What happens to AI Builder credits on 1 November 2026? Microsoft says new and renewed Power Platform and Dynamics 365 licences will stop receiving seeded AI Builder credits. Eligible credits from contracts purchased or renewed before that date remain available through the applicable contract term. Existing AI Builder capacity add-on credits also remain usable until their contracts expire. AI Builder features continue to be supported using Copilot Credits.
For the person funding an automated document or prompt workflow, the immediate decision is to identify when its entitlement ends and what will fund execution afterwards. There is no universal overnight shutdown, and Microsoft does not automatically convert your AI Builder entitlement into Copilot Credits.
The transition follows two contract calendars
Microsoft’s end-of-credits guidance distinguishes seeded credits included with eligible premium licences from separately purchased capacity add-ons. Both move towards Copilot Credits, but the contract attached to each source determines continuity.
| Capacity source | What changes from 1 November 2026 | What to establish before renewal |
|---|---|---|
| Seeded credits from eligible premium licences | Future new or renewed licences no longer supply seeded AI Builder credits; existing eligible entitlement continues through its contract term | Which licence supplies the credits, its term end and which environments depend on them |
| Existing AI Builder capacity add-ons | Existing credits remain usable until the add-on contract expires; Microsoft’s guidance tells customers to turn off auto-renew | The add-on expiry, the purchasing owner and replacement capacity for dependent workloads |
| Copilot Credits | These fund supported AI Builder execution under the relevant consumption rules | The environment’s capacity or billing arrangement and a budget matched to its workload |
Treat the auto-renew instruction as a task for the authorised licensing administrator. Record the replacement funding decision alongside it; turning off renewal alone does not establish continuity.
Two departments can therefore reach different renewal decisions. One workflow can still draw on an active contract while another environment has already moved to Copilot Credits. Tell each workflow owner which entitlement pays for their runs and when it ends; a blanket “credits end in November” email leaves them unable to plan.
Fallback depends on the environment, not the spreadsheet budget
In Power Apps and Power Automate, AI Builder features first consume available AI Builder credits. When those credits are absent or exhausted, execution attempts to use Copilot Credits. If neither currency is available, the features fail. In Copilot Studio agents and agent flows, AI Builder capabilities consume Copilot Credits rather than AI Builder credits.
These rules are documented in Microsoft’s licensing overview and Copilot Credit guidance. They make environment-level capacity an execution dependency. A finance approval to buy credits is insufficient if the workload cannot use the resulting capacity where it runs.
Build a renewal sheet with one row per business workflow:
| Field | Evidence to record |
|---|---|
| Business output | The completed item an operator accepts, such as a verified document record |
| Runtime | Power Apps, Power Automate, or a Copilot Studio agent/agent flow; name the environment |
| Current capacity | Seeded licence or add-on supplying AI Builder credits, or the existing Copilot Credit arrangement |
| Term end | The actual contract date, confirmed by the licensing owner |
| Replacement route | Copilot capacity or billing arrangement, its administrator and the approval still outstanding |
| Operating consequence | Who handles work if the AI step is blocked, and where the backlog is visible |
Give the licensing administrator and workflow owner the same row to sign off. The administrator confirms the term end and replacement capacity; the workflow owner confirms which work stops if that capacity is missing. That joint check catches a funding gap before a renewal changes what the workflow can consume.
Reprice the workload in its actual units
There is no single entitlement-conversion ratio. Microsoft’s guidance states that the currencies have scenario-specific rates. Rebuild the estimate from the capability used and the units it processes, rather than translating an old credit balance into an equal number of Copilot Credits.
For example, the published capability table lists receipt, invoice and identity-document analysis at 8 Copilot Credits per page. Using Microsoft’s stated pay-as-you-go reference of US$0.01 per Copilot Credit, an illustrative workload processing 10,000 pages consumes 80,000 Copilot Credits, equivalent to US$800 for that AI capability alone.
That is a planning illustration using the documentation checked on 29 October 2026, not a customer result or a full invoice quote. Check the capability classification and your purchasing terms before using it in a budget. Prompt workloads have different units and model tiers; an invoice-analysis example is not a rate for every document or AI action.
Keep the surrounding platform costs separate. Microsoft states that AI Builder actions make a Power Automate flow a premium flow. Under pay-as-you-go, cloud-flow run billing is separate from AI Builder credit consumption. The full implementation-cost guide covers the wider budget boundary, including integration and ongoing operation.
The useful comparison is the cost of producing work the business accepts. If a changed extraction step saves credits but creates more manual correction, the operating result can worsen. Use the accepted-outcome ROI model to keep that correction effort inside the decision.
Verify the replacement route before the old entitlement expires
The following is a proposed acceptance check for your authorised administrator and workflow owner. It is not a report of tests run against a customer tenant.
Select a representative, non-sensitive input in an authorised test environment. Record the capability and processing units, confirm that the intended Copilot Credit arrangement is available, then run the workflow and inspect both its output and consumption record. A successful model response alone does not establish that the receiving system stored a usable result.
Where the platform configuration permits a controlled check, verify the replacement route without depending on remaining AI Builder credits. If you cannot isolate that condition safely, record the limitation and confirm the transition plan with the licensing administrator or Microsoft. Do not remove production capacity to manufacture a test.
Agree who receives a blocked-run alert and how pending work will be handled. A prompt feature failing midway through a business workflow requires an owner even when the underlying cause is billing capacity rather than model quality.
One monitoring detail deserves care: Microsoft’s transition FAQ says AI Builder credit-consumption computation can be delayed by five or more days. A workload continuing to run is therefore insufficient evidence that its long-term funding is correct. Inspect the relevant entitlement and consumption records rather than inferring them from a green run history.
Separate the funding fix from the redesign decision
A credit transition does not, by itself, justify rebuilding a working automation as an agent. First establish continuity and calculate the revised operating cost. Then decide whether the workflow still earns its place, whether its expensive AI step is necessary, or whether correcting poor inputs would remove more work than changing platforms.
Model Operator is a Dubai-based AI product studio led by Alexander Chrisostomou. Through AI Initiative Consulting, Alexander works with teams to prioritise consequential workflows and agree the implementation and adoption work they require. This licensing change is a reason to inspect an existing workflow’s economics, not proof that a new company-wide system is the right purchase.
If the revised cost exposes a workflow that consumes budget without reliably completing the work, bring its renewal sheet and a representative output to a Model Operator build conversation. Use that evidence to decide what to keep and what to change.