Workday moved AI to a consumption model in 2025. Flex Credits ship inside every subscription and meter per agent skill. Read the pillar before your next Workday renewal prices the agents in.
Workday Flex Credits are a consumption model, included in every subscription and metered per agent skill on a rate card. The pillar covers the mechanics, what is bundled, how Illuminate and Sana agents burn credits, how the Agent System of Record governs them, and the buyer side levers at renewal.
Workday spent two decades pricing on named workers and modules. In 2025 it introduced a second axis. AI now bills on consumption, and the unit is the Flex Credit. The shift matters because a subscription line that reads as fixed now carries a variable tail.
The design goal Workday states is simple. Align cost to the value an agent delivers, measured by outcomes, not by tokens or seat counts. That framing is reasonable. It also moves the cost risk. When you meter by action, the buyer owns the volume, and volume is the one thing a vendor estimate cannot predict for your estate.
Flex Credits are a consumption based currency, included in every Workday subscription, that you spend across AI agents and platform innovations as you use them. Each metered action draws a set number of credits from your balance.
Workday meters credits per agent skill. A skill is a discrete capability an agent performs. The Flex Credits model assigns each metered skill a fixed credit value tied to the work it does.
The rate card is the price list. It maps each metered skill to a credit value based on estimated time saved and efficiency gained. A Self Service Agent retrieval skill draws about 1 credit per action. An autonomous task completion skill draws about 5 credits per action.
Workday meters by task completion and specific actions rather than tokens consumed or minutes of interaction. The stated intent is to tie cost to delivered value. The practical effect is that an agent that finishes multi step work costs more per run than one that answers a question.
Illustrative Flex Credit draw by skill type
| Skill type | Example action | Illustrative credits per action | Cost driver |
|---|---|---|---|
| Retrieval | Self Service Agent answers a policy question | About 1 | Low, single step |
| Guided task | Agent drafts a document for review | 2 to 3 | Medium, human in the loop |
| Autonomous completion | Agent completes a task end to end | About 5 | High, multi step |
| Industry skill | Purpose built agent runs a domain workflow | Varies by rate card | Depends on scope |
Every Workday subscription includes an annual allotment of complimentary Flex Credits, sized to company headcount, plus access to the agent catalog and the Agent System of Record. The consumption exposure begins only above the allotment.
Workday grants an annual allotment of complimentary credits based on company size. The stated purpose is to let buyers test and explore agents in production, generate real usage data, and budget with confidence. Treat this window as free telemetry, not free capacity.
Credits apply universally across agents and platform innovations. That means the same balance funds an Illuminate HR agent, a Finance agent, and newer AI features as Workday ships them. One currency, many draws.
Workday designed the allotment to grow as customer needs evolve. That flexibility is a selling point. It is also a reason to fix the growth terms in writing, because a balance that grows on the vendor's terms is not the same as one that grows on yours.
Illuminate and Sana agents both meter through the same Flex Credit rate card, so the credit draw depends on the skills each agent runs rather than which brand it carries. The mix of skills is what sets the burn.
Workday expanded Illuminate in September 2025 with new agents for HR, Finance, and industry. These agents are embedded in the flows people already use and are powered by Workday data and context. Each metered skill they run draws its rate card value.
Workday completed its acquisition of Sana in November 2025, adding AI powered search, agents, and learning. Sana agents streamline workflows and run under the same governance and metering as first party agents.
Two estates with the same agent count can burn very different credit volumes. An estate that leans on retrieval skills sits near 1 credit per action. An estate that leans on autonomous completion sits near 5. The Workday AI agents catalog shows how broad the skill range now is.
The standard account team line is that AI is included, so there is nothing to price and nothing to worry about. We disagree. In roughly half of the Workday estates we benchmarked in 2024 and 2025, the complimentary allotment was exhausted before renewal once agents ran in production, and the buyer then faced a consumption true up on a rate card they never negotiated. The word included describes the base allotment, not the ceiling. The buyer side move is to treat the complimentary window as a measurement period, capture real burn, and pin the dollar per credit rate and the per skill values in the order form before signature, not after the meter has already run.
Source: Redress Compliance advisory engagement file, 2024 to 2025.
Included is a description of the floor, not the ceiling. The moment an agent runs autonomously in production, the meter is on, and the rate card you never negotiated becomes the bill you have to pay.
The Workday Agent System of Record is the control plane that registers, governs, and monitors every AI agent in the estate, whether it comes from Workday, a partner, or a third party. It brings the whole agent fleet under one accountable system.
Workday describes the Agent System of Record as the place to manage the entire fleet of agents in one location. It provides centralized governance, security, compliance, and visibility across every agent the organization runs.
Third party agents, with or without access to Workday data, are registered and monitored through the Agent Gateway. Governance is not limited to Workday delivered agents. Partner and third party agents fall under the same accountability at the skill level.
Workday unveiled the Agent System of Record in February 2025 and moved it to general availability later that year. It reframes agents as a managed workforce with roles, cost, and oversight.
Governance and cost are the same lever here. An agent you cannot see is an agent whose credit burn you cannot forecast. Registering every agent through the gateway is what makes the burn model complete and the renewal defensible.
Forecast burn by mapping each active agent to the skills it runs, multiplying expected monthly actions by the per skill credit value, and comparing the total against the complimentary allotment. The complimentary window exists to give you that data.
List every registered agent and the metered skills it invokes. Use the Agent System of Record as the source of truth so third party agents are counted, not missed.
Estimate monthly actions per skill from real usage in the complimentary window. Separate retrieval actions from autonomous completions, because the credit draw differs by roughly five times.
Total the credit draw and compare it against the complimentary allotment. The gap between projected burn and free allotment is the number you negotiate the paid tier around.
A worked example makes the math concrete. Take an estate with 5,000 employees running two agents in production over a month.
Illustrative monthly credit burn for a 5,000 employee estate
| Agent and skill | Monthly actions | Credits per action | Monthly credits |
|---|---|---|---|
| Self Service Agent, retrieval | 40,000 | 1 | 40,000 |
| Self Service Agent, guided draft | 6,000 | 2 | 12,000 |
| Finance Agent, autonomous completion | 4,000 | 5 | 20,000 |
| Total | 50,000 | Blended 1.44 | 72,000 |
The same 50,000 actions cost 72,000 credits, not 50,000, because the autonomous skill draws five times the retrieval rate. The blended cost per action is what matters, and it moves with the skill mix.
Now project the year. At 72,000 credits a month the estate draws about 864,000 credits over twelve months. If the complimentary allotment covers, say, six months of that pace, the paid tier has to fund the rest. That gap is the number to negotiate, and it only appears if you measure the mix rather than the headcount.
Run the forecast a second time with autonomy dialed up. If adoption pushes autonomous completions from 4,000 to 12,000 a month, the monthly draw jumps to well over 100,000 credits. A small shift in how people use the agents produces a large shift in cost. Model the heavy case before you sign, not after.
The strongest levers are the dollar per credit rate, the per skill values, rollover, allotment size, and a true up cap. Fix these in the contract because the consumption model shifts volume risk onto the buyer.
Workday has not published a universal dollar per credit price. That means the rate lives in your contract and is negotiable. Fix it, and fix how it escalates over the term.
Rate card values for the same named skill varied across the contracts we reviewed. Treat the per skill credit values as negotiable inputs, not fixed list prices.
Ask for rollover of unused paid credits and a cap on any retroactive true up. Without both, a spike in autonomous agent use late in the term can produce a bill you never modeled.
The complimentary allotment is set by company size, but the exact number and how it grows belong in the contract. Confirm the allotment for your headcount, confirm the annual reset, and confirm what happens to the allotment if headcount changes mid term.
Ask for consumption reporting and threshold alerts as a contractual right, not a courtesy. You cannot manage a meter you cannot read. Alerts at 60 and 85 percent of the allotment give the buyer time to act before an overage lands.
The consumption model changes a Workday renewal by adding a variable line to what used to be a fixed subscription. The renewal is no longer only about seats and modules. It is also about the rate card and the burn.
Named worker counts and module scope still drive most of the contract value. Do not let the AI conversation distract from the core commercial review. Rightsize the fixed base first, then layer the consumption terms on top.
The new work at renewal is the Flex Credit rate card, the allotment, and the burn forecast. Treat these as a separate negotiation track with its own numbers. A strong fixed deal with an open ended meter is not a strong deal.
Begin the burn measurement at least two full quarters before renewal. The complimentary window is your data source, and it only helps if you capture usage while it is live. A late start leaves you negotiating the meter on the vendor's numbers.
Workday Flex Credits are a consumption based currency included in every Workday subscription and spent across AI agents and platform innovations. Credits are metered per agent skill on a Flex Credit Rate Card, so cost tracks task outcomes rather than user counts or tokens.
No. An annual allotment of complimentary Flex Credits is included with every subscription based on company size. The exposure begins only when consumption exceeds that allotment and you buy additional credits, so the model is included at the base but metered above it.
A Flex Credit is consumed when a metered agent skill runs. Each skill on the rate card is assigned a fixed credit value tied to the work it does. A Self Service Agent retrieval skill draws about 1 credit per action while an autonomous task completion skill draws about 5 credits per action.
Workday has not published a universal dollar per credit price. The conversion sits inside a customer specific rate card negotiated in the subscription, which is why the buyer side move is to pin the credit to dollar rate and the per skill rates in the order form before signature.
The Workday Agent System of Record is the control plane that registers, governs, and monitors every AI agent in the estate, whether built by Workday, a partner, or a third party. Third party agents connect through the Agent Gateway and are governed at the skill level alongside first party agents.
Illuminate agents embedded in HR and Finance and the Sana agents Workday acquired in 2025 both meter through the same Flex Credit rate card. Each skill they invoke draws its assigned credit value, so an estate that leans on autonomous multi step skills burns credits several times faster than one that only runs retrieval.
Forecast burn by mapping each active agent to the skills it runs, multiplying expected monthly actions by the per skill credit value, and comparing the total against the complimentary allotment. Use the complimentary period to capture real usage data before the allotment expires, then size the paid tier against observed burn rather than a vendor estimate.
The main levers are the dollar per credit rate, the per skill credit values on the rate card, rollover of unused credits, the size of the complimentary allotment, and a cap on retroactive true up. Fix these in the contract because the consumption model shifts cost risk onto the buyer once the allotment runs out.
Rollover is a negotiated term, not a guarantee. Complimentary credits are designed to be used inside an annual window, so a buyer who wants unused paid credits to carry forward must secure rollover language in writing before signing the order form.
Rate card mechanics, complimentary credit math, the agent burn forecast model, and the buyer side moves for the next Workday renewal.
Used across more than five hundred enterprise engagements. Independent. Buyer side. Built for procurement leaders running the next renewal cycle.
Engage independent buyer side Workday licensing experts. We do not resell. We do not implement. We sit on your side of the table.
Open the Workday Practice page
See engagement scope, comparison versus resellers, and the buyer side framework.
Visit page →The consumption model does not raise your price on day one. It moves the volume risk onto your side of the table. The buyer who measures burn early is the buyer who signs on their own numbers.
