01 · The $193M Benchmark

For fifty years, enterprise employees used software to do their jobs: developers used IDEs, sales representatives used CRM, analysts used BI tools, and accountants used ERP. Software's value depended on how much it helped those employees accomplish.

Executives now also buy AI products to perform work previously funded through headcount or outside services. CFOs evaluate sales-development agents; engineering leaders buy Cursor and Claude Code, increasingly judging value by pull requests rather than developer hours. General counsel buy Harvey's contract review by matter, and support leaders buy Sierra by resolved interaction. These examples move purchasing toward work delivered.

For fifty years, an additional software-company employee commonly corresponded to roughly $1 million to $5 million in enterprise value. Gross margins and recurring revenue supported the upfront expense, while investors priced future cash flow. AI-native businesses change that relationship.

The report's contact-center example uses SaaS seats at roughly $1,000 to $5,000 per agent annually and a 500-agent account with at most $2 million ARR. At $1 to $3 per resolution across five million annual contacts, AI pricing produces $5 million to $15 million ARR against the same customer's $25 million labor budget. The vendor gains revenue and the buyer can save costs, while effects on displaced labor remain outside the software contract. The example illustrates why work pricing can exceed seat pricing.

Two effects are visible in the register. AI competition weakens the revenue expansion required by leveraged PE software deals. Private AI companies also limit public access and LP liquidity. Piece 2 examines both through ownership cohorts.

As buyers and the work purchased change, vendor economics and valuation comparisons change with them.

Across 75 years, eight software waves, 2,398 companies, and $28.2T of value, value per full-time employee provides a common comparison. The cohort benchmark stayed between roughly $2.7M and $4.3M for forty-three years across mainframe, client-server, on-prem, and SaaS businesses.

It reached $18.3M in the seven years since 2019, roughly three and a half times the cloud-native cohort's $5.2M. The AI-native cohort contains 413 companies founded in 2019 or later.

The register below lets readers compare all 2,398 leaders by founding era, ownership, and value. The series examines how three of the four ownership cohorts face different financial constraints.

Value per employee provides a financial comparison for vendors, especially in PE-held portfolios approaching refinancing. In our view, the new benchmark is already affecting how investors and buyers assess existing software businesses.

Over the next three years, we expect those differences to affect which of the 2,398 leaders retain their place in customer budgets.

Company workflows help explain our interpretation of the valuation differences.

Boris Cherny described his Claude Code workflow at Sequoia AI Ascent on May 4, 2026: five to ten parallel sessions, often monitored from his phone, with a few thousand sub-agents working overnight and twenty to thirty merged pull requests on a typical day. He reported roughly seventy percent growth in Anthropic's per-engineer productivity over 2026 while headcount tripled, and estimated Claude Code's involvement in four percent of public GitHub commits. These are reported figures from that presentation.

Cursor is valued at roughly $60 billion with about 700 employees. Harvey prices legal contract review by matter. Both illustrate the shift toward buying completed workflows discussed above.

With roughly 5,000 employees, Anthropic's reported value is more than four times IBM's at about two percent of its headcount. The $193M-per-employee ratio is a valuation comparison; the workflows above offer context for how the company operates.

The cohort trend matters alongside the individual company examples.

The report values OpenAI at $852B, Anthropic at $965B, and xAI at $250B (now part of the merged X/SpaceX entity), totaling roughly $2.1 trillion with about 15,100 employees. Twelve SaaS pure-plays founded between 1999 and 2009 total about $794 billion: Salesforce, Workday, ServiceNow, Shopify, Atlassian, HubSpot, Okta, Twilio, Veeva, Dropbox, Box, and Zendesk. The comparison places the AI cohort at about 2.6 times their value in under a decade, with about a fifteenth of the people.

Those figures provide a comparison for boards reviewing software vendors.

The $193M-per-employee ratio reflects more than the abilities of individual staff. Our interpretation is that Anthropic uses the agentic systems it sells across engineering, sales, finance, marketing, and HR. Software takes on functions previously filled through headcount, including coordination work traditionally handled by middle managers.

Some leaders are making headcount cuts of 20 to 40%, with substantial backlash. The report describes a tenfold change in the productivity benchmark and an eighteen-to-twenty-four-month window to grow value toward the new benchmark or reduce costs. We expect the leaderboard to record uneven outcomes over the next five years.