Skip to content

AI Builder → Copilot Credits Estimator

Map what you run on AI Builder today to the Copilot Credits you'll need. Runs in your browser — nothing is uploaded.

November 1, 2026: AI Builder credits reach end of life

Features keep working — they draw Copilot Credits once AI Builder credits run out. What changes ↓

1 How do you have your numbers?

2 Which capabilities do you use?
Quick start:
3 Enter your monthly numbers
↑ Pick a capability above (or tap a preset) and a row will appear here to enter your monthly numbers.
2 Import your data

Assume a mix:

4 Add planning assumptions

Use this when customer telemetry is incomplete. Keep Current as the observed baseline, then model the expected migration month and a peak/burst month for provisioning.

Known volume growth, new teams, more flows, or increased automation throughput.
Use because capacity is monthly and production workloads rarely follow a perfect average.
Use lower confidence when the activity export lacks environment, user, page, token, or process-owner validation.

What changes on November 1, 2026

Per Microsoft's End of AI Builder credits calendar:

  • AI Builder capacity add-ons stopped selling to new customers on November 1, 2025 and reach end of life on November 1, 2026. Existing add-on credits stay usable until those licenses expire — and Microsoft notes customers must turn off auto-renew in the Microsoft 365 admin center or Partner Center.
  • Seeded AI Builder credits (in Power Apps / Power Automate premium, Dynamics 365, and similar licenses) also reach end of life on November 1, 2026: future premium licenses won't include them, and existing seeded credits are honored through the end of the contract term.
  • There's no automatic conversion. Environments consume AI Builder credits first, then Copilot Credits; if neither is available, AI Builder features are blocked. That's why sizing the Copilot Credits you need — below — matters now.

Expedited conversion assessment workflow

Use this when a customer has Power Platform reports but the reports do not match the actual conversion planning process.

Fast but defensible model

1. Anchor the total Load the AI Builder consumption export first. Sum the actual consumption field over the agreed reporting window by environment and day.
2. Explain the mix Use AI Builder Activity to group estimated consumption by tool/model. This creates the capability split needed for the rate table.
3. Enrich only the gaps Ask process owners for pages, documents, tokens, peak periods, production/test status, and expected growth for the top consumers only.

Recommended order of operations:

  1. Normalize both files to the same date window and timezone. Use actual timestamp columns, not filenames.
  2. Treat the consumption export as the top-down ledger baseline.
  3. Treat the activity export as the bottom-up workload mix and reconciliation view.
  4. Do not row-join the files unless both expose stable environment/user/execution keys.
  5. Ask Delta to enrich the top 10-20 tools/models or the workloads that represent roughly 80% of usage.
  6. Use the estimator's planning assumptions to publish Current, Expected, and Peak Copilot Credit ranges.

Manual enrichment template:

Column Purpose
BusinessProcess Human-friendly process name, such as vendor invoice intake or claims extraction
FlowOrAppOrAgent Owning Power Automate flow, Power App, Copilot Studio agent, or agent flow
EnvironmentId Join back to the consumption export where possible
ToolName Tool/model/prompt name from AI Builder Activity
CapabilityMapping One of the estimator capabilities, reviewed by the process owner
MonthlyRuns Current monthly run volume
AvgPagesImagesPerRun Required for document, OCR, image, and ID processing
AvgInputTokensOrChars Required for prompt and text workloads
AvgOutputTokens Required for prompt workloads
ProductionStatus Production, pilot, test, retired, or unknown
IncludeInForecast Yes or no
ExpectedGrowthPct Known growth for this workload
PeakMultiplier Burst factor for busiest month
Confidence High, medium, or low
Notes Assumptions, data gaps, or owner comments

The goal is not to ask every maker for perfect telemetry. Start with the reports, reconcile the obvious totals, then use this short intake only where the export cannot tell the difference between estimated activity, production usage, page/token volume, and future growth.

Where do I find my current AI Builder usage?

You don't need exact numbers to get a useful estimate — but the closer you are, the tighter the result. The Import my data tab lets you paste real numbers instead of typing each one.

  • By usage volume (most accurate). Count what you actually process each month: invoice/receipt pages, custom document pages, OCR pages, object-detection images, and — for prompts and text analysis — thousands of tokens or characters. Your process owners usually know these volumes.
  • Paste the per-tool activity grid (exact, recommended). In Power Automate, open Automation Center → Monitor → AI Builder activity. That grid lists each model/prompt run with a Tool name and a Consumption value (it also covers Power Apps and Copilot Studio usage). Set the timeframe, then copy the Tool name + Consumption columns and paste them into Import my data → Paste a table. The tool auto-matches each label to a capability; you confirm the mapping and apply.
  • Paste the tenant total (quickest). In the Power Platform admin center, go to Resources → Capacity → Summary, select Download reports in the Add-ons section, choose +New → AI Builder → Submit, then download the Excel report. It shows credits consumed per environment/date in an AIConsumption column — a single aggregate number, with no capability breakdown. Sum that column and drop it into Import my data → One total number, then pick the capability mix that matches your workload. Because a flat total can't reveal the split, that result is a modeled estimate — adjust the mix to see how much it moves.

The rate card this tool uses

Borrow, don't build — this is grounded, but Microsoft owns the truth

Every rate in this estimator is transcribed verbatim from Microsoft's official AI Builder Capability Rate table, retrieved 2026-08-10. Rates and packaging change; for any customer-facing number, go to the source:

Straight from Microsoft's AI Builder Capability Rate table (retrieved 2026-08-10). "Copilot Credit rate" is what the capability costs in Copilot Studio; "AI Builder credit rate" is what the same unit costs today.

Capability Unit Copilot Credit rate AI Builder credit rate
Prompt — basic LLM model 1k tokens 0.1 1.2 ᵃ
Prompt — standard LLM model 1k tokens 1.5 24 ᵃ
Prompt — premium LLM model 1k tokens 10 182 ᵃ
Text recognition (OCR) 1 page 0.1 3
Simple text analysis (sentiment, language, key phrase) 1k chars 0.1 2
Advanced text analysis (classification, entity extraction) 1k chars 1.5 20
Text translation 1k chars 1.5 22
Custom document processing 1 page 8 100
Receipt, invoice, identity document analysis 1 page 8 32
Contract, health insurance card, image description 1 image 8 32
Object detection 1 image 8 8
Business card reader, prediction n/a free free

Dollar conversions (from the doc's footnotes): 1 Copilot Credit = $0.01 pay-as-you-go; the AI Builder $ reference is the yearly prepaid Tier 1 add-on (1,000,000 credits for $500/month = $0.0005/credit).

ᵃ For prompts, the single AI Builder credit rate is Microsoft's estimate at a 90% input / 10% output token mix (exact per-1k: 1/3 basic, 20/60 standard, 140/560 premium). The Copilot Credit rate is a flat per-1,000-tokens number, so the Copilot Credits this tool reports are exact.

How to read the result

  • Copilot Credits / month is the number you provision. It's the currency Copilot Studio meters, pooled across your whole tenant. Size to your busiest month — capacity is enforced monthly and unused credits don't roll over.
  • Compare dollars carefully. The credit count usually drops sharply moving from AI Builder to Copilot Credits, but the pay-as-you-go $/credit is higher ($0.01 vs the $0.0005 prepaid AI Builder reference). This tool shows both so you plan honestly — and remember prepaid Copilot Credit packs lower the effective Copilot rate.
  • Not a bill. These are directional planning numbers. Confirm customer-facing figures against the Licensing Guide and your own pricing.

Once you have a Copilot Credit budget, size the agents that will spend it in the Credit Estimator, and build the return story in the ROI Estimator.

Sources

All rate and licensing claims are grounded in official Microsoft documentation, retrieved 2026-08-10:

Claim Source Retrieved
AI Builder Capability Rate table (Copilot Credit + AI Builder credit rate per capability, units, $ footnotes: $0.01/Copilot Credit PAYG, Tier 1 prepaid $500/mo per 1M AI Builder credits); AI Builder features in Copilot Studio always consume Copilot Credits; add-on renewal-only from Nov 1 2025 and seeded credits removed Nov 2026 AI Builder — Overview of licensing 2026-08-10
End-of-life calendar: add-ons end of sales 11/1/2025 and end of life 11/1/2026 (turn off auto-renew); seeded credits end of life 11/1/2026, honored through contract term; no automatic conversion; AI Builder credits consumed first, then Copilot Credits, else blocked End of AI Builder credits 2026-10-28
How Copilot Credits are bought, pooled per tenant, and enforced monthly with no carryover Licensing and Copilot Credits — Copilot Studio 2026-08-10
PPAC AI Builder consumption report is a downloadable Excel (admin-only, rolling 30 days) with columns Date, UserId, EnvironmentId, EnvironmentName, AIConsumption, IsTrial — an aggregate credit total, no per-capability breakdown AI Builder consumption report 2026-08-11
Power Automate AI Builder activity grid (Automation Center → Monitor) lists per-run Tool name + Consumption, covering Power Automate, Power Apps, and Copilot Studio Monitor AI Builder models and prompts activity 2026-08-11

Unofficial and community-built. This page is not endorsed by or affiliated with Microsoft. It summarizes public documentation to help you plan — always confirm customer-facing numbers against the Microsoft sources linked above.

Was this page helpful?

Found an issue or want to suggest an improvement? Open a GitHub issue  ·  Star on GitHub