Perplexity chief executive Aravind Srinivas wrote on X on 15 November that the company’s internal use of its Teammate agents is worth “roughly $80M to $180M a year in labor value against about $10M in tokens, so 9x to 19x depending on the job”. The figures are his estimates from Perplexity’s internal analytics; they are not audited, and the valuation method is not published.
On those rounded figures, the labor value per token dollar is $8 to $18, by our arithmetic, a little below his stated 9x to 19x; a token bill near $9.5 million would give 8.4x to 18.9x. Net of the tokens, his range is $70 million to $170 million a year.
Business Pill · Usage-Based Pricing Explained
An evergreen Business Engineer explainer on usage-based pricing, paying for work done rather than for seats, the model behind per-token rates. It teaches the general idea only and does not discuss Perplexity or teammates.
The key insight: As we read it, the claim rests on two numbers Perplexity sets itself: the hours its agents run, which it charts week by week, and the value it puts on those hours, $80 million to $180 million a year against about $10 million of tokens, for which it has published no method.
The 15 November post, at 03:10 UTC, reads in full: “A week later, it’s 11x more agent hours per week from Teammate. Growing more than 50% week over week (was 7x last week). The number of employees getting an extra workday (8+ hours) and extra work week (40+ hours) from Teammate has more than tripled in a week since September. That’s roughly $80M to $180M a year in labor value against about $10M in tokens, so 9x to 19x depending on the job.”
It quotes his post of 6 November, which reads: “The median Perplexity employee now gets 7x more agent-hours per week from teammates than in September. And this has been consistently up every single week. 1 in 5 are already at 40+ agent-hours a week, which is basically a full extra workweek.”
Median agent-hours per active employee per week in Perplexity’s internal workspace, as shown in the chart Aravind Srinivas posted on X on 15 November 2026. Company figures, unaudited; 13.5% is our arithmetic.
What Was Posted on X
A week later, it’s 11x more agent hours per week from Teammate. Growing more than 50% week over week (was 7x last week). The number of employees getting an extra workday (8+ hours) and extra work week (40+ hours) from Teammate has more than tripled in a week since September. That’s roughly $80M to $180M a year in labor value against about $10M in tokens, so 9x to 19x depending on the job.
Perplexity chief executive Aravind Srinivas posted this at 03:10 UTC on 15 November. It gives the 11x figure, the more-than-50% weekly growth and the $80M to $180M labor-value estimate against about $10M in tokens, with the company chart this piece draws on.
The median Perplexity employee now gets 7x more agent-hours per week from teammates than in September. And this has been consistently up every single week. 1 in 5 are already at 40+ agent-hours a week, which is basically a full extra workweek.
The 6 November post the newer one quotes. It gives the earlier 7x figure and says 1 in 5 employees were at 40+ agent-hours a week; its chart also shows the human-typed share of teammate turns.
What the Company Charts Show
The chart attached to the newer post shows median agent-hours per active employee per week in Perplexity’s internal workspace. The median rose from 0.5 hours in the week of 3 September to 5.4 hours in the week of 5 November, and the chart is labelled “Up every single week”.
Two side panels compare the first week with the latest. The share of active employees getting 40 or more agent-hours a week went from 7% to 21%, and the share getting 8 or more went from 16% to 42%.
The 6 November chart had a different panel: teammate actions typed by a human fell from 31% to 22% of all teammate turns, with the rest coming from sub-agents, follow-ups and automations.
The Numbers, by Our Arithmetic
Growth. 5.4 hours divided by 0.5 is 10.8, which rounds to his 11x. 5.4 divided by 3.6, the prior week, is 1.5: the 50% weekly rise, on the chart’s rounded labels.
The median vs the heavy users. A median of 5.4 agent-hours is 13.5% of a 40-hour week. The 21% of active employees at 40 hours or more sit far above that median.
Shares vs counts. The 40-hour share rose 3.0x, from 7% to 21%, and the 8-hour share about 2.6x, from 16% to 42%. His post says the number of employees in those groups “more than tripled”; a count can rise faster than a share if the number of active employees also grows, and neither chart gives that number.
Value per token dollar. $8 to $18 on the rounded $80 million, $180 million and $10 million; $70 million to $170 million a year after the token cost.
Where Teammates Fit in Perplexity’s Agent Push
Perplexity launched teammates on 5 November, rolling out first to Perplexity Max subscribers at $167 a month billed annually, as we reported. The internal series starts on 3 September, two months before that launch.
Outside the company, SemiAnalysis founder Dylan Patel wrote on 11 November that 45% of his firm’s AI spend is now persistent agents, mainly Perplexity Computer. On 5 November, Perplexity said the medium effort level of its Agent API cost almost 90% less while scoring 77% on the BrowseComp and DSQA research benchmarks.
The Structural Read
As we read the two charts together, the growth sits on the machine side: the share of teammate turns typed by a human fell from 31% to 22% while median agent-hours rose, and agent-hours count sub-agents and automations running in parallel.
Aravind Srinivas, Perplexity chief executive, on X, 6 November 2026
“1 in 5 are already at 40+ agent-hours a week, which is basically a full extra workweek.”
Three Implications
FOR TOKEN BUDGETS On the chief executive’s figures, about $10 million of tokens sits against $80 million to $180 million of labor value, $8 to $18 per token dollar by our arithmetic.
FOR AGENT-SEAT BUYERS The posts give figures for Perplexity’s internal workspace only; teammates launched on 5 November for Max subscribers at $167 a month billed annually.
FOR THE METRIC Agent-hours include agents working in parallel, per the chart footnote, so they are machine time rather than a count of human hours.
The framework says the token “is simultaneously the unit of cognition the model produces, the unit of compute the data center serves, the unit of price the lab charges, and the unit of value the enterprise extracts.”
As we read it, the Perplexity posts put a number on the last of those roles for one company’s own staff: $80 million to $180 million of labor value on about $10 million of tokens, by its chief executive’s estimate. The framework predates these posts and does not discuss them.
First week charted (week of 3 September) vs latest, from the charts Aravind Srinivas posted on X on 6 November (human-typed share) and 15 November 2026 (8+ and 40+ shares). Company figures, unaudited; multiples and the 9-point change are our arithmetic.
What Is Not Established
All figures are Perplexity’s internal analytics, posted by its chief executive. None is audited, and the company has not published how it values an agent-hour as labor or which jobs give the 9x and 19x ends of the range.
Agent-hours sum the running time of teammates, their sub-agents and automations, including agents working in parallel, per the chart footnote. They measure machine time, not hours of human work replaced. The post does not say over what period the $10 million in tokens is counted.
The two charts also differ. The week of 29 October is 3.4 hours in the 6 November chart and 3.6 hours in the 15 November one, and the footnotes word the definitions differently: an active employee sent “at least one prompt” in the first and “at least one message” in the second. Neither chart gives the number of employees.
This Week In The AI Supercycle
AI Lost Its Central Banker, the Lab Oligopoly, and the Future of Open Source
This week’s issue: the Nvidia leadership change, read as the AI build-out losing its central banker: who now finances the GPU landlords, backstops the data centres and sponsors open models, and why two labs are becoming the bank.
Perplexity’s chief executive puts the company’s internal Teammate use at $80 million to $180 million a year of labor value on about $10 million of tokens, or $8 to $18 per token dollar by our arithmetic. The median employee’s agent-hours rose from 0.5 to 5.4 a week between the weeks of 3 September and 5 November, per the company chart; the labor-value method has not been published.
A note on sourcing. Every Perplexity figure comes from two posts on X by its chief executive, Aravind Srinivas, and the charts attached to them, and is labelled by Perplexity as internal analytics; the per-token-dollar values, net figures, ratios and the share of a 40-hour week are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
