数据显示 AI 初创公司销售占团队比例未降，人力压缩集中在客服与后台

The common view for the past two years has been that AI-native companies reach $100M ARR with very small teams, that agents handle much of the selling, and that the traditional B2B sales org is shrinking away.

Part of that holds up. AI-forward companies do run smaller go-to-market teams in absolute terms. ICONIQ’s State of Go-to-Market 2026, a January 2026 survey of 150+ B2B GTM executives, has AI-forward companies at $10M to $25M ARR running about 20 GTM FTEs against 35 for their lower-adoption peers, or 43% fewer people at the same revenue.

New data from Pave, pulled from its real-time compensation database of 10,000+ companies, measures something different: what percent of each company works in each function. On that measure, sales at AI hypergrowth companies looks close to sales in the rest of tech.

Median % of workforce in sales:

• AI hypergrowth: 18.4%
• Rest of tech: 20.7%
• Non-technology: 10.4%

The gap is just 2.3 points. AI hypergrowth companies staff sales at roughly 89% of the share a traditional tech company does. They have fewer reps because they have fewer employees overall, and the sales org is about the same proportion of the company.

The bigger differences are in other functions. Customer support is 1.8% of headcount at AI hypergrowth companies versus 4.4% in the rest of tech. Engineering is 47% versus 33.9%. The “Other” bucket of marketing, ops, finance, IT, data and related roles is 8 points lower.

One methodology note: these are medians by function, so they don’t add up to exactly 100% (94.7% for AI hypergrowth, 91.5% for the rest of tech). The comparisons are directional rather than a precise org chart.

#1. Sales Is 18.4% of Headcount at AI Hypergrowth Companies vs. 20.7% in the Rest of Tech

If agents were replacing account executives in large numbers, sales would show the biggest drop as a share of the company. It shows one of the smallest. Sales went from about one in five employees to slightly under one in five.

Putting the two datasets together: ICONIQ shows AI-forward companies are smaller overall, and Pave shows sales keeps most of its traditional share inside those smaller companies. The lower rep count comes mostly from a smaller company, not from cuts aimed at sales.

ICONIQ’s function-level AI adoption data is consistent with this. The share of companies where more than half the function uses AI daily: SDRs 71%, marketing 65%, AEs 57%, RevOps 54%, account management 45%, customer success 41%. AEs are among the heaviest AI users, mostly for research, call prep, drafting and follow-up, and companies are still hiring them at close to the traditional ratio.

Those reps are also hitting their numbers more often. ICONIQ has AI-forward teams at 67% quota attainment versus 59% for everyone else.

#2. Customer Support Fell From 4.4% to 1.8% of the Company

Support’s share of the workforce is 59% smaller at AI hypergrowth companies than in the rest of tech, the largest relative drop of any function in the Pave data.

As a ratio, the rest of tech has about 4.7 salespeople for every support person. AI hypergrowth companies have about 10.2.

Support work is high volume, well documented, repetitive, and measured ticket by ticket, which is the kind of work AI agents handle well today. Enterprise sales involves security reviews, procurement and multiple stakeholders with different priorities, and that work is still mostly done by people.

For a 2026 hiring plan, the data points to support as the first place agents reduce headcount, well ahead of sales.

#3. Engineering Went From 33.9% to 47% of the Company

Engineering is nearly half of the median AI hypergrowth company. That works out to about 2.6 engineers per salesperson, versus 1.6 in the rest of tech.

Part of this is what AI products require: model work, evals, infrastructure, and a product that changes every quarter.

Part of it is likely forward deployed engineering. At many AI companies, FDEs sit in the engineering org and do work that a solutions engineer, implementation lead or CSM would have done at a traditional B2B company. AI products usually need to be configured and trained on the customer’s data from Day 1 to deliver value, and that work is now often staffed by engineers. Some post-sales headcount has moved into engineering rather than going away.

#4. The “Other” Bucket Dropped From 28.8% to 20.8%, While HR Went Up

Pave’s “Other” category covers IT, data, marketing, professional services, operations, creative and finance. It is 8 points smaller at AI hypergrowth companies, which makes the back office the second place, after support, where AI is clearly reducing headcount.

Two small functions moved the other way:

• HR: 5.6% vs. 3.2%. AI hypergrowth companies hire at a very high rate and need recruiters to do it. ICONIQ’s headcount data (as of August 2026) has companies growing 100%+ adding a median 133% to headcount in the first half of 2026, more than they added in 2022-2023.
• Legal: 1.1% vs. 0.5%. Larger contracts, data processing agreements, model and IP terms, and enterprise security reviews all add legal work.

Both are costs of growing quickly into large customers.

5. ICONIQ’s Leaner-Team Gap Shrinks From 43% to 21% as Companies Scale

ICONIQ’s GTM headcount by ARR band, AI-forward vs. lower adoption:

• $10M to $25M: 20 vs. 35 FTEs (43% leaner)
• $25M to $100M: 45 vs. 65 (31% leaner)
• $100M to $250M: 125 vs. 165 (24% leaner)
• $250M to $500M: 275 vs. 350 (21% leaner)

AI-forward companies are leaner at every stage, and the difference is largest at the smallest companies. Small deals, self-serve and product-led motions see the most compression. As deal sizes grow, GTM teams move back toward traditional size.

This lines up with how I invest. I don’t invest in an agentic B2B company unless it unlocks budget beyond the software budget, and the clearest evidence is deal sizes at least 2x what pre-AI competitors get. Deals that size involve more stakeholders and longer evaluations, and they are closed by salespeople. That is the segment where the ICONIQ gap is smallest.

The Modern GTM Org in 2026: 20-30% Leaner, 9x Flatter, ~2x More Net New Revenue Per Rep. The Latest from ICONIQ Growth.

Planning Sales at 18% of Headcount in 2026

Across both datasets, the sales org at AI companies looks much like it did before. Most of the change is in the functions around sales.

The differences that show up in the data:

• Higher output per rep. 67% vs. 59% quota attainment at AI-forward teams, on smaller teams overall.
• A smaller SDR layer. 71% of companies have more than half their SDRs using AI daily, the highest of any GTM function. Pipeline is still mostly generated by sales: ICONIQ has 62% of pipeline sales-sourced versus 19% marketing-sourced.
• Much smaller support teams. Support is 59% smaller as a share of the company, and agents in support are now standard at AI hypergrowth companies.
• Post-sales work moving into engineering. FDEs and implementation work are increasingly staffed from the engineering side.

The parts that look the same:

• Sales is still about one in five employees. 18.4% vs. 20.7%.
• Larger deals still require larger sales teams. The efficiency gap at $250M+ ARR is 21%, half what it is at $10M.
• Fast-growing companies are hiring heavily overall. 133% median headcount growth in H1 2026 for the 100%+ growers, with sales keeping its share of that growth.

For a 2026 or 2027 plan, 18% to 20% of headcount in sales is still a reasonable benchmark for a B2B + AI company selling real contracts. The efficiency gains in the data come from reps who use AI and close more, from agents in support and the back office, and from engineers who get customers live on Day 1. None of the AI hypergrowth companies in the data have planned around agents closing their enterprise deals, and a new hiring plan shouldn’t either.