Posted on: 26 October 2026
Gainsight is a platform that software companies use to follow their own customers after the sale, measure how much they use the product and predict who will renew. On 20 October it announced that from next year it will no longer charge per user but in proportion to the customer's recurring revenue and the number of accounts managed. In the same month Microsoft switched on consumption billing by default on new Copilot licences for small and medium businesses: a flat fee per person remains for the everyday functions, while agents that work on their own burn credits billed separately. Two small signals that tell one large story.
The SaaS model, meaning software sold as a subscription rather than a perpetual licence, never billed for value, but it billed for something that could be counted: the person sitting in front of the screen. The seat was never the right measure, it was the convenient one, the one a vendor and a procurement department could agree on without a fight, because all it took was counting heads. An industry worth hundreds of billions rests on that convention, while AI agents, programs that carry a task from start to finish with nobody logged in, remove precisely the chairs. Deloitte expects the share of revenue tied to the seat to fall from 21 to 15 per cent of the total over the next few years. Intercom, which sells customer support software, has already stopped counting operators and charges 99 cents for every conversation the machine resolves, while Salesforce asks two dollars a conversation for its agents.
There is a precedent all of us have paid on a phone bill. The minute of a call made sense when a call occupied a physical circuit between two exchanges, so that time really did measure a resource consumed. When voice moved onto packet networks the circuit vanished, yet operators went on billing minutes for years, because it was the unit customers and regulators knew how to read, until flat-rate plans closed the matter. People were paying for an object that no longer existed because nobody had yet invented the new unit of measure. It was not a scam, it was a bridge, but whoever crossed it paying should have known.
The detail that gives the real incentives away sits in an April survey of 300 chief executives of software companies: 97 per cent want to abandon seat pricing within two years, yet 94 per cent say the seat represents the value of their product well. Read together, the two figures say they are not running because the measure is wrong, but because the customer's procurement department has worked out that with agents the seats go down and will use the next renewal to cut. The change of model is not born of vision, then, but of fear of a negotiation.
The meter can run backwards too, and I have watched it happen as a customer. Wodify, the management software we use in our CrossFit gyms in Switzerland, billed per active member for years, until at some point we discovered that the active members by their definition were quite a few more than the active members by ours, meaning those who paid and actually trained, while the difference was paid by us. After repeated arguments with customers they dropped the count and moved to a flat fee on three tiers, with add-on modules priced separately. It looks like the opposite of Gainsight's direction, whereas it is the same rule read from the other side of the table: when the unit of measure is defined by the vendor alone and the customer cannot reproduce the count, trust in the meter collapses and everyone goes back to the unit everybody knows how to read, even at the cost of a cruder measure. The flat fee is not the right measure, it is the measure nobody can cheat on.
Gainsight's move should be read in that light, which is to say that tying the price to the customer's revenue means ceasing to sell software and starting to take a percentage, like the agent who takes ten per cent of an actor's earnings or the advertising agencies that for decades kept fifteen per cent of the media they bought. That model held until the client did the sums and found that the growing was theirs while the collecting was somebody else's. In the nineties the large advertisers dismantled the fifteen per cent commission within a few years, moving to fees per project, and nobody has gone back since.
What not to do, if you buy software for a company that is growing. Do not sign multi-year seat renewals when the headcount using that tool is shrinking, because you are committing to pay for empty chairs. Do not accept a price pegged to your revenue without a cap, because you are handing the vendor an option on your growth without them sharing the risk when revenue falls. Do not leave the vendor to define "outcome" or "active user" when they propose that you pay by consumption, because whoever controls the meter controls the price, as we learned in our gyms.
The transition, if history is any guide, will be long and full of hybrids like Microsoft's: a flat fee that lets whoever signs off the budget sleep at night, while the meter beside it does the real work and produces the vendor's growth. That meter holds only as long as the customer can check it against their own numbers, otherwise the cycle turns back the way it did for us. Buyers would do well to look at the meter before the fee, because software has entered its years of the minute billed for nothing, and like telephony it will only end when somebody finds a way to count the right thing.