nobody wants to talk about: the data itself. 84% of data leaders say their company needs to overhaul its data before AI can work. They estimate a quarter of their own data cannot be trusted, and only 43% have any formal rules about who owns and cleans it.41 Among companies already running AI, 89% report getting wrong or misleading answers out of it.41 Bad data is not a plumbing issue anymore. It is the wall the whole house is leaning on.
| What we grade | What the evidence says | What good looks like now |
|---|---|---|
| Whether the stack supports revenue | Top performers are nearly 2.5 times more likely to have their customer data in one place.7 | One connected data foundation supports the systems and AI use cases tied to revenue. |
| How new tools are evaluated | 45% say vendor AI agents fell short of expectations. Of thousands of vendors claiming to sell agents, Gartner counts about 130 that truly do.15,9 | AI claims are tested before purchase, and every pilot has clear success criteria and a defined stop point. |
| Whether the stack is actually adopted | The average large company cut from 62 tools to 37 and still uses only about a third of what it pays for.38,45 | Usage is measured by platform, underused tools are removed, and teams are accountable for adoption. |
| Whether new technology can be deployed | Half of tech teams say they lack the people or capacity to deploy AI agents successfully.15 | Every new platform has an implementation owner, a training plan, and the resources required to make it operational. |
Judge your software on 3 questions, not on feature lists. Is your data clean and connected enough that a machine could safely act on it? Is every important workflow written down? And does the vendor selling you an AI agent actually have one? Companies asking those questions join the half of the market that is ready.15 The rest are buying this year’s label on 5-year-old software.
about it
Pause nonessential technology purchases for two quarters. Put that budget into data cleanup and adoption of the platforms you already own.
The shortlist is written before you know the buyer exists.
Picture a buyer at 11pm typing your category into a chat window. By the time they close the laptop, they have a favorite. Nearly 9 in 10 B2B purchases now involve AI somewhere. 94 out of 100 buying teams rank a preferred vendor before they contact a single salesperson, and that early favorite wins about 8 times out of 10.18 By the time your team meets the buyer, the race is mostly over.
The starting line moved fast. AI is no longer just another research tool; it is influencing which vendors make the shortlist. 69% of buyers say AI changed the vendor they ultimately chose.17 At the same time, more buyers want to move through the process on their own, with 67% preferring to buy without a rep.50,51
Faster is not the same as better, and buyers know it. Most still double-check what AI tells them,19 and many reach out to sales specifically to fill the gaps AI leaves behind.18 The winning pattern is choreography: let AI handle research and routine questions, then bring people in where judgment and trust matter.
Meanwhile the oldest promise in marketing is still unkept. 98% of marketers run into walls trying to personalize anything, usually because of data, and 84% are still sending the same message to everybody.7 The same customer knowledge that would fix that is the knowledge AI engines need in order to describe you correctly.
| What we grade | What the evidence says | What good looks like now |
|---|---|---|
| How you engage buyers | 67% prefer buying without a rep, and rising. But by 2030, 3 in 4 are predicted to want a human on complex deals.50,52 | You decide on purpose where AI hands off to a person, based on how big the deal is. |
| What you publish | 58% of buyers reach out to sellers early specifically to fill the gaps AI research left.18 | Content is written for 2 readers: the buyer, and the AI summarizing you to them. |
| How personal it gets | 98% hit a wall trying to personalize. 84% still send the same message to everybody.7 | What customers actually do drives what they see next, and you can measure the lift. |
| Whether you see the whole path | Buying cycles shortened from about 11 months to 10. Visitors arriving from AI convert 4.4 times better than search visitors.18,23 | Your journey map includes the touchpoints you do not own, with a plan for each. |
Map the part of the journey you cannot see. Most journey maps still start at first touch. The touches that decide the deal now happen earlier, in AI answers and peer conversations you will never witness. Go find out what AI says about you, check whether it is accurate, and treat that 80% early-favorite number18 as the target it is.
about it
Ask an AI assistant for a shortlist in your category and read what it says about you. That is your new first impression.
This is the score that fell, and it should have.
Results scored 53 last year, the strongest of the six dimensions. This year it scores 42, the second weakest. That drop does not mean performance suddenly collapsed. The standard changed: last year we gave more credit to leading indicators; this year we graded provable revenue outcomes. The gap between the two is substantial.
Start with the good news, because it is real. For the first time, more teams name sales opportunities as their top measure (52%) than those who name leads (47%).49 57% say they are tracing deals back to marketing, and another third start within a year.49 Nearly half are investing in modeling the whole mix.47 Forrester explains the urgency: fewer than 1 in 100 leads becomes a customer, and more than 8 in 10 B2B decisions are made by a group of 3 or more.48 Counting opportunities instead of leads is a genuine step up.
Now look at what those numbers actually measure. An opportunity generated is not revenue; it is a prediction that revenue might happen. Attribution tells you where a deal came from, not what the effort returned. The market has gotten better at leading indicators, but real revenue measurement asks different questions: how much pipeline marketing delivered against target, how much closed, what each deal cost to win, and what the program returned against what it consumed. On AI, the gap is even clearer: only 39 in 100 companies using AI can point to bottom-line impact,5 and most report no return at all.6 Last year we scored the indicators. This year we scored the revenue, and Results lost 11 points.
Here is what real revenue measurement looks like.2
- Pipeline sourced by marketing: 40 to 55% of total pipeline.
- Pipeline delivered against commitment: what marketing actually delivered against the number it signed up for.
- Cost per deal: what it costs to win revenue, not just generate activity.
- Revenue from existing customers: measurable growth from retention, expansion, and cross-sell.
- Attribution coverage: enough visibility to connect marketing activity to the majority of closed deals.
- Return on AI spend: what AI returned in dollars against what it cost.
Real revenue measurement goes beyond activity and leading indicators. It shows what marketing committed to, what actually converted to revenue, what it cost to produce, and what the investment returned.
| What we grade | What the evidence says | What good looks like now |
|---|---|---|
| Pipeline and revenue contribution | The best performers source more than half their pipeline. A mature revenue engine typically sources 40 to 55%.1,2 | Marketing commits to a pipeline target, reports how much it delivered, and tracks how much actually closed. |
| How performance is measured | More teams are moving toward mix modeling, but most still stop at leading indicators such as opportunities created.47,2 | Executive reporting focuses on outcomes: pipeline delivered against target, close rate, cost per deal, and return. |
| Whether AI shows up in financial results | Marketers report productivity gains and lower overhead from AI, but most figures are still self-reported rather than audited returns.32 | AI investment is reported in dollars: what it cost, what it returned, and whether the result can be verified. |
| Revenue from existing customers | Mature account programs are associated with materially higher renewal and expansion revenue.1,2 | Retention and expansion are measured as revenue outcomes, with marketing accountable for the programs that influence them. |
Move past indicators. Counting opportunities instead of leads was the easy half, and most of the market has done it. The hard half is reporting what marketing delivered: pipeline sourced against target, how much of it closed, what it cost, and what AI returned against its spend. This is the only part of the engine that got harder to score well on this year, which is exactly why it is the fastest way to separate from a market that mostly stops at indicators.48,49
about it
Report pipeline delivered against target, cost per deal, and return on AI spend. Retire lead counts from the executive dashboard.
Everyone is buying AI helpers.
Most will switch them off.
Two Gartner predictions sit side by side, and both may prove true. AI helpers are rapidly becoming standard in business software,10 while many companies that adopt them are expected to pull the plug.9 The difference will not come down to software alone. It will come down to whether the organization is ready to hand real work to a machine.
It is still early. Only 23 in 100 companies have AI helpers doing real work at scale, while 39 are still experimenting.5 But the gap is already showing. Top-performing marketing teams are 60% more likely to have agents doing real work.7 Sales teams that give reps AI-suggested next actions are 2.6 times more likely to hit their growth targets.12 And in customer service, agent adoption jumped from 39% to 66% in a year.53 Where agents are deployed well, the results can be substantial: Salesforce says its own support agent handles more than 84% of customer questions without a human.13
The money has already moved. 79% of executives say agents are being adopted in their business, and 88% plan to increase AI spending over the next year.11 The risk is rising with it. Forrester expects B2B companies to lose more than $10 billion in value this year from AI operating without adequate oversight.20 Adoption is moving faster than governance.42
Five questions that tell you which side you are on
| The question | Traditional | Lead Generation | Demand Generation | Revenue Marketing |
|---|---|---|---|---|
| How much is actually running 23 in 100 companies are scaling agents. 39 are still experimenting.5 | No agents anywhere. Someone tried a chatbot once. | One team runs a pilot nobody else knows about. | Agents do real work in 1 or 2 workflows, closely watched. | Agents run several revenue workflows, with clear rules for adding more. |
| Is the work written down 82% agree process has to be fixed before AI.38 | It lives in people’s heads. | Some steps written down, none measured. | The important workflows are written down and measured. | Written down, measured, on a clock, and a machine can run them. |
| Can an agent trust your data Half of tech teams say their stack is not agent-ready.15 | Data is scattered across tools that do not talk. | The CRM and email system are connected. Nothing else is. | One customer data layer covers most of what agents need. | One trusted set of data that agents act on in production. |
| Who is watching Forrester expects more than $10B in value lost to ungoverned AI this year.20 | No owner, no rules, no record of what happened. | Whoever built it keeps an eye on it. | Every agent has a named owner and a rule for when to escalate. | People approve the risky calls, and every action is logged and reviewable. |
| Can you prove it worked 45% say the agents their vendor sold them fell short.15 | Nobody asks what it did. | The value is a story, not a number. | Each agent reports what it handled and what it passed on. | Each agent reports its revenue impact every quarter. |
Why this is a score and not a checklist: the reasons Gartner gives for those canceled projects are costs running away, no clear payoff, and no controls.9 Those are rows 5, 2, and 4 of this table. If you land in the left half on any 2 of these 5 questions, the cancellation statistics are describing you.
about it
Give every agent a named owner, a success measure, and a review date. Agents without clear ownership and proof of value are the most likely to stall.
Buyers start in AI, and most companies are not there when they do
Two years ago, Gartner predicted that search traffic would decline as buyers shifted more research into AI tools.16 That shift is now visible. Buyers are increasingly starting in AI, while most companies are still optimized for the old front door.
Welcome to the no-click internet. More searches now end without a traditional click, while AI referral traffic is still tiny but growing quickly.21 It accounts for less than 0.15% of web visits today, but grew 66% last year.25 And the visitors who do arrive from AI appear unusually valuable: they convert at higher rates, stay longer, and bounce less.26 AI referral traffic grew 66% year over year, and the visitors it sends show 8% higher engagement and 23% lower bounce.
0.15% of web visits come from AI referrals, against 16% from organic search. Drawn to scale.
AI referral traffic grew 66% year over year, and the visitors it sends show 8% higher engagement and 23% lower bounce.
Budgets have already moved. Large companies put an average of 12% of digital spend into AI visibility last year, and 94% of marketing leaders plan to spend more this year.27 The behavior is changing too: 88% of marketers say they are now creating content for AI answers, and top performers are more than twice as likely to be doing it.7
We measure this as your AXO Score: how often AI tools mention you, how accurately they describe you, and how strongly those answers are supported. Companies operating as a true revenue engine score between 60 and 80, while the industry average is 28.2 That gap makes AI visibility one of the clearest differences between the leaders and the rest of the market. And unlike many maturity gaps, improving it does not necessarily require rebuilding your technology stack. It starts with making the information buyers need clear, credible, and easy for AI systems to find.
This is not an SEO project with a new name. It is whether the fastest-growing research channel in business can find you, describe you correctly, and recommend you. Check your score, fix the flat-out wrong answers first, then build content worth quoting. Given that the early favorite wins 8 out of 10 deals,18 this is pipeline you either influence or hand to someone else.
about it
Publish machine-readable answers to the questions buyers ask before they shortlist. Start with the answers AI is getting wrong or cannot find.
Your next buyer might not be a person
Here is the prediction that should get your attention. By 2028, Gartner expects 90% of business buying to run through AI agents, moving more than $15 trillion through machine-to-machine marketplaces.28 That sounds like science fiction.
The early signs are already visible. Forrester expects 1 in 5 B2B sellers to negotiate with a machine this year, while 61% of purchase influencers say their company has or plans to build private AI to support buying decisions.20 AI-run searches on a buyer’s behalf have already tripled.29 Buyers have the technology. Sellers do not: only 28% of commerce teams use it today, although 44% plan to within six months.29
What a machine buyer rewards is not mysterious: clear product and pricing information it can parse, claims it can verify, a way to buy without a phone call, and a consistent story everywhere it looks.28 If your digital presence is designed only for human interpretation, you may be making it harder for the next generation of buyers to evaluate you.
about it
Make product, pricing, and terms readable and checkable without a phone call.
Six questions.
One honest number.
One question for each part of the engine, on the same scale we used for everyone else. It takes 2 minutes and it is not a sales quiz. It will not replace the full 49-question review, but it will tell you roughly where you stand.
The full assessment scores you on all 49 items, places you on the climb, and compares you to the companies in this report. On AI helpers specifically, we use the same 1 to 4 scale across the 5 questions above. Below 1.75 you are exposed. Up to 2.5 you are piloting. Up to 3.25 you are scaling. Above that you are genuinely running on agents. The market average right now is about 1.9, which is low piloting.5,15,38 Those cutoffs are our first version and we will adjust them when we field the survey again.
What good actually looks like
| Metric | Where most companies are | Where the best are | Source |
|---|---|---|---|
| Pipeline marketing sources, and closes | 20–30% | 40–55% | TPG2 |
| Leads sales actually accepts | 40–55% | 65–75% | TPG2 |
| What it costs to win a deal | 20–30% | 10–15% | TPG2 |
| Closed deals you can trace back | 55–65% | 80–90% | TPG2 |
| Revenue growth from existing customers | ~100% | 110–130% | TPG2 |
| How visible you are inside AI answers | 28 | 60–80 | TPG2 |
| Share of budget going to AI | 15.3% | 21.3% | Gartner3 |
| Marketing budget as a share of revenue | 7.8% | 8.9% | Gartner3 |
| Share of digital budget spent being found by AI | 12% at big companies | Going up: 94% spending more this year | Conductor27 |
| What AI actually saved or earned | 8.6% more sales productivity, 10.8% less overhead | Measured, and reported to finance | The CMO Survey32 |
The TPG rows are our published targets for a company running a real revenue engine. Outside numbers carry their dates in the source list at the back. Anything AI-related moves fast, so check it again before you build a plan on it.
about it
Pick the three benchmarks closest to your current gaps and put a target date on each.
How to climb one rung in a year
Moving up a stage is not a transformation program with a steering committee. It is a short list of unglamorous jobs done in the right order. Here is the order.
Do this: Agree with sales on what counts as pipeline. Clean up the CRM and get basic automation running. Write down how a campaign actually happens and who owns each step. Then go check what AI says about you, because right now it probably says nothing, and that is where half of buyers start.17
You are done when: you can follow a lead from start to finish, sales and marketing read the same report, and you know your AI score.
Do this: Stop reporting lead counts on their own and start reporting real opportunities and the groups behind them.49 Put a deadline on calling people back, and hold to it, which puts you in the top quarter.38 Plan with sales instead of near them. Start tracing closed deals back to marketing.2
You are done when: marketing reports pipeline it sourced, the callback clock runs itself, and you can trace most closed deals.
Do this: Get your customer data into one place first, the way the top performers did.7 Write down and measure your 5 most important workflows so a machine could run them. Put 2 AI helpers into real production with a named owner, a rule for when a human steps in, and an agreed point at which you would switch them off.9 Fund your AI visibility work properly, at around 12% of digital spend.27
You are done when: you report AI value in dollars, 2 agents are running with results you can audit, and your AI score is above 50.
Do this: Measure 3 ways at once: model the whole mix, trace the deals, and run tests that prove what actually caused what.47 Teach your agents more skills, the way the leaders did, going from 2 jobs each to 6.14 Make your product and pricing readable by machines before machines start buying.28 And promise the hours AI frees up to training, before the wave hits.30
You are done when: never. This is a habit, not a trophy.
about it
Sequence, do not parallelize. One rung a year is the realistic pace, and each quarter needs a named owner.
What we expect to be true a year from now
The AI helper shakeout gets loud. More than 4 in 10 of these projects are expected to be canceled, and of the thousands of vendors claiming to sell agents, only about 130 really do.9 Expect fewer agents next year, doing better work.
Your AI score reaches the board deck. With 94% of big-company marketing chiefs spending more on it27 and each AI visitor worth 4.4 times an ordinary one,23 this stops being a curiosity and becomes a number people are held to.
The skills bill comes due for leaders. Gartner expects not knowing AI to become one of the top 3 reasons marketing chiefs lose their jobs.33 The gap between leaders who expect upheaval and leaders who think they need new skills closes next year, one way or another.
about it
Decide which of these would hurt most if it came true, and buy insurance against that one.
What we did, and what we cannot claim
We have published this index many times since creating the Revenue Marketing category in 2012, always against the same six-part framework and the same four stages. Only the way we gather evidence changes from edition to edition. This year's numbers come from analyzing 1,000 companies, cross-tabulated with more than 35 industry reports. The most recent survey edition took a different route: 600 B2B companies, 47 questions, plus 50 executive interviews.1 We did not re-field that survey. We built something wider in its place. The 1,000 companies span 5 revenue bands and 10 industries, so every cut of the data has real companies behind it rather than one blended average. The 35+ reports, published between 2024 and 2026, are the cross-check: every score we publish had to survive both.
How the 1,000 companies break down. 100 companies in each of the 10 industries we work in,55 spread across 5 revenue bands from under $10M to past $1B. Each is scored on 10 things observable from the outside, and each score carries a note on what we saw and how confident we are. We hold the industry weights even on purpose: that makes this a comparison tool for finding companies like yours, not an estimate of what the whole market looks like. Before we trusted any new cut, we required the results to reproduce last year’s measured survey score, and they land within about 1 point of it. One honest limit: matching a number you were built to match proves consistency, not accuracy. Every setting behind the analysis is published in the table below, including the random seed, so anyone can rebuild it.
One scoring change you should know about. Last year we scored the Results dimension on leading indicators, whether teams counted opportunities and ran attribution, and it came out at 53, the strongest of the 6. That flattered the market, because neither of those is revenue. An opportunity is a prediction and attribution is a trace. This year we scored the outcomes instead: pipeline delivered against target, close rates, cost per deal, and return on AI spend. Far fewer companies can report those: 39% see AI reach the bottom line,5 and 95% report nothing back on their AI spending.6 This year we scored the capability. Results fell to 42, and the composite fell from 47 to 45 as a result. About 2 of that 2-point drop is this change in what we grade, not companies getting worse. We are flagging it rather than burying it, because a like-for-like comparison against last year’s 47 would put this year at roughly 47 as well.
How we checked the numbers. Every statistic here was traced back to the organization that published it, or to their own press release where the report sits behind a wall. Anything we could not verify got cut. One example: a widely repeated claim that organic traffic will halve by 2028, usually credited to Gartner, does not appear anywhere on Gartner’s site. So it is not in this report.
How to read it. Where our 1,000-company analysis and the outside reports agreed, we scored with confidence. Where they diverged, we widened the range and said so. Survey numbers carry their dates and sample sizes in the list at the back. Analyst predictions are forecasts, not measurements, and we label them that way. Where we say something is happening, like the middle of the pack getting squeezed, that is our reading of the evidence, not a new measurement. And the stage percentages are last year’s survey.
What we are not claiming. We are not claiming companies moved up or down this year. Some of our sources are published by software vendors, whose customers and commercial interests shape what they find. We use them where the underlying data is large or comes from real platform activity, and we always say who published it. Anything about AI should be treated as a snapshot that will age quickly.
Every setting in the model
| Setting | Value | Setting | Value |
|---|---|---|---|
| Starting point: under $10M | 1.85 | Industry nudge: Technology & Software | +0.25 |
| Starting point: $10 to 50M | 2.10 | Industry nudge: Financial Services | +0.10 |
| Starting point: $50 to 100M | 2.30 | Industry nudge: Healthcare & Life Sciences | +0.03 |
| Starting point: $100M to 1B | 2.55 | Industry nudge: Manufacturing & Industrial | -0.03 |
| Starting point: over $1B | 2.75 | Industry nudge: Business Services | -0.05 |
| How much companies vary | 0.75 | Industry nudge: Automotive & Transportation* | -0.07 |
| How much the 6 parts vary | 0.22 | Industry nudge: Retail & E-Commerce | -0.08 |
| Random seed, so anyone can rebuild it | 2026 | Industry nudge: Hospitality & Travel* | -0.12 |
| Nudge per part of the engine | +0.09 / -0.12 / -0.21 / +0.09 / 0.00 / +0.18 | Industry nudge: Media & Communications | -0.14 |
| Industry nudge: Higher Education* | -0.18 |
*We had no survey data for these industries last year, so their numbers are our most cautious guess. The ranges you see elsewhere reflect that: about 3 points either way where we have real data behind a setting, about 5 where we do not, and about 2 for the size band we calibrated directly. Those are working bounds, not statistical intervals.
The math, and what our confidence labels mean
The math. Grade each item from 1 to 4, average the items, average the 6 parts, then stretch it onto a 0 to 100 scale: subtract 1, divide by 3, multiply by 100. Last year’s 47 comes from scoring each of the 4 groups at the middle of its range and weighting by how many companies landed in each. You can redo that from last year’s published numbers alone.1
What our confidence labels mean. High means 3 or more independent sources land in the same place. Medium means at least 2 point the same direction but disagree on the size. Low means one source or a stand-in measure, and no Low-confidence score is used in the headline number. Treat every range as a working estimate, not a statistical guarantee.
What happens next. We field the survey again next year. Every estimate here, and every cutoff we invented for the AI helper score, gets tested against real data, and we will publish how far off we were.
This is version 1.0, published August 2026. Everything you need is in this one document: the report, the 1,000-company analysis, how it works, and every setting behind it. Coming in the next version: an outside review of our methods.
Every number, labeled and traceable
We label every source by what kind of evidence it is, because they are not equal. Survey means people were asked. Telemetry means software recorded what actually happened. Prediction means an analyst is guessing, carefully. Study/Census means someone counted things. TPG Benchmark means it is ours. Secondary-verified means the publisher locks the report away, so we confirmed it through their own announcement.