> ## Content Index
> Fetch the complete content index at: https://www.doitwithfriday.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Cost of AI at Work: What Microsoft's GigaOm Study Shows
- URL: https://www.doitwithfriday.com/news/cost-of-ai-at-work/
- Published: 2026-11-18T13:10:01.000Z
- Updated: 2026-11-20T14:36:29.000Z
- Description: A Microsoft-funded study with GigaOm prices Copilot, ChatGPT Enterprise and Claude Enterprise per user each month. Here is what it found and what it assumed.
- Author: Kevin Anderson
- Tags: News, What It Costs

Microsoft published the figures on November 16, 2026\. They are estimates from Microsoft's own tests, and the vendors' US pricing pages show why the totals move.

For a small business, the gap between a flat seat price and a metered usage bill decides which total is real, and the small-team plans are priced differently.

## Microsoft puts a monthly price on the cost of AI at work

On November 16, Microsoft's Copilot blog published what it calls the cost of AI at work for three products. The post discloses that Microsoft paid for the research.

### Who funded the study and who ran the tests

The post, signed by Nicole Herskowitz, Microsoft's corporate vice president for Microsoft 365 and Copilot, says Microsoft "commissioned and funded research with GigaOm" on a report titled Demystifying the TCO of Enterprise AI. It compared Microsoft Copilot, ChatGPT Enterprise and Claude Enterprise through hands-on testing in October 2026.

The split of work matters. Microsoft says it "ran the testing and cost analysis." GigaOm contributed to the research design, validated the methodology, results and cost analysis, and co-developed the findings. OpenAI and Anthropic, whose products were priced, are not named as participants. In Friday's view, that makes the study a vendor's estimate of the cost of AI at work, reviewed by an outside firm, and it should be read that way.

### The six numbers Microsoft published

For an everyday user, the post puts Copilot at $30 a month, ChatGPT at $81 and Claude at $205, "once usage charges were added to their subscriptions." For an advanced user, it puts Copilot at $96, ChatGPT at $149 and Claude at $299\. Microsoft says that makes Claude more than 6x and ChatGPT 2.5x Copilot's cost for everyday use, and more than 3x and 1.5x for advanced use.

![Slope chart of the modeled monthly cost per user in US dollars from the Microsoft-funded GigaOm study, with one line per product from an everyday column to an advanced column: Copilot from $30 to $96, ChatGPT Enterprise from $81 to $149, Claude Enterprise from $205 to $299. The advanced column is highlighted, and no product is singled out.](https://storage.ghost.io/c/67/23/6723d566-f4a4-4c03-8b45-4f8f0948342b/content/images/2026/11/gigaom-ai-cost-slope.webp)

From everyday to advanced use: +$66 Copilot, +$68 ChatGPT, +$94 Claude. Modeled estimates on Microsoft's workloads, not list prices; Microsoft-funded, directional. Source: Microsoft Copilot blog, Nov 16, 2026, Microsoft-funded GigaOm research, checked Nov 18, 2026.

The headline claim of this Copilot vs ChatGPT and Claude comparison is narrow. Microsoft says Copilot had the lowest estimated monthly cost "on the study's modeled workloads," while quality was broadly comparable across the three products on the tasks tested.

## What the study assumed before it priced anything

Every estimate of the cost of AI at work depends on its inputs: the users, their task volume, the prices applied and how quality was scored. Microsoft's post states each one.

### Two kinds of users, 100 and 140 tasks a month

The study modeled two users. Everyday users rely on AI for finding information, preparing for meetings, analyzing data and creating content. Advanced users do all of that and also hand off multi-step work to agents that run on their own. The evaluation covered 37 unique prompts across Copilot Chat and Cowork, plus Copilot in Word, Excel and PowerPoint. Over a modeled month, that came to 100 tasks for an everyday user and 140 for an advanced user, including 20 agentic workflows. Each product was tested against the same test company and data, with prompts run repeatedly.

### List prices, assumed prices and no discounts

Microsoft says all cost estimates "are based on October 2026 list prices, or on an assumed price where list pricing isn't published, and exclude negotiated discounts." The post does not say which product used an assumed price. It ends with its own caveat: "read the findings as directional." In Friday's view, any published figure for the cost of AI at work is only as good as its stated price inputs, and here one of them is unnamed.

### How the tests scored quality

Microsoft says that before testing, human evaluators wrote a checklist for every prompt: the right sources, people, facts and figures, a complete deliverable, and no actions the user did not ask for. A run passed only if it met every item. Microsoft says results varied by task, so the quality finding holds for the workload as a whole. In Friday's view, the test setting favored Copilot in the Copilot vs ChatGPT match-up: the work lived in Microsoft 365, and Claude and ChatGPT reached that data through connectors and add-ins.

## Seat price vs usage-based pricing: where the gap comes from

Microsoft credits three factors: everyday use inside the Copilot seat, its Work IQ context layer and Auto model routing. In Friday's view, the first is a pricing choice behind the everyday figures.

### Copilot keeps everyday work inside the seat

Microsoft says chat and Microsoft 365 app usage is included in the Copilot user subscription, so the everyday tasks "remained within the user subscription." That is why the everyday Copilot figure equals the $30.00 list price on Microsoft's enterprise pricing page for the Microsoft 365 Copilot add-on, paid yearly. The same page says a separate license for a qualifying Microsoft 365 plan is required, and the post does not say whether that base license sits inside the $30.

Advanced work is metered on Copilot too. Microsoft's business pricing page says Copilot Cowork, Code, Autopilot and the Work IQ APIs are billed based on usage through Copilot Credits. In the study, agentic charges in Copilot Credits take the advanced Copilot figure to $96, still the lowest of the three under usage-based pricing.

### Claude Enterprise pricing adds usage at API rates

Anthropic's pricing page describes Claude Enterprise pricing as "$20 per seat per month plus usage billed at API rates," billed annually with a 20 seat minimum. Microsoft says the assumed competitor subscriptions "may be $10 less than Copilot," which fits a $20 seat. If the model used that seat, roughly $185 of the $205 everyday Claude figure is usage. Claude Enterprise pricing also lets admins set user and organization spend limits. In Friday's view, usage-based pricing without a cap is the line item to watch.

### ChatGPT Enterprise has no public list price

OpenAI's ChatGPT pricing page says Business plans start at 2 users and asks buyers to contact sales about Enterprise. A footnote says Enterprise and Business can purchase credits for more access. In Friday's view, next to Claude Enterprise pricing, which is public, the ChatGPT side is opaque. With no published Enterprise list price, the $81 and $149 ChatGPT figures rest at least partly on the study's own assumptions, which Microsoft's post does not itemize.

## Copilot vs ChatGPT vs Claude for a business under 20 seats

The study priced enterprise plans. A small business may not want their minimums, so the plans built for small teams are the fairer place to compare Copilot vs ChatGPT and Claude.

### What the small-team plans list

Microsoft's [Copilot business pricing page](https://www.microsoft.com/en-us/copilot/pricing/business?ref=doitwithfriday.com) lists the Microsoft 365 Copilot Business add-on at $21.00 a user a month, paid yearly, or $25.20 paid monthly. The page also shows "now starting from $18.00," a promotional price that runs from July 1 to December 31, 2026, for the first year only. It also lists Microsoft 365 Business Standard with Copilot at $23.50 and Business Premium with Copilot at $32.00, paid yearly. Each line notes usage-based billing for select advanced AI features, so usage-based pricing applies on Microsoft's small-business plans too.

Anthropic's [Claude pricing page](https://claude.com/pricing?ref=doitwithfriday.com) lists a Team plan for 2 to 150 people: a Standard seat at $20 a month billed annually, or $25 monthly, and a Premium seat with 5x the usage at $100, or $125 monthly. OpenAI's ChatGPT Business seat price was not readable in Friday's check, so it is left out here.

### Where the cost of AI at work hides

Seat prices are only the visible part of the cost of AI at work for a small team. The Copilot add-ons require a Microsoft 365 plan underneath. Claude Enterprise pricing has a 20 seat minimum, so a 12-person firm would pay for 20 seats. Agent work on Copilot draws Copilot Credits, and Claude Team users who hit a limit can turn on usage credits at standard API rates. In Friday's view, any quote for usage-based pricing should come with the spend cap you plan to set, written next to the seat price.

## How to price your own cost of AI at work

GigaOm recommends applying the study's approach to your own tasks and pricing, and Microsoft calls its findings directional. In Friday's view, a small team can run that test in a month.

### For example, a 12-person accounting firm

For example, a 12-person accounting firm already on Microsoft 365 adds the Copilot Business add-on at the $21.00 list price: 12 × $21 = $252 a month, paid yearly. During the promotion, first year only, 12 × $18 = $216\. The same firm on Claude Team Standard seats pays 12 × $20 = $240 a month, billed annually. Under Claude Enterprise pricing, the 20 seat minimum means 20 × $20 = $400 a month before any usage. In each case the seat is the floor of the cost of AI at work, and agent tasks or extra usage are billed on top.

### Run a one-month test on your own tasks

The post says GigaOm recommends choosing tasks that reflect your organization's work, running each repeatedly, assessing quality against defined criteria and applying your own pricing. Friday's recommendation for a small team: pick ten tasks people repeat every week, run them on two products, such as Copilot vs ChatGPT, for a month with spend caps on, and record the seat cost, any usage charges and the share of runs that passed your checklist. That total is the cost of AI at work for your team, and the usage-based pricing line on your own invoice is the figure to compare with a vendor chart.

### When a small workflow beats another seat

In Friday's view, not every task needs a seat. A job that runs the same way each time, such as sorting invoices or drafting a weekly report, can run as a small workflow billed by the model call, and the per-token prices in our [GPT-6.2 Astra and Sol pricing news](https://www.doitwithfriday.com/news/gpt-6-2-astra-sol/) show how that math works. A [workflow built around your own tools](https://www.doitwithfriday.com/services/workflows/) also leaves the records in systems you own.

## Read the study as a starting point

The study gives six figures on the cost of AI at work, from tests Microsoft ran and paid for. The useful part for a small business is the method, applied to its own work and its own invoices.

Before you settle Copilot vs ChatGPT or add more seats, write down your plan's seat price, its minimums, what it meters and the spend cap you will set. Then test with your own tasks for a month. The [Microsoft post](https://www.microsoft.com/en-us/copilot/blog/2026/11/16/what-ai-at-work-really-costs/?ref=doitwithfriday.com) lists its method and caveats. If you want help pricing one workflow against a seat, [send us the task you want to automate](https://www.doitwithfriday.com/contact/).

Last checked: November 18, 2026, against Microsoft's Copilot blog post of November 16, Microsoft's Copilot enterprise and business pricing pages, Microsoft Learn, Anthropic's Claude pricing page and OpenAI's ChatGPT pricing page.

Study figures, assumptions and US list prices come from Microsoft's, Anthropic's and OpenAI's own pages, checked November 18, 2026\. The study was funded by Microsoft, and the full GigaOm report was not reviewed. Friday has no relationship with Microsoft, GigaOm, OpenAI or Anthropic. Statements marked as Friday's view are opinion.