Workday sells its agents on Flex Credits: an annual pool of credits that is drawn down each time an agent completes a defined task. That makes the old way of budgeting, multiplying a per-employee fee by headcount, useless. The new question is how many completed actions of each kind the business will ask agents to perform, and how many credits each one costs.
Workday has not published a dollar price per credit. You can still build a defensible estimate in credits, which is the unit the contract and the consumption console use. This guide shows how.
Step 1: List the agent skills you actually plan to switch on
Start narrow. For each agent you are considering, write down the individual skills, not just the agent name. The Self-Service Agent is a good example of why: Workday's page prices instant information retrieval at 1 credit per action and autonomous task completion at 5 credits per action. The same agent can be cheap or five times as expensive depending on which skill does the work.
Get the current per-skill rates from Workday's Flex Credits Rate Card through the customer's account team. Rates are set per capability, so a budget built from the wrong card is wrong everywhere.
Step 2: Count today's volume for each skill
An agent mostly replaces or speeds up work that already happens. That work leaves a trail in Workday you can count:
- Questions people ask today. Help-desk tickets, HR case volumes and "how do I" requests are the base for retrieval skills.
- Transactions people submit today. Count completed business processes of the relevant type over the last twelve months: time-off requests, expense reports, job changes, address changes. These are the base for task-completion skills.
- Seasonality. Open enrolment, year-end and review cycles create peaks. A monthly average hides the month that drains the balance.
A report on completed business process events by type and month is usually enough. You are looking for an order of magnitude, not a precise forecast.
Step 3: Apply an adoption rate, not 100%
Agents rarely take over all of a process in year one. Model three cases, for example 15%, 35% and 60% of eligible volume moving to the agent, and keep each case separate in the spreadsheet. The low case is what you commit to; the high case is what you set alerts for.
The arithmetic for each skill is simple:
annual credits = eligible annual volume × adoption rate × credits per action
Add the skills together for each case.
Step 4: Pilot in non-production to find the real actions per task
The estimate above assumes one action per task. Real agents do not always work that way. A single "change my address" request might involve a lookup, a validation and a submission, and each may be metered differently.
Workday says customers can test most agents for free in non-production. Use that. Run a set of realistic requests in a sandbox, then ask the account team how those runs would have been metered in production. Replace your assumed actions-per-task with the observed number before the budget goes to finance.
Step 5: Set alerts before the first production action
Workday's consumption console supports planning, monitoring and custom alerts and budget tools. Before anything goes live:
- name the person who owns the balance;
- set alerts at points that give you time to react, for example at half and three-quarters of the annual pool;
- agree who decides whether to add credits, move credits between agents or switch a skill off.
Credits can be added at any time through Workday Marketplace, and capabilities can be switched on or off without a new procurement cycle. Decide the rule now so nobody improvises it in month nine.
Questions to ask the account team
Workday's public page leaves several commercial points to the contract. Get written answers to:
- What is the price per credit, and does it change with volume?
- How many complimentary credits does this customer receive each year, and when do they reset?
- What happens to unused credits at the end of the term?
- What happens if the balance runs out mid-year: does service continue, and how are overages billed?
- Which of the skills we plan to use are excluded from free non-production testing?
- Are per-skill rates fixed for the contract term?
- How are custom agents we build ourselves metered?
Where kiweely fits
The cheapest place to learn how an agent behaves is a tenant where testing is free. kiweely connects to a client's non-production tenants with consent, builds and validates Workday Extend apps there, and tests them in the browser before anything reaches Production. Its own AI runs on the ChatGPT or Claude account you connect, so the build work itself does not draw on the client's credits.
Once you have a budget, the next step is keeping consumption inside it: see keeping custom agents from burning credits.