Forecast · Workday · Illuminate
Workday Illuminate agent cost forecast
Illuminate agents meter on Flex Credits per skill. This guide gives a step by step burn forecast a buyer can run in the complimentary window and take into a renewal.
Illuminate agents meter on Flex Credits per skill. This guide gives a step by step burn forecast a buyer can run in the complimentary window and take into a renewal.
Forecasting Illuminate agent cost means turning agent skills and action volume into a monthly Flex Credit burn. This guide gives a step by step model a buyer can run during the complimentary window and take into a renewal.
You build the forecast by mapping each Illuminate agent to its skills, estimating monthly actions per skill, and multiplying by the credit value from the Flex Credits rate card. Workday expanded Illuminate across HR, Finance, and industry in 2025, so the agent list can be long.
List every active Illuminate agent and the metered skills it runs. Use the Agent System of Record as the source so nothing is missed.
Estimate actions per skill from real usage, separating retrieval from autonomous completion. The Workday AI agents catalog shows which skills each agent exposes.
Multiply actions by the per skill credit value and sum across agents. That total is the monthly credit burn.
Illuminate monthly burn forecast, worked example
| Illuminate agent and skill | Monthly actions | Credits each | Monthly credits |
|---|---|---|---|
| HR Self Service, retrieval | 35,000 | 1 | 35,000 |
| HR Self Service, guided draft | 6,000 | 2 | 12,000 |
| Finance Agent, autonomous | 5,000 | 5 | 25,000 |
| Total | 46,000 | Blended 1.57 | 72,000 |
The size of the forecast is driven by the share of autonomous actions, because they draw five times a retrieval action. A small shift toward autonomy moves the total sharply.
Hold actions flat and raise the autonomous share, and the burn climbs fast. This is why the mix matters more than the count.
As users trust the agents, they hand over more complete tasks. Model that drift, or the forecast ages badly.
The common advice is to forecast AI cost from the number of agents or the number of users. We disagree. In the forecasts we built, agent count barely predicted burn, while the skill mix predicted it closely, and the autonomous share moved the total more than any other input. The buyer side move is to forecast from actions per skill, run a heavy autonomy case as a sensitivity, and size the paid tier so it absorbs adoption rather than the launch month alone. A forecast built on agent count is a forecast that breaks the first time usage matures.
Source: Redress Compliance advisory engagement file, 2024 to 2025.
Agent count tells you almost nothing about cost. The mix of skills, and how fast it tilts toward autonomy, tells you almost everything.
You use the forecast to size the paid tier against observed burn and to negotiate the rate card from evidence. The number that matters is the gap between projected burn and the complimentary allotment.
Compare annual projected burn against the allotment and fund the gap, using the complimentary allotment described on the Workday AI Flex Credits page. Do not extrapolate from a quiet first month.
Run a second forecast with autonomy dialed up and take both to the table. A tier that only fits the light case is a tier that fails on adoption.
You forecast Illuminate agent costs by mapping each agent to the skills it runs, multiplying expected monthly actions by the per skill credit value, and totaling the credit draw against the complimentary allotment. The complimentary window supplies the usage data.
Illuminate agents vary in cost because they run different skills, and each skill has its own credit value. An agent that mostly retrieves information draws near 1 credit per action, while one that completes tasks autonomously draws near 5.
You need the list of active agents, the skills each one runs, and the monthly action volume per skill from real usage. The Agent System of Record supplies the agent and skill inventory, and the complimentary window supplies the volume.
The forecast should run at least two quarters ahead of renewal so the complimentary window can supply real usage data. A late forecast leaves you sizing the paid tier on a vendor estimate rather than your own numbers.
