Full narration of the briefing. Click a section heading to jump the player to that moment.
1 billion dollars for Sana, and began repositioning its entire AI strategy around it: away from features embedded inside modules, toward a single interaction layer that sits on top of your tenant, searches your enterprise knowledge, and takes actions across systems. It is a genuinely interesting product. It is also, commercially, an undefined one: Workday has not settled how Sana will be packaged or priced long term. What is defined is the meter underneath it, Flex Credits.
This briefing explains how that meter actually works, what it costs in practice, and a framework for deciding when to commit, using evidence instead of enthusiasm.
First, be precise about the product. Sana is a front door for work: one interface over your Workday data and beyond it. The former Illuminate branding continues underneath as the governance layer, but the commercial center of gravity has moved to Sana, and pricing is expected to drift toward models that scale with usage and adoption. Recognize the strategic frame too: Workday is competing to become the primary AI interface of the enterprise, which means broad deployment is worth more to the vendor than to you in the early phase.
That asymmetry should shape every commitment you make this cycle. An undefined product deserves a defined contract.
Second, the meter. Every agent action consumes Flex Credits, and the published action costs are worth memorizing. A simple self-service retrieval: one credit. An autonomous task completion: five.
Grading a single resume: about six. And one Talent Rediscovery run: seven hundred fifty credits per requisition. Your organization receives a complimentary annual allotment sized by employee band, from fifteen thousand credits for smaller enterprises up to two hundred thousand at the very largest. Now the asymmetry: the credit cost of each action is published, but the dollar price of a credit is not.
It is negotiated customer by customer. You are budgeting consumption of a currency whose exchange rate is set privately. That is not a reason to avoid the product. It is a reason to set the rate in writing before you depend on it.
Third, run the burn model, because the arithmetic surprises people. Take a real profile: thirty-four thousand employees, one hundred twenty thousand complimentary credits a year. Recruiting grades twelve thousand resumes: at six credits each, that is seventy-two thousand credits. Talent Rediscovery runs on sixty-four requisitions: at seven hundred fifty each, another forty-eight thousand.
Those two use cases alone consume the entire annual allotment before any other agent takes a single action, and with API overages and storage, this profile ends the year roughly seventeen thousand credits short, purchased at whatever rate was, or was not, locked. Three operational notes. Credits expire annually and do not roll over. They burn only in production, so sandbox piloting is free, with one exception, the BP Optimize agent, which consumes everywhere.
And the Platform Consumption Console exists for exactly this: stand it up with thresholds, alerts, and one named owner before agents go live.
Fourth, the protections, in priority order. One: the dollar-per-credit rate for additional credits, locked in writing for the full term, not quoted at the moment of shortfall. Two: carryover rights for unused credits, or at minimum a credit against next year's purchase. Three: a hard consumption cap with notification before any overage accrues.
Four: extension of the API billing exemption that currently runs to January 31st, 2027. Five: an escalation clause, if consumption runs more than twenty percent above forecast for two consecutive months, the parties meet and the rate reopens. And six: remember that the updated master agreement itself is the gate, required for new and renewing customers and for anyone exceeding complimentary credits. That signature is leverage.
Trade it for the protections on this list, not for goodwill.
Fifth, the decision itself. A three-gate framework. Gate one: pilot inside the free window, core HCM and Financials customers have Sana and the Self-Service Agent at no cost through August 31st, 2026, but treat the pilot as measurement, not adoption theater: instrument consumption from day one, because your own usage curve is your negotiating baseline. Gate two: if the pilot earns it, buy on separate paper: Sana and credits commercially severed from the core renewal, with their own exit rights, price protections, and proof-of-value milestones tied to measurable outcomes, not activity counts.
Gate three: scale only on data. The history of the Workday platform argues for patience here: roughly two-thirds of Extend and Prism estates were bought above forecast and underutilized at first renewal. The buyers who did best bought those platforms at the second renewal, on evidence. Sana will reward the same discipline.
Enthusiasm is not a business case. A usage curve is.
The summary is simple. Learn the meter, run the burn model on your own numbers, lock the rate before dependence, and let evidence set the pace of commitment. If you would like the burn model run against your actual Workday census and recruiting volumes, that is work we do at Redress Compliance, on pure contingency: our fee is 25 percent of what we save you, and if we save you nothing, you pay nothing. com.
This concludes the three-part Workday Buyer Briefing: the contract, the renewal, and the AI decision.
Redress Compliance works on contingency: our fee is 25 percent of what we save you. Nothing saved, nothing paid. Independent, buyer side only, never vendor funded.
Talk to a Workday Deep Dive · Part 3 negotiator