Agentic Layer Check
Named companies, read against the nine layers of the Agentic Commercial Model: what's working, where there's room to extend it, and how each finding ripples through the rest of the model. Public signal only: what's on the page, not what's in the contract.
An allowance quoted in real work and shared across the whole company, two things almost nobody else does, set at a level that empties in an afternoon and upgrades into no meter at all.
More capacity handed back at the same price and a meter that only counts the expensive work, sitting next to a sales motion that still closes seats while the growth lands on a meter nobody sold.
A meter that splits on who picks the model, and a cost saving handed back to customers as capability, sitting next to monthly balances that pool with nobody and carry nothing.
A published per outcome rate card and spend caps switched on by default, next to a seat line the customer can convert into credits mid contract and a credit balance that lapses every month.
A decade old meter that still only bills work that worked, and a real overflow valve when the limit hits, sitting next to a second meter for agents that has neither.
A meter that charges for work completed rather than attempts, and a half finished build billed at half, sitting next to a goal run whose price the customer cannot know before committing to it.
A genuinely seat-tax-free meter that counts executions, not headcount — sitting next to an execution unit that bills a two-node webhook exactly the same as a forty-step, thirty-LLM-call AI agent.
A clean split between platform cost and data cost, a 0% markup passthrough on reasoning models, and a single credit balance that now has to carry two very different margins at once.
A genuinely good fix for outcome based AI metering, sitting inside the most fragmented packaging surface of any major platform vendor, and a three week old partnership with Anthropic that adds a construct Salesforce itself doesn't invoice for.
One of the cleanest outcome based AI meters from a major platform vendor, sitting right next to a packaging story that's starting to fragment into three separate mechanics.
Every Agentic Layer Check is a public signal read, built from a company's own pricing pages, product pages, and public statements, dated to when it was written, never from private briefings or company input. Companies covered are not consulted beforehand, not paid for inclusion, and inclusion doesn't imply a relationship with or endorsement of this work. Pricing pages change; where a company updates its packaging after a piece is published, that will be noted rather than silently edited away.