The Companies Selling You AI Are Shrinking Their Own Teams First

Appier reported its June quarter on August 13. The Tokyo-listed company sells AI marketing software, most of its revenue coming from Northeast Asia, and the release carried the numbers you’d expect: revenue of JPY 12.9 billion, up 24.6% on the year, and gross profit of JPY 7.7 billion, up 33.5%.

Then it published a figure most software companies keep to themselves. Gross profit per employee, up 38% YoY.

Call it profit per head. It’s the amount of margin each person on the payroll generates, and Appier put it in a shareholder release because it now believes that number is the proof. The company credited it directly to its own use of agents, meaning AI that carries out multi-step work on its own rather than answering one question at a time.

“Agentic AI adoption across R&D has shortened development cycles and accelerated product improvements, driving the quarter’s margin expansion and record core free cash flow.”

Gross margin crossed 60% for the first time in the company’s history, landing at 60.1%. Operating profit rose 82.8%. Core free cash flow more than tripled. Chief Executive and co-founder Chih-Han Yu framed the quarter as an answer to the skeptics:

“While the broader market remains focused on AI’s potential, Appier’s results demonstrate proven, measurable P&L impact.”

Customer count grew 12% over the year. Revenue per customer grew 11%. Neither of those explains a 38% jump in profit per head. The denominator did most of the work.

Same week, same thesis, opposite direction

Sprout Social held its own earnings call the same day, August 13. The social media management platform, which brands use to schedule and measure their posts, spent part of it explaining a decision its board had approved on July 8.

The company is cutting about 20% of its workforce, roughly 260 people. Chief Executive Ryan Barretto put it to analysts plainly:

“As we announced on July 15th, we made the very difficult but important decision to reduce the size of our team by approximately 20%.”

Asked why, he pointed at the org chart, not the technology:

“We saw too many layers and an org structure that was slowing down decision-making, and saw an opportunity to really streamline our work and areas of focus.”

The filing tells it differently. Sprout’s 8-K, the disclosure a public company files when something material happens, says the plan will:

“streamline the Company’s organizational structure and align its cost base with its strategic priorities, including its ongoing investments in AI-powered social intelligence”

That’s the sentence that matters. On the call, the cut is about org layers. In the legal filing, the cost base is moving to fund AI. Both can be true, and companies word these things carefully for good reason. Only one of the two documents carries liability.

The restructuring will cost $18 million to $20 million up front. It’s targeted to take $50 million a year out of the cost structure, with the full benefit landing in 2027.

Matt Navarra, a social media industry analyst who writes the Geekout newsletter for a stated audience of 30,000+ marketers, creators and social media leaders, made the observation the day after the cut was announced:

“Social media tools are selling AI as productivity magic, while the workers behind those tools are learning what ‘efficiency’ really means.”

The customers aren’t buying it yet

HubSpot reported on August 5 and supplied the other half of the picture. It cut its expectation for new customers from 9,000 to 10,000 a quarter down to roughly 5,000 to 6,000 for the back half of the year.

Chief Executive Yamini Rangan explained the slowdown as a change in how software gets bought:

“traditional software used to be about customers looking at features, implementing features and then making sure that users got trained on those features to drive outcomes. But with AI, the real change is that we deliver outcomes. And they got to make sure that it works within their environment with their data, and they got to understand the ongoing economics before they commit.”

Buyers now run trials, demand proof, and route the decision up to finance. HubSpot moved several of its agents to outcome-based pricing this summer, which we covered when it started charging 50 cents only when the AI resolves a ticket.

And HubSpot is running the same play on itself. Rangan told analysts the company had reorganized into smaller teams on six-week sprints, and put it in one line:

“We’re doing more with less and getting faster as we go, and it is showing up in the numbers.”

Chief Financial Officer Kathryn Bueker named where it shows up. Operating margin rose 3 points, and she attributed the expansion to “our continued disciplined approach to head count spend, partially offset by AI costs.” She also guided to another 2 to 3 points of margin expansion in 2027, which Rangan tied to “the operating leverage we are building as an AI-first company.”

So the vendors are getting leaner on the promise of AI, and their customers are getting slower about paying for it.

Update, August 18: the ad platforms are on the same script

Cardlytics, which places offers inside banking apps using purchase data, reported on August 12 and added a fourth data point at a smaller scale. Revenue fell 36%, most of that from restructured bank partnerships. Adjusted operating expenses fell 31%, which Chief Financial Officer David Evans said was “largely due to reduction in force actions taken in the second half of 2025 and optimization of our cloud infrastructure.” Asked where headcount goes from here, he expects it “staying the same.”

Chief Executive Amit Gupta told analysts the platform rebuild behind that math is finished:

“Last year, we invested in cleaning up our tech debt and building an AI-forward tech stack. Now that these investments are behind us, we are now operating more efficiently and moving faster.”

Meanwhile new advertiser signings grew 59% QoQ. Fewer people, flat hiring plans, more customers. It’s the same shape Appier put a number on.

Your budget went the other way

Here’s the part that should stop a marketing leader mid-scroll. While these three companies were reporting tighter headcount and higher output per person, the average marketing organization did the opposite.

Gartner’s 2026 CMO Spend Survey was fielded from January through March among 401 marketing leaders across North America, the UK and Europe, most of them at companies above $1 billion in revenue. It found that labor’s share of the marketing budget rose from 21.9% in 2025 to 24.5% in 2026. AI took 15.3% of the budget in the same year. While 70% of those CMOs called AI leadership a critical goal, only 30% said their organization was ready to scale it.

Marketing teams are spending more on people and more on AI at the same time. Their software vendors are spending less on people and reporting the result to investors every quarter.

That gap won’t stay open. Appier didn’t invent profit per head, but it has now demonstrated that a marketing company can publish it, credit AI for it, and get rewarded. Public companies copy each other’s disclosures faster than they copy each other’s products, and a metric that flatters one CEO’s quarter tends to show up on the next one’s slide.

What to do before someone asks

Know your own number before finance works it out for you.

Pick a denominator you can defend. Pipeline per marketer, qualified leads per marketer, revenue influenced per marketer, published assets per marketer. Measure it across the last four quarters rather than this one, so you hand over a trend instead of a snapshot. Then measure what your AI spend moved, and say so plainly when the answer is nothing. The Gartner number says two thirds of your peers are in that same position, and we found the same shape in the gap between AI budgets and what teams can prove.

The vendors have already run this exercise on themselves. Appier ran it and published the win. Sprout ran it and cut 260 jobs. Both companies had the same technology available to them. What separated the two outcomes was timing: whether output per person moved before the pressure arrived.