Most revenue teams already use AI somewhere in their motion. That is no longer what separates them. Our 2026 GTM Maturity Report finds the gap now sits in one place: how much of the work after a sales conversation your system does for you, and how much revenue per head that produces.
The report maps that gap as a four-stage curve, from call recording to an agentic GTM harness. Below are the headline findings, plus the customer data Devang Agrawal, Airspeed’s cofounder and CTO, presents in his talk “If AI isn’t increasing your revenue per head, you’re doing it wrong.” The full report, with the scoring table, stage-by-stage playbook, and customer story, is a free download on the GTM Maturity Curve page.
Why has revenue growth decoupled from headcount?
In 2026, headcount growth and revenue growth are moving apart across B2B sales organizations. Valuation and durable margin increasingly track revenue per head, meaning output per employee, rather than raw sales capacity. The operating model is shifting from adding reps to augmenting them with software.
Two forces drive it. The first is persistent inefficiency in how reps spend their time. The second is AI moving from pilot to committed enterprise spend. The numbers the report compiles make both plain:
| Finding | Figure | Source |
|---|---|---|
| Share of a rep’s time spent on non-selling work | 60% | Salesforce, 2026 |
| Revenue growth at AI-enabled sales teams vs. others | 83% vs. 66% | Salesforce State of Sales, 2026 |
| Time saved per seller per week by AI | 4.8 hours | Gartner, 2026 |
| Orgs that never reinvest that saved time | 72% | Gartner, 2026 |
| Worldwide AI spending in 2026 | $2.59T, up 47% in a year | Gartner, 2026 |
| Organizations deploying AI agents | 53%, up from 12% in 2024 | KPMG, 2026 |
The 72% figure is the one to sit with. Most teams already bank the time AI saves. Almost three in four never turn it into selling.
Why isn’t AI showing up in revenue?
Because most teams measure the wrong thing. The hours AI saves are real, but they get absorbed into the rest of the week instead of reinvested into selling, so nothing compounds and nothing reaches the P&L. The gains go to the teams that redesigned the work, not the ones that bought the tools.
Devang’s argument is to change the question before judging the answer:
| The old question | The question that counts | |
|---|---|---|
| Ask | How much time did AI save us? | Did revenue per head go up? |
| Counted in | Seats, licenses, and hours | One number already on the board’s page |
| What happens | Saved time is absorbed, not reinvested | It rises as the system takes on execution |
| Over time | Never shows up in the P&L | It compounds as the system learns |
If revenue per head does not move, the initiative did not work, however many seats are live.
How has buyer behavior changed what sales teams need?
Buyers now do most of their evaluation before a rep ever joins the call. According to Gartner (2026), 61% of B2B buyers prefer a rep-free buying experience. The 6sense Buyer Experience Report (2026) finds 81% have a preferred vendor before the first sales conversation. Gartner also expects 95% of seller research workflows to start with AI by 2027, up from under 20% in 2024.
That raises the stakes on the conversations reps do get. There are fewer of them, each carries more weight, and there is less room to misread a signal or miss a follow-up.
A model built on retrospective review cannot keep pace. Static recordings and dashboards read after the fact deliver the insight long after the moment that needed it. The report’s conclusion: keeping pace now depends on turning each conversation into action while the deal is still live. For a closer look at where conversation data leaks value, see why conversation intelligence misses sales insights.
What are the four stages of GTM maturity?
Every revenue organization sits somewhere on a single curve. Each stage is defined by three things: what the system does, what the human does, and the business impact that results. The human role is the clearest tell.
| Stage | Mode | Focus | Human role | Impact |
|---|---|---|---|---|
| 01 Call recording | Passive | Historical capture and compliance | Manual listener | Basic auditability, high friction |
| 02 Conversation intelligence | Reactive | Keywords, sentiment, rep coaching | Active consumer | Better visibility and rep training |
| 03 Revenue execution | Active | Pipeline management and process enforcement | System manager | Standardized playbooks, better forecasts |
| 04 Agentic GTM harness | Proactive | Autonomous execution and real-time orchestration | Human in the loop | Compounding gains in revenue per head |
At stage 01, calls are stored but inert. Review happens after the deal is lost, if it happens at all. At stage 02, recordings become searchable and coachable, but the system sees everything and decides nothing. Managers translate dashboards into action by hand. If you are weighing tools at this stage, our conversation intelligence buyer’s guide covers what to test.
At stage 03, signal starts to drive workflow. The system scores pipeline, fills the CRM, and enforces playbooks, so manual CRM upkeep drops and forecast confidence rises. It still does only what it is told. At stage 04, background agents run the research, prep, follow-up, and handoffs, and route consequential actions to a person for approval. Each interaction updates the system’s guidance, so it gets better as it runs.
What is the difference between an agent and a harness?
An agent is a worker. A harness is the workplace. Every standalone agent starts from zero: it needs its own briefing, data, tools, and rules before it can do anything useful. Agents inside a harness join a workplace that already knows your business, so each new one starts on the task, not the setup.
What the harness knows is your GTM context graph. Every call, email, meeting, and support ticket becomes a node: who said what, to whom, and what happened next. The graph keeps growing, and each node makes the next answer better informed. That is why Devang frames maturity as how much of your buyer, deal, and playbook knowledge the system can act on.
Take a simple question: “Why are my deals not progressing from discovery to demo?” A task agent answers the words. A harness knows what they mean for your business:
- Which deals? Segment, owner, source, and size, and which CRM records are actually still live.
- Compared to what? Time in stage against your own history, not an industry benchmark.
- By whose definition? Your exit criteria, and whether the calls actually met them.
- What happened in between? Every call, email, and ticket between the stages, and who from the buying committee showed up.
Where do most teams stall?
Progress up the curve is continuous, not binary, and every step pays back. But the report is direct about where the largest gains, and the most common stall, sit: the move from operating a system to supervising one, between stages 03 and 04.
Below stage 04, people still assemble the work by hand, deal by deal: the prep, the follow-up, the admin. Crossing over means the organization directs the system rather than operating it.
Teams stall here for one reason more than any other. They move the technology or the people, but not both. Invest only in the setup and you get a well-built agent nobody adopts. Invest only in the team and you get motivated reps with generic tools that do not fit how they sell. As Mathew Fitzgerald, Global Revenue Enablement at Foleon, puts it in the report:
“You have to build an environment where people feel comfortable tinkering, where it’s okay to try something with AI and have it fail. That is what turns a tool into adoption.”
How do you move up the curve?
The report lays out three steps, each with a technical half and a people half that have to move together.
- Understand your own challenges. Use AI to read your real picture: ICP, blockers, common objections, and where you actually lose deals, not where best practice assumes you do. Align managers and reps on those problems so the team solves the same thing.
- Build what fits your motion. Design the specific agents you need and be explicit about what they should not do. A generic play, like an instant email follow-up, may be wrong for your industry. Then train reps to read the output and act on it, so the tooling becomes how they sell.
- Build trust, and keep it learning. Put human-in-the-loop checkpoints in place so nothing goes out unreviewed, and route every outcome back so the system keeps improving. Reps learn to review, approve, and correct the AI against proven sales methodology.
Does moving up the curve actually raise revenue per rep?
In Airspeed’s own customer data, yes, and the effect scales with agent use. Devang’s talk indexes each customer’s median won deal value per sales rep in the most recent period to its own year before Airspeed (= 100), then groups customers by how many agent runs each seat makes per year:
| Agent use | Runs per seat per year | Won value per rep (index) | Customers improved |
|---|---|---|---|
| Before Airspeed (n=116) | None | 100 | Baseline |
| Light (n=13) | Under 15 | 109 | 46% |
| Regular (n=35) | 15 to 50 | 136 | 69% |
| Heavy (n=68) | 50 or more | 154 | 74% |
The curve in the data has the same shape as the maturity curve. Across the customer base, the talk reports:
- +38% median increase in won deal value per rep since adopting Airspeed, with 7 in 10 customers improving
- +45% deals won per rep per year for the median customer, with 79% improving
- +56% vs. +23%: deals won grew more than twice as fast as sales headcount for customers with a full year of data, so output is outgrowing the team
- About 6x growth in agent usage per seat, from 9 to 55 runs per year, between a customer’s first six months and months 12 to 24
What does moving up look like in practice?
Foleon started at stage 01. The team ran a legacy call recorder, pasted transcripts into ChatGPT or Gemini to draft follow-ups, and sat on thousands of transcripts with no way to see which conversations led to closed-won versus closed-lost.
After switching to Airspeed, the report documents:
- 54 hours saved per rep, per month, from automated follow-ups and call prep
- 3 minutes 37 seconds average time to transcribe a call, with follow-up emails ready instantly
- 145 agents in production across sales, customer success, and product
- 11 days from MQL to closed-won for a rep in their first month, the team’s fastest
“The license paid for itself in two months,” Fitzgerald says. “100% of the team onboarded in an hour.” Read the full Foleon customer story.
What should revenue leaders take away?
Devang closes his talk with three points:
- Measure AI initiatives on revenue per head. Not seats, not hours saved. If the number does not move, the initiative did not work.
- Mature GTM context makes mature agents. An agent is only as good as what it knows about your buyers, your deals, and your playbook.
- Agents execute, humans verify. Agents inside a harness have the context to do the work. Your team keeps them on track.
Get the full report and find your stage
The summary above covers the headlines. The full 2026 GTM Maturity Report adds the mechanics and cost of staying at each stage, the four-question self-assessment with its scoring table, and how teams put each step in place, from self-serve setup to a Forward Deployed Engineering team that builds the system for you.
Airspeed is built for the move this report describes: from recording what was said to doing the work that follows. Start by placing your team on the curve. The GTM Maturity Curve assessment takes four questions and returns your stage plus a recommended next step, and the full report is a free download on the same page.
Market statistics are compiled in the report from primary sources: Gartner; Salesforce State of Sales (2026); KPMG AI Pulse (2026); 6sense Buyer Experience Report. Customer statements are drawn from recorded Airspeed customer calls and quoted as spoken. Revenue-per-rep figures are Airspeed customer data presented by Devang Agrawal: median won deal value per sales rep, indexed to each customer’s own year before Airspeed.
Frequently asked questions
What is the GTM maturity curve?
A single path every revenue organization sits on, from passively recording sales conversations to proactively orchestrating the work that follows them. It has four stages: call recording, conversation intelligence, revenue execution, and an agentic GTM harness. As a team moves up, the system takes on more of the execution and revenue per head rises, modestly at first and then sharply.
How do I know which GTM maturity stage my team is at?
Look at the human role first. If people listen to recordings by hand, you are at stage 01. If they consume dashboards and insights, stage 02. If they manage a system that enforces playbooks, stage 03. If they approve work that agents already did, stage 04. The four-question self-assessment on the GTM Maturity Curve page places you in about a minute.
Where do most revenue teams stall on the maturity curve?
At the move from operating a system to supervising one, between revenue execution (stage 03) and the agentic GTM harness (stage 04). The report finds this is also where the largest gains sit. Teams stall when they invest in the technology or the people but not both: a strong agent nobody adopts, or motivated reps with generic tools that do not fit.
What is an agentic GTM harness?
The mature stage of the curve. Background agents run GTM workflows such as research, call prep, follow-up, objection handling, and cross-functional handoffs, and route consequential actions to a human for approval. Every outcome feeds back into the system, so playbooks adapt to what actually wins deals and revenue per head compounds.
How should revenue leaders measure the ROI of AI in sales?
On revenue per head, not on seats deployed or hours saved. Gartner (2026) finds AI saves sellers 4.8 hours a week, yet 72% of organizations never reinvest that time into selling, so it never reaches the P&L. Revenue per head is one number already on the board's page, and it rises only when the system takes on real execution.
What is the difference between an AI agent and an agent harness?
An agent is a worker and a harness is the workplace. A standalone agent starts from zero and needs its own briefing, data, tools, and rules. Agents inside a harness draw on a shared GTM context graph built from every call, email, meeting, and ticket, so each new agent starts on the task instead of the setup and answers in the context of your business.
How do I get the full 2026 GTM Maturity Report?
Download it free from the GTM Maturity Curve page on goairspeed.com. It includes the full breakdown of each stage, the self-assessment and scoring table, a three-step plan for moving up, and the Foleon customer story.
