Through the first half of 2026, the public markets have begun to re-price enterprise software in a manner that the median analyst report has not yet caught up to. The median multiple is flat year-over-year. The dispersion of the multiple is the largest it has been in fifteen years.
A handful of categories trade at premiums that price in not only growth but the assumption of agentic compounding; a larger group trades flat; a third group trades at discounts that price in not only slower growth but a thesis that the seat-based licensing model is ending. The thesis is not the consensus. It is becoming the consensus.
The Signal
The shape of the dispersion is the signal. In ordinary years, the distribution of enterprise software multiples is narrow — the category trades as a category, give or take quality. In 2026, the distribution is bimodal, and the gap between the two modes is widening month over month. The public markets are no longer pricing enterprise software as one thing. They are pricing it as two.
The Mechanic
The seat-based licensing model of enterprise software priced the user. Every additional employee using the system was an additional licensable unit, and the software vendor's revenue grew with the customer's headcount. The model required no opinion about what the user did with the software. The opinion was the user's; the software's responsibility was to be present.
The agentic model of enterprise software is changing what is licensed. The agent — not the user — is the unit of work. The agent's volume is not bounded by the customer's headcount; it is bounded by the customer's appetite for work the agent can perform. A category in which the agent can perform work the customer would otherwise have paid a human to do becomes, in the agentic model, a beneficiary of the change. A category in which the agent threatens to subsume what the software previously charged for becomes, in the agentic model, a casualty of the change.
The Categories That Compress
The categories whose multiples are compressing share a feature: the work the category sold to the seat is increasingly performed by the agent, and the seat is therefore a declining unit of value.
The category of seat-based productivity tools — calendar, conferencing, generic horizontal collaboration — is exposed because the agent is now able to perform the discrete work the tool previously charged for. The tool's value migrates upstream, into whichever surface the agent runs in.
The category of seat-based specialist tools whose specialist is now performing work in the agent's company — drafting, code generation, design — is exposed for the same reason. The specialist did not stop using the tool; the specialist's volume fell, and the seat count fell behind it.
The category of seat-based vertical workflow tools whose vertical is the customer's labor — first-line customer service, basic financial close, certain elements of legal review — is the most exposed of the three. Here the agent performs the work that the seat enabled. The vendor's defensive thesis — we will add agents to our product and price the agent — is a defensible thesis, but it converts a high-multiple recurring revenue stream into a lower-multiple usage-based one. The market is pricing the conversion.
The Categories That Expand
The categories whose multiples are expanding share a different feature: the agent's work flows through the category, and the category becomes the metering layer for the agent's output.
The category of data and context infrastructure — the systems that store, organize, and make discoverable the corpus the agent reasons over — is expanding because every agent the customer deploys is a marginal demand for the infrastructure. The customer's spend on context grows even as the customer's spend on seats falls.
The category of governance, audit, and observability — the systems that capture, evaluate, and reconstruct the agent's behavior — is expanding because the new fiduciary obligations described elsewhere in this series are creating a regulated requirement for the spend. The category was, in 2024, an adjacent purchase. It is becoming, in 2027, a required purchase.
The category of orchestration — the systems that connect agents to one another, to the customer's data, and to the customer's existing systems of record — is expanding because the customer cannot extract value from any single agent without it. The orchestration layer is, in the language of the prior brief, the platform. Customers who underinvest here pay for the underinvestment in their next forty pilots.
The category of vertical specialist software whose data is rare and whose workflow is regulated — certain corners of healthcare, certain corners of financial services, certain elements of industrial operations — is expanding because the agent cannot replicate the category's data or the category's compliance footprint. The category becomes the agent's host, and the host charges accordingly.
What the Strategic Owner Should Do
The implications, for chief financial officers and strategic owners of enterprise software exposure, are four.
First, the inventory of the firm's enterprise software spend should be re-categorized along the lines above — not by vendor, but by exposure to the agentic shift. The exercise will reveal, in most firms, a portfolio that has more compressing exposure than the chief financial officer would have guessed.
Second, the firm's investment in the expanding categories — context infrastructure, governance, orchestration — should be funded in advance of need. The categories are repricing upward; the cost of delay is the cost of buying the same capability later, at a higher multiple, under regulatory or competitive pressure.
Third, the firm's contracts with the compressing categories should be examined for term, price, and renewal mechanics. Multi-year contracts struck at 2024 seat prices are, in 2026, a balance-sheet item with a quietly impaired carrying value. The chief financial officer who has not modeled the impairment may discover it at renewal.
Fourth, the firm's strategic owners should be asked the harder question: which of the firm's competitive advantages was a function of the seat-based licensing model of our enterprise software, and which would survive a transition to agentic? The honest answer, in many industries, is that some advantages — speed, leverage, throughput per employee — were a function of how the firm spent on software and not of what the firm was. The advantages will not survive the transition unless the firm re-establishes them under the new economics.
The re-rating of the enterprise software multiple is not an event. It is a process, and it is already underway. The chief financial officers who will be remembered as having read the half-decade correctly will be the ones who, in the early months of 2026, treated the dispersion as the signal it is.