How Power Pals credits work

What Credits Measure

Credits measure the amount and type of AI work performed for you. A short reply uses fewer credits. A long answer, deep analysis, large document, or workflow that uses files, search, images, or code uses more.

Credits are based on four main factors:

1. How much new information the AI reads.
2. How much previous context is reused.
3. How much new content the AI writes or generates.
4. Whether specialist tools are used, such as web search, file search, image generation, or code analysis.

How Chat Credits Work

Normal chat uses credits based on the amount of context read and the amount of answer produced.

Chat activity	Credit impact
Reusing existing context already seen by the AI	Lowest
Reading new instructions, files, history, or context	Medium
Writing new output	Highest

This means a short answer to a focused question usually uses only a few credits. A long response that reads a lot of previous conversation and writes a detailed answer uses more.

Why Long Chats Can Use More Credits

As a chat grows, Power Pals may need to include more previous context so the active Pal understands what has already happened. That can make later replies use more credits than earlier replies.

To manage this, you can start a fresh chat when the topic has changed, keep prompts focused when you do not need the full previous discussion, or use document and file tools when you want a specific artifact.

How Different Activities Consume Credits

Short Chat Replies

Short, focused replies generally use the fewest credits. Examples include quick explanations, simple rewrites, short planning help, or direct answers.

Standard Help and Analysis

Standard responses use more credits when the Pal reads more context or provides a fuller answer. This includes ordinary business support, drafting, research planning, comparison, and structured advice.

Deep Analysis

Deep analysis uses more credits because the Pal spends more effort reasoning through the request and often produces a longer, more structured response. Examples include technical reviews, complex strategy, detailed document analysis, or multi-step problem solving.

Memory Updates

Power Pals uses background memory processes to keep useful context organized, such as chat summaries, topic organization, and workstream intelligence. These can consume credits because they ask AI to read and organize your working context.

RoundTable and Multi-Pal Work

RoundTable and other multi-Pal workflows can use more credits than a normal chat turn because several Pals may contribute, evaluate, challenge, or synthesize the answer. Each meaningful AI contribution can consume credits.

DocuThink and Document Generation

DocuThink can consume more credits because it may involve planning the document, creating the structure, writing sections, reviewing the flow, and adding visual or media planning where relevant.

Longer documents, more sections, richer source material, and generated visuals all increase credit use.

Personalized Suggestions

Personalized prompt suggestions on your home screen are generated by AI from your recent work, so they consume credits each time they are produced. They are turned off by default and only run when you switch them on - while off, no suggestion credits are used. When on, a small amount of credit is used to tailor the suggestions to your recent chats, topics, and Pals.

Specialist Tool Credits

Some specialist tools add a fixed credit amount when used, in addition to any credits used by the Pal to read context and produce the response.

Tool action	Typical fixed credit use
File search call	4 credits
Web search call	15 credits
Code analysis session	50 credits

Privacy Scanning and Redaction

Privacy scanning can consume extra credits because Power Pals performs additional processing before supported content is used by a Pal.

For supported file uploads, File Scanning can inspect a document before it is used by a Pal. Users can normally choose this in the file-scanning modal before upload. Account admins can also enable Force File Scanning so supported uploads are scanned even when users skip the preview scanner, which is useful for structured documents and forms used in automations.

Force Full Scanning is available on Business packages and above. It requires the deeper scan path whenever forced scanning runs, which is best for chat, prose, and less-structured business content. It adds Azure Language scanning at higher credit cost and is only available when File Scanning is enabled. When Redact PII is enabled, Power Pals creates a privacy-redacted version of the extracted text before it is passed into the normal workflow.

Privacy credit use is based mainly on the size of the document and the level of scanning selected.

Privacy action	Typical credit use
File scan	Usually about 2 credits per page
Full scan	May use additional credits depending on document content

Example privacy action	Typical estimated privacy credit use
File scan on a 5-page document	About 10 credits
Full scan on a 5-page document	Usually 25 credits or more
Redaction on a 10-page document	About 20 credits
Full scan with redaction on a 10-page document	Usually 50 credits or more

The in-app privacy preview lets you review scan results and redacted text before deciding whether to use the file in a normal workflow. Running a preview scan still consumes credits immediately. After the scan, the app shows the actual credits used.

Privacy scanning is additive: it can be charged in addition to any later chat, file search, or Pal response credits generated when the processed content is actually used.

Image Generation

Image generation uses credits when Power Pals creates images or visual assets. The exact amount can vary based on the image request, the generation settings, and any related planning or document work.

What Usually Does Not Consume User Credits

Some operations are treated as part of running Power Pals rather than direct user consumption. Examples may include setup tasks, account administration, internal routing, system maintenance, and other background operations that are not exposed as user-triggered AI work.

Opening the app, browsing existing messages, viewing settings, or looking at previously generated content does not normally consume credits.

How to Reduce Credit Use

• Start a new chat when the old context is no longer needed.
• Ask for a short answer when you only need a quick result.
• Use clear instructions so the Pal does not need to infer too much.
• Avoid asking for unnecessary web or file searches.
• Use RoundTable or deep analysis when the extra depth is worth it.
• Break very large document tasks into planned stages when you want more control.

Allowances, Top-Ups, and Remaining Credits

Your plan may include a regular credit allowance. Some accounts can also add extra credits through top-ups, depending on how the account is managed.

In the app, your credit view may show credits used in the current period, credits remaining, any available top-up balance, and whether the account is in a consumption or managed billing mode.

If your account is managed by a partner or through Microsoft Marketplace, some billing and credit actions may be handled outside the app by your billing owner.

Why Credits Are Useful

Credits make AI usage easier to understand across very different tasks. A single short answer, a multi-Pal discussion, a document generation workflow, a file search, and a code analysis session are not the same kind of work. Credits provide one shared way to track and manage them.

The goal is transparency: you should be able to see what has been used, understand why heavier workflows consume more, and choose the right level of AI effort for each task.