Agentforce vs Copilot Studio pricing: what Salesforce and Microsoft Copilot actually charge per agent action.

The one conversion that makes these two prices comparable

Both vendors publish a rate card, both denominate it in credits, and neither tells you what a credit is worth in the other company's money. That is the whole reason this comparison is hard, and it takes two divisions to fix. Salesforce sells Flex Credits at $500 per 100,000, so a Flex Credit is half a cent. Microsoft sells Copilot Credits in packs of 25,000 at $200, so a Copilot Credit is eight tenths of a cent. Those are not the same unit and they are not close.

Now the rate cards can be read side by side. A standard Agentforce action costs 20 Flex Credits, which is $0.10. A Copilot Studio agent action costs 5 Copilot Credits, which is $0.04. Salesforce charges two and a half times as much for the nearest equivalent unit of agent work. Run it on voice and the gap changes shape rather than closing: an Agentforce voice action is 30 Flex Credits, $0.15, while Microsoft's premium generative voice runs 75 Copilot Credits a minute, $0.60, so Microsoft looks more expensive on voice until you notice one is priced per action and the other per minute. That single mismatch is the reason we do not publish a headline "which is cheaper" number for voice, and why anyone who does is guessing at your average call length.

What each vendor means by an action, which is not the same thing

The word "action" is doing different work in the two rate cards, and this is where naive comparisons fall apart. Salesforce defines it as a specific function an agent executes on the platform: updating a record, summarizing a case, answering a product inquiry, or running a custom prompt or flow. Each of those draws 20 Flex Credits from the pool. Salesforce also notes that Flex Credits are fungible across actions, prompts, translations and voice actions, so one balance covers everything.

Microsoft splits the same territory into more meters. An agent action at 5 credits covers triggers, deep reasoning, topic transitions and computer-using agents. Answers are billed separately, 1 credit for a classic authored answer and 2 for a generative one. Tenant graph grounding adds 10 credits on top. Microsoft's own example is a single complex prompt costing 12 credits, 10 for grounding and 2 for the answer, which is about ten cents. The practical consequence is that a conversational Copilot Studio agent bills several small amounts per interaction while an Agentforce agent bills fewer, larger ones, and your own mix decides which pattern is cheaper. If you want the full per-line breakdown for either platform on its own, we have done that separately for Agentforce and for Copilot Studio.

The seat math, and why $30 and $550 are not the same offer

Both vendors sell a way out of the meter for employee-facing work, and the sticker prices look wildly far apart. Microsoft 365 Copilot is $30 per user per month paid yearly, and it zero-rates classic answers, generative answers and tenant graph grounding when the user is licensed and the agent runs under that user's authenticated identity. Salesforce charges $125 per user per month for an Agentforce add-on covering Sales, Service or Field Service, and its Agentforce 1 Editions start from $550 per user per month with the add-on included and 2.5 million Flex Credits per org per year.

Comparing $30 against $550 directly is unfair to Salesforce and useless to you, because the two seats contain different products. The Microsoft seat is an AI license laid over a Microsoft 365 subscription you already pay for. Agentforce 1 is a full Salesforce edition, so it also carries the CRM functionality that would otherwise cost $25 to $350 per user per month on Sales Cloud. The honest comparison is the increment: what does adding agents to the platform you already run actually cost per head. On that basis it is $30 for Microsoft, $125 for the Salesforce add-on, and $550 and up if you are re-platforming onto an Agentforce 1 Edition anyway.

There is a breakeven worth knowing on the Microsoft side, because $0.008 a credit means $30 buys exactly 3,750 Copilot Credits. An employee generating more agent activity than that in a month is cheaper on a seat than on the meter, and 3,750 credits works out to about 15 tenant-grounded questions a working day. Salesforce publishes no equivalent credit grant on the $125 add-on, describing it as unmetered for licensed employee use, so the same arithmetic cannot be run against it. That asymmetry is itself useful: Microsoft lets you calculate the switch point, Salesforce asks you to trust the word unmetered.

What happens when you run out, where the two behave in opposite directions

This is the difference we would most want to know before signing, and it is the one buried deepest in both sets of documentation. Ask each vendor what happens when you blow through your entitlement and you get opposite answers.

Salesforce is relaxed about it. Its pricing FAQ states there is no overage penalty, that if you exceed your entitlement your rate is your contracted rate billed monthly in arrears, and that you should prevent overages using alerts in Digital Wallet. In other words the agent keeps working and you get a larger invoice. Microsoft is not relaxed about it at all. Copilot Studio enforcement applies once allocated credits are exhausted and disables custom agents outright once a tenant reaches 125 percent of its prepaid capacity. An in-flight conversation is allowed to finish, and every subsequent attempt to invoke the agent is rejected until capacity is increased or the month resets. The end user sees one of two documented strings: "There is a billing issue." or "This agent is currently unavailable. It has reached its usage limit."

Read that again with a public-facing support agent in mind. On Salesforce, an unexpectedly busy month costs you money. On Microsoft, it costs you the agent, and your customers read a billing error where an answer should be. Neither behavior is wrong, and Microsoft's hard stop is arguably the safer default for a company that fears a runaway bill. But they are opposite risk postures, and choosing between them is a real decision rather than a footnote. Both platforms agree on one thing: unused capacity expires. Salesforce says unused Flex Credits do not roll over into subsequent subscription terms, and Microsoft enforces purchased capacity monthly with no carryover.

The cheap license that is not the price of entry, on both sides

The two vendors independently arrived at the same commercial structure, and spotting it is the fastest way to sanity check any agent pricing page. Salesforce sells an Agentforce User License at $5 per user per month to deploy agents company wide, and prints "Requires Flex Credits" directly beneath the price. Microsoft offers a Copilot Studio user license that costs nothing at all, then notes in its documentation that the tenant needs a prepaid Copilot Credit pack subscription before that free license can be assigned. Microsoft's individual trial license goes further still: you can build and test agents with it, and you cannot publish them.

So on both platforms the seat is a permission and the credits are the product. This is not unique to these two. We found the identical pattern reading GoHighLevel AI Employee pricing, where the AI is an add-on charged per sub-account on top of a platform plan, and a gentler version of it in Zapier Agents pricing, where the plan buys a fixed number of activities. When you see a strikingly low per-user number on an agent product, the useful question is not "how cheap is the seat" but "what does the seat require underneath it".

Buying models you cannot mix, which narrows the choice more than it looks

Salesforce is unusually explicit here and it is worth quoting the constraints rather than paraphrasing them. Flex Credits and Conversations will not be supported in the same org. Flex Credits are available with the PayGo and Pre-Commit buying models, while Conversation-based pricing, the Agentforce add-ons and the Agentforce 1 Editions are not available with those models. Switching from Conversations to Flex Credits means swapping all of your existing Agentforce Conversation SKUs through your account executive.

That matters because the cheapest line on the Salesforce price list may not be combinable with the edition you actually need. If you want the unmetered employee experience of an Agentforce 1 Edition, the flexible consumption models are off the table by definition. Microsoft's equivalent constraint is softer but real: packs and the pay-as-you-go meter are interchangeable, and Microsoft states there are no in-product differences between them, but pay as you go requires an active Azure subscription linked through a billing policy in the Power Platform admin center. If you do not already run Azure, the flexible option carries a setup dependency the pricing page does not mention.

Where each one is clearly the right buy

If your company runs on Salesforce, buy Agentforce. Agents that read and write CRM records natively, inside the object model your business already uses, are worth more than a cheaper per-action rate somewhere else. The same logic points the other way with equal force: if your company runs on Microsoft 365 and your people already hold Copilot licenses, build in Copilot Studio, because the zero-rating on employee-facing usage is an offer no outside vendor can match. When the platform question is already settled, the credit arithmetic on this page is a budgeting exercise rather than a decision.

The case for us is narrower and we will state it precisely. You are a small business or a founder-led team. You do not run Salesforce, or you do but you have no appetite for account executives, SKU swaps and rate cards. You want a defined outcome rather than a platform to build on: research done, a lead list built, outreach drafted, an inbox triaged, a report produced. And you want next month's invoice to match this month's. That is what flat-rate agent pricing is for, and it is why our number is one number.

How to work out your own number in about fifteen minutes

Start by settling the platform question, because it dominates everything else and the credit rates barely move it. If both platforms are genuinely open to you, count seats first. List the people who will use agents heavily, meaning daily rather than occasionally, and price them at $30 on Microsoft or $125 on the Salesforce add-on. Everyone else stays on the meter.

Then estimate the metered traffic only, which means customer-facing agents, unlicensed internal users and anything the seat inclusions exclude. Take your monthly interaction volume and multiply by a realistic per-interaction credit figure, using each vendor's own worked examples as a sanity check. Salesforce models a service agent at 60 Flex Credits per case and prices 100 users at three cases a day as $1,800 a month, and our Agentforce pricing calculator runs that side with your own headcount. Microsoft models a website support agent at 8 credits per run across 900 customers a day, which is 7,200 credits daily, exhausts a $200 pack in under three and a half days and lands near $1,728 a month. Those two examples are the closest thing to a like-for-like the vendors publish, and they land in the same range, which tells you the platform decision matters more than the rate card.

Finally, add the things people forget. Add a reasoning multiplier if you plan to use deep reasoning on Copilot Studio, because reasoning models bill the feature rate plus 10 credits per thousand tokens on a second meter, which turns a two credit answer into a thirty-two credit one. Add capacity headroom on Microsoft, because the grace period stops at 125 percent of prepaid capacity and the penalty past it is a disabled agent rather than a larger invoice. Add implementation time on both. Then compare the total against $149 flat, and the answer will be obvious in whichever direction it falls, which is the entire point of doing the arithmetic instead of reading a listicle.