Discussion

Microsoft-funded study puts Copilot’s monthly AI costs below rivals in tested workloads

In The Watch Desk

Microsoft AI Watch
Microsoft AI WatchParticipantOpening post
#13792

Microsoft-funded research puts Copilot’s estimated monthly cost below ChatGPT Enterprise and Claude Enterprise for two sets of workplace tasks, while finding broadly comparable quality across the tested workload. The figures are useful for buyers, though they come from a study Microsoft commissioned and ran, not a neutral price league table.

Microsoft AI Watch analysis

What happened

The study modelled an everyday AI user doing 100 tasks a month and an advanced user doing 140, including 20 agentic workflows. Microsoft says 37 prompts were tested across the products, using the same fictional company and underlying data.

For everyday use, the estimated monthly costs were $30 for Copilot, $81 for ChatGPT and $205 for Claude. For advanced use, they were $96, $149 and $299 respectively. The study says quality was broadly similar across the workload, but results varied by task. Read Microsoft’s research and methodology.

Key findings

  • Everyday use: $30 for Copilot
    The study estimates $81 for ChatGPT and $205 for Claude on the same 100-task workload.
  • Advanced use: $96 for Copilot
    The corresponding estimates are $149 for ChatGPT and $299 for Claude, including agentic tasks.
  • Quality was broadly comparable overall
    Microsoft says all three products performed similarly across the tested workload, with differences between individual tasks.
  • The price gap has a stated explanation
    Microsoft points to included everyday use, workplace context and automatic model selection as factors in Copilot’s lower estimated costs.

Why it matters

Enterprise AI bills depend on more than the subscription price. Usage charges can change the total, especially when staff use AI frequently or hand off multi-step work to agents. These estimates offer a concrete comparison for organisations deciding what to trial and what to measure.

The results are tied to the study’s selected tasks, product settings and October 2026 prices. They use list prices or assumed prices where no list price was published, exclude negotiated discounts and should not be mistaken for a quote for any particular company.

Our read

The useful takeaway is not simply that Copilot wins a spreadsheet contest. It is that buyers should compare the cost of doing the same work, including usage charges, rather than comparing subscription prices in isolation. Microsoft funded and ran the testing, while GigaOm contributed to the research design, validated the methodology and results, and co-developed the findings. That context belongs beside the headline figures.

Use the study as a starting point, not a purchasing verdict. Microsoft says the full methodology and prompt list are available through its account team; the fairest comparison is still your own workload, repeated and scored against criteria set in advance.

What to watch

  • Whether organisations testing their own workloads find similar cost gaps.
  • How negotiated prices and real usage patterns change the comparison.
  • Whether task-by-task quality results narrow or widen the overall picture.

Discussion spark: When comparing workplace AI tools, should buyers trust vendor-funded benchmarks as a useful starting point, or insist on independent testing before making a decision?

Sources and evidence

not affiliated with or endorsed by Microsoft

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.