Workday Flex Credits are how Workday now sells its AI. Instead of adding a per-employee fee for each new AI feature, a customer buys a pool of credits for the year and draws it down as agents and platform capabilities do work. If you advise Workday customers, or run a tenant yourself, you will be asked what a credit actually buys. This guide sets out what Workday has published, and what it has not.
What Flex Credits are
Workday describes Flex Credits as credits "purchased in bulk and applied across any eligible agent or platform innovation." The subscription is an annual purchase meant to cover the year's needs, and more credits can be added at any time through Workday Marketplace.
The idea is that one pool pays for many capabilities. A customer can switch a capability on or off, or move credits from one agent to another, without starting a new procurement cycle.
What Flex Credits cover
Workday's Flex Credits page lists five areas:
- AI Agents — the agents Workday builds, such as the Self-Service Agent.
- Data Cloud — Workday's data platform.
- Sana from Workday — the AI assistant and search experience.
- Agent-Ready Tools — Workday's pre-built tools that let agents read Workday data and take actions.
- Developer Agent — Workday's AI that builds custom agents, apps and orchestrations. Its product page states that pricing is based on Flex Credits and that it is available through Workday Extend Professional.
The exact list of eligible capabilities, and the credits each one uses, lives on Workday's Flex Credits Rate Card. Treat the Rate Card, not a blog post or slide, as the source of truth, and check its date: rates can change as skills move from early access to general availability.
When an agent is metered
This is the part most people get wrong. Workday says it meters "when a defined task with known value is complete," and that metering happens almost entirely in production.
In practice that means:
- You are not charged per token. A long conversation that never completes a defined task is not priced by how much text the model read or wrote.
- You are not charged per employee. Consumption depends on usage, not headcount.
- Each skill has its own price. "Each metered capability or skill is assigned a specific number of credits," and Workday says those numbers reflect estimated time saved and efficiency gained.
The published examples
Workday's own page gives one concrete pair of numbers, for the Self-Service Agent:
| Skill | Credits per action |
|---|---|
| Instant information retrieval | 1 |
| Autonomous task completion | 5 |
The gap matters. An agent that answers "How much vacation do I have?" uses one credit. An agent that actually submits the time-off request on the worker's behalf uses five. When you estimate cost, the question is not "how many people will talk to the agent" but "how many completed actions of each kind will it perform."
Complimentary credits and free testing
Two things lower the cost of getting started:
- Complimentary Flex Credits. Workday offers an annual allotment "based on your company size." Workday's public page does not list the amounts; your account team or contract will.
- Free non-production testing. Workday says customers can "test most agents for free in non-production." Note the word most: confirm on the Rate Card that the specific agent or skill you plan to trial is included before you assume it is free in a sandbox.
For consultants, this second point shapes how you work. Prove an agent's behaviour in a sandbox or Implementation tenant first, where testing is free, and only move it to Production once you know how many actions it will take per real-world task.
How customers see what they are spending
Workday describes three cost controls in its consumption console:
- Plan — model usage before committing, based on current products and growth plans.
- Monitor — see current credit use and detailed usage and cost over time.
- Control — set custom alerts and budget tools.
Ask early who in the customer organisation owns this console. AI consumption is new territory for many Workday teams, and the person who switches on an agent is often not the person watching the balance.
What Workday has not published
As of this guide's date, Workday's public pages do not state:
- a dollar price per credit (pricing is through your account team);
- the size of complimentary allotments by company size;
- rollover or expiry rules for unused credits;
- per-action rates for custom agents you build yourself.
Partners and analysts have published figures for some of these. Treat them as useful context, not contract terms, and confirm against your own Order Form and the current Rate Card.
Where kiweely fits
kiweely is a desktop co-pilot for Workday consultants. It builds, validates and deploys Workday Extend apps, and reads and changes a client's tenants only with your consent. Two points are relevant to Flex Credits:
- kiweely's AI runs on the ChatGPT or Claude account you connect. The model that plans and writes your Extend app is billed by that provider, not drawn from the client's Flex Credits pool.
- kiweely works in non-production by default. Apps it publishes go to the Development environment, and its testing happens in the tenant you connect, so the experimentation that takes the most attempts happens where Workday says testing is free.
If you are planning an agent rollout, continue with the guide on budgeting Flex Credits before you sign.