By Marc Johnson, CTO & Co-founder, TeamCentral | April 2026
On September 1, 2025, Microsoft replaced the “message” as the standard billing unit for Copilot Studio with a new model: Copilot Credits. The change was technical in nature, but its revenue implications for Microsoft partners are anything but minor. Partners who are still scoping deployments without modeling credit consumption are designing for a revenue opportunity they cannot actually see.
This article explains how the credit model works, why the type of workflow you build matters more than the number of users you deploy to, and what architectural decisions determine whether Partner Admin Link (PAL) attribution generates meaningful Azure Consumed Revenue (ACR) for your practice.
What Copilot Credits Actually Measure
Copilot Credits are not a flat per-message fee. They’re a usage-weighted billing unit that varies by the type of work an agent performs.
The key credit tiers are:
- 1 credit per basic non-AI query
- 1–2 credits for classic or generative responses from static document sources
- 5 credits per tool or workflow trigger (e.g., creating a ticket, sending a confirmation)
- 10 credits per Tenant Graph Grounding query (live retrieval from connected enterprise systems)
- When an agent uses a reasoning-capable model, Copilot Studio bills two meters: the standard feature rate for the operation, plus Text and generative AI tools (premium) at 100 Copilot Credits per 10 responses.
At $0.01 per credit, the difference between an agent that answers HR policy questions from a SharePoint document (1–2 credits per interaction) and an agent that checks live inventory in an ERP system (10 credits per interaction) is a 5–10x difference in Azure consumption per query. Multiplied across daily operational usage, that gap defines whether a deployment is a meaningful ACR contributor or a nominal one.
Why Workflow Architecture Is a Revenue Decision
Two customers with the same Copilot license counts can generate dramatically different Azure consumption depending on what their agents are doing. The determining factor is how frequently the agent accesses live, external operational data.
Agents designed around high-frequency operational workflows (order status checks, procurement queries, customer service lookups spanning CRM and WMS data) drive consistent daily credit consumption. These workflows meet real business needs multiple times per day per user, and each interaction triggers the higher-cost credit events (Tenant Graph Grounding, tool triggers) that compound into substantial ACR.
Agents built for episodic use cases (like policy lookups, onboarding guides, infrequent document retrieval) generate far less consumption, even when they serve the same number of users. The usage pattern is the consumption driver.
The practical implication here is that, before scoping a Copilot Studio deployment, partners should map the target workflows to their likely credit profile. An order management agent handling 1,000 queries per day at an average of 15 credits per query generates 450,000 credits monthly ($4,500 in ACR). A document-retrieval agent handling the same query volume at 2 credits each generates $600. They’re the same size deployments with differing workflow designs.
PAL Attribution: The Setup That Makes Architecture Revenue-Eligible
Partner Admin Link (PAL) is the mechanism that connects Azure consumption to a partner’s PartnerID. Without correct PAL configuration, or its equivalent Digital Partner of Record (DPOR), consumption generated by deployed agents does not flow back to the partner as ACR, regardless of how well the agents are designed.
Two configuration requirements matter most:
1. The Power Platform environment must be linked to an active Azure subscription. If an environment runs on prepaid message packs without an associated Azure subscription in a pay-as-you-go billing configuration, consumption does not register on Azure meters. PAL attribution requires PAYG metering to be active.
2. PAL should be configured before production usage begins so Azure-consumed revenue can be attributed correctly. Microsoft documents that a linked partner ID can be changed, added, or removed, but its published guidance does not clearly state that attribution is guaranteed retroactively for usage that occurred before the partner link was in place.
When both conditions are met, every credit-generating agent interaction (whether a Tenant Graph Grounding call, a workflow trigger, or a premium reasoning query) generates ACR attributed to the partner. For customers with a Microsoft Azure Consumption Commitment (MACC), this consumption also reduces their committed spend balance, which creates additional alignment between partner revenue and customer purchasing priorities.
Modeling Consumption Before You Build
Microsoft provides the Copilot Studio agent usage estimator as a planning tool, but the underlying methodology should be considered as well.
Effective pre-deployment modeling covers the following:
- Query frequency per workflow type: how many times per user per day does each agent action occur?
- Credit tier per interaction: does this workflow trigger Tenant Graph Grounding, tool actions, or static retrieval?
- Number of active users and expected adoption curve: what is the realistic daily active user count in month 1 vs. month 6?
- Environment configuration: is PAYG enabled, and is the subscription correctly linked to the partner’s PartnerID?
This modeling converts an architecture decision into a revenue forecast. It also surfaces whether a proposed deployment justifies the integration investment required to connect to external systems, and which workflow to prioritize for a proof-of-concept.
The Partner Takeaway
The September 2025 credit model change makes it clear that Copilot Studio deployments are not all equivalent from a partner revenue standpoint. The consumption profile of a deployment is set by the workflows it serves, the systems it connects to, and the PAL configuration that attributes that consumption correctly.
Partners who model this before go-live are designing for a recurring revenue stream. Those who scope based on user counts and license tiers are likely underestimating both the value of the engagement and the ACR opportunity it represents.
The architecture questions (i.e.; which workflows, which data sources, which credit tiers) are the revenue questions. They belong in the scoping conversation, and not in the post-deployment review.
About the Author: Marc Johnson is the CEO/CTO, and Co-founder of TeamCentral. He works with Microsoft partners to design Copilot Studio deployments that generate measurable Azure consumption and sustained business value.
