Four Models Hiding Behind One Word
Not long ago, licensing Microsoft AI was straightforward. If someone wanted Copilot, you bought them a licence and moved on. Today, the picture is very different. Microsoft’s AI portfolio now spans four distinct commercial models, each with its own licensing approach, pricing structure and billing behaviour.
This matters because the shape of the cost, not just the size, affects how you budget, govern and control your AI investment. A fixed per-user licence behaves very differently from a consumption-based service where costs rise and fall with usage. For small and mid-sized businesses and non-profits, where budgets are often tight and predictability matters, choosing the wrong licensing model can mean the difference between achieving measurable value and facing unexpected costs.
To help make sense of it all, this guide explores the four Microsoft AI offerings you’re most likely to encounter: Microsoft 365 Copilot, Copilot Studio, Copilot Cowork and Microsoft Scout. We’ll explain how each licensing model works, where it’s best suited, and the key considerations before you commit.
A quick note on pricing: Microsoft AI licensing evolves rapidly, and pricing varies by region, licensing agreement and over time. Rather than quoting figures that may change, this guide focuses on the commercial models themselves so the advice remains relevant long after publication. Always confirm current pricing for your organisation and region before making purchasing decisions.
Model 1: Per-user Seats (Microsoft 365 Copilot)
How it works
The most familiar licensing model is the per-user licence. Microsoft 365 Copilot is purchased on a per-seat basis: you assign a licence to a named user, pay a predictable monthly or annual fee, and that person gains access to Copilot across Word, Excel, Outlook, PowerPoint and Teams. Their experience is grounded in your organisation’s data through Microsoft Graph, allowing Copilot to work with the files, emails and conversations they already have permission to access. The licence also includes Microsoft’s Researcher and Analyst reasoning agents, along with expanded capabilities for building agents in Copilot Studio.
Why organisations choose it
The biggest advantage of the per-user model is predictability. You know exactly what each licensed user costs, making budgeting straightforward and eliminating unexpected usage charges. The trade-off is that every assigned licence incurs the same cost, regardless of how frequently it’s used. To maximise return on investment, licences should be allocated deliberately to the people who will benefit most and reviewed regularly to ensure they’re being used effectively.
Getting the best value
For smaller organisations, a phased rollout is often the most practical approach. Start with employees who spend much of their day working in documents, email and meetings, where even small productivity gains can quickly add up. A grants officer preparing funding applications or a manager with a constantly full inbox is likely to see far greater value than someone who only uses Microsoft 365 occasionally.
After the first month or two, review adoption, reclaim any underutilised licences and use demonstrated value to guide future licence assignments.
Don’t overlook Microsoft 365 Copilot Chat
It is also worth knowing what you may already have. Microsoft 365 Copilot Chat (a secure, web-grounded chat experience with enterprise data protection) is included with most eligible Microsoft 365 business subscriptions at no additional per-user cost. For many employees, this provides an excellent baseline AI experience.
The difference between Copilot Chat and Microsoft 365 Copilot is where organisations can either save—or overspend. Copilot Chat provides a secure assistant for general questions, drafting and summarising. Microsoft 365 Copilot builds on those capabilities by grounding responses in your organisation’s files, emails and chats through Microsoft Graph, while also unlocking reasoning agents and enhanced Copilot Studio capabilities.
Rule of thumb: Give everyone the included Microsoft 365 Copilot Chat experience first. Then assign Microsoft 365 Copilot licences only to employees who need AI to work directly with organisational data inside the Microsoft 365 apps they use every day.
Model 2: Consumption Credits (Copilot Studio)
How it works
Once you move beyond using Copilot and start building your own AI agents, the commercial model changes. Copilot Studio, Microsoft’s low-code platform for creating custom agents, is licensed primarily through consumption rather than per-user seats.
Instead of paying for individual users, organisations purchase Copilot Credits, which are consumed as agents perform work.
How you pay
Microsoft introduced the Copilot Credits model in late 2025, replacing the earlier per-message pricing approach. Organisations can obtain credits in several ways, including:
- Pay-as-you-go usage billed through Azure.
- Prepaid credit packs purchased as a subscription.
- Pre-purchased capacity plans for more predictable workloads.
Organisations with Microsoft 365 Copilot licences also receive a level of included Copilot Studio capacity, which may be sufficient for lighter internal use.
Why organisations choose it
Consumption pricing is designed to align cost with value. A rarely used agent costs very little to run, while a heavily used agent consumes more credits in proportion to the work it performs. This makes Copilot Studio highly scalable, particularly for organisations whose AI workloads fluctuate over time.
The trade-off is that consumption needs to be forecast and monitored. Microsoft provides a usage estimator to help predict credit consumption based on factors such as agent type, traffic and complexity, while built-in capacity controls help prevent unexpected spending. For most organisations, the safest approach is to begin with a small pilot and expand only once the cost per outcome is well understood.
A practical example
Imagine a non-profit building an internal agent to answer volunteers’ common questions—where to find policies, how to submit expense claims, or who to contact about rostering.
While only a handful of volunteers are testing the agent, credit consumption remains minimal. During a major fundraising campaign, however, hundreds of volunteers may begin using it at once, significantly increasing credit usage simply because the agent is handling many more conversations.
The key is to plan for both scenarios. Forecast expected consumption using Microsoft’s estimator, set a capacity limit that aligns with your budget, and decide in advance how the organisation will respond if that limit is reached.
Getting the best value
For many organisations, the included Copilot Studio capacity that comes with Microsoft 365 Copilot licences may be enough to support smaller internal agents without purchasing additional credits.
As usage grows, regularly reviewing consumption ensures costs remain aligned with business value rather than quietly exceeding the included allowance.
Rule of thumb: Start with the included capacity, build a small pilot, monitor credit consumption closely, and only purchase additional capacity once you’ve demonstrated clear business value.
Model 3: Usage-based Billing (Copilot Cowork)
How it works
Copilot Cowork is Microsoft’s agentic AI experience, allowing users to delegate multi-step tasks for AI to complete more autonomously. It is available to Microsoft 365 Copilot customers, enabled by administrators through the Microsoft 365 admin centre, and accessed by users from the Copilot app.
Rather than charging a flat per-user licence, Copilot Cowork uses a usage-based pricing model, where organisations pay for the work their AI agents actually perform.
Why organisations choose it
Usage-based pricing is a natural fit for agentic AI because the work it performs is often valuable but inconsistent. Unlike day-to-day AI assistance, delegated tasks tend to happen in bursts, with periods of heavy activity followed by little or no use.
This model allows organisations to scale spending with genuine demand instead of paying for licences that may sit idle. For lean teams, it means paying for outcomes rather than reserved capacity.
A practical example
Imagine a team preparing for a board meeting. During that week, they rely heavily on Copilot Cowork to compile competitor research, draft reports and complete other multi-step tasks. Once the meeting is over, usage drops significantly.
Under a traditional per-user licensing model, the cost would remain the same regardless of activity. With usage-based billing, costs increase during busy periods and reduce when demand falls, allowing organisations to pay for additional AI capacity only when they actually need it.
Governing usage
The same elasticity, however, makes cost governance essential. Usage-based models reward clear use cases and active oversight. Sensible organisations define where agents are allowed to operate, monitor consumption against the value delivered, and set the guardrails that keep spending aligned with budget. Treated this way, Cowork’s model is a strength; left unmanaged, it is a risk.
This governance doesn’t need to be complex, but it should be in place before usage begins to scale. Deciding which tasks are genuinely worth delegating, and reviewing consumption during the early stages of adoption, can prevent unnecessary costs later.
Even a simple internal guideline can make a significant difference, such as:
Use Copilot Cowork for multi-step research, analysis and drafting, not quick questions that Copilot Chat already answers at no additional cost.
Model 4: Preview and Program Access (Microsoft Scout)
What it is
Microsoft Scout is Microsoft’s first “Autopilot”, an always-on AI agent designed to work autonomously in the background. At the time of writing, Scout is not yet generally available. Instead, it is being offered through Microsoft’s Frontier early adopter program, where selected customers can trial pre-release AI capabilities and provide feedback before broader release.
How access works
Unlike Microsoft’s other AI offerings, Scout is not something organisations can simply purchase.
Access currently requires participation in the Frontier program, a GitHub Copilot subscription, administrator configuration (including device policies through Microsoft Intune), and acknowledgement of how the agent operates. At the time of writing, Microsoft has not announced general availability pricing or standard commercial licensing.
What this means for organinsations
For most SMEs and non-profits, Scout is something to be aware of rather than something to budget for today.
It provides a clear indication of where Microsoft’s AI strategy is heading, with autonomous, always-on agents becoming an increasingly important part of the Microsoft ecosystem. Organisations that invest now in strong identity management, security and data governance will be better positioned to adopt these capabilities when they become generally available.
Planning ahead
It’s also important to understand what a preview program means from a budgeting perspective. Preview programs exist to evaluate emerging technology, not to support business-critical workloads.
Any investment in Scout should be treated as research and experimentation. Avoid building operational processes around preview features until Microsoft announces general availability, commercial licensing and long-term support.
Rule of thumb: Treat Microsoft Scout as a technology to watch, not one to build your AI strategy around just yet.
The Four Models at a Glance
Microsoft 365 Copilot
AI embedded across Microsoft 365 applications. It works within Word, Excel, Outlook, Teams, and PowerPoint, helping employees work faster using organisational data they already have access to.
Best for: Role-based productivity and organisation-wide AI adoption.
Commercial model: Per-user seat.
How you pay: Fixed monthly fee per licensed user.
Copilot Chat (included)
A secure AI assistant included with many Microsoft 365 plans. It provides web-grounded AI for drafting, research, brainstorming, and everyday questions without requiring additional user licensing.
Best for: Everyday AI assistance and getting started with AI.
Commercial model: Bundled.
How you pay: No extra per-user cost on eligible plans.
Copilot Studio
A platform for building custom AI agents and automations. Organisations can create agents tailored to specific business processes, workflows, and customer interactions.
Best for: Variable, multi-step delegated tasks
Commercial model: Usage-based.
How you pay: Pay for the work agents perform
Microsoft Scout
An emerging agent-focused offering designed for always-on AI experiences. It represents Microsoft’s vision for proactive, autonomous agents operating across business environments.
Best for: Early adopters exploring next-generation AI agents.
Commercial model: Preview / program.
How you pay: Frontier program + GitHub Copilot Subscription
Budgeting and Governing a Mixed Estate
Most organisations will end up with a blend of these models, and the art is managing them together. A few principles help keep a mixed AI estate under control:
- Match the model to the demand. Use predictable seats where usage is steady and broad; use consumption and usage-based models where demand is variable or experimental. Paying a fixed seat fee for occasional use, or running unbounded consumption for a constant workload, both waste money.
- Set controls before you scale. Consumption and usage-based models are powerful precisely because they grow with use, which means they need budgets, capacity limits and monitoring from day one, not after the first large invoice.
- Govern access and data. Every model assumes a sound foundation of identity, permissions and information governance. This becomes critical as agents act more autonomously across your systems, because an agent inherits whatever access you give it.
- Name an owner for each agent and seat group. Someone should be accountable for whether a paid seat is earning its keep and whether a consumption agent is staying inside budget. Costs drift when no one is watching a particular line.
- Review regularly. Because pricing and products in this space change quickly, an annual “set and forget” approach does not work. Periodic reviews keep your licensing aligned with both your needs and Microsoft’s evolving line-up.
To make the blend concrete, picture a not-for-profit organisation with twenty-five staff. Everyone gets the included Copilot Chat experience at no extra cost; eight document-heavy roles get paid Microsoft 365 Copilot seats; the finance lead uses Copilot Cowork in the weeks it is needed; and one Copilot Studio agent for volunteers’ questions runs inside the included Studio capacity. Scout stays on the watch-list.
The result is a layered estate where most spend is predictable, the variable parts are bounded, and nothing is paid for that the organisation does not use.
The Bottom Line
Microsoft AI licensing has expanded well beyond a single seat fee into a portfolio of commercial models: per-user, consumption, usage-based and preview. None is inherently better than the others; each suits different kinds of work and different appetites for predictability versus flexibility.
The organisations that get the most value are those that understand the shape of each cost, match it to real demand, and put governance in place before spending scales.
At 365 Architechs, we help SMEs and non-profits navigate exactly these decisions: choosing the right mix of Microsoft AI licences, forecasting consumption, and putting the cost controls and governance in place to adopt with confidence.
Unsure how to license Microsoft AI without overspending? Contact 365 Architechs for clear, independent guidance.
