A manager approves five AI seats, then discovers that the expensive part is finding time for five people to use them properly. That is a useful warning for a Microsoft Copilot pilot: the subscription is only one line in the experiment.
This guide offers a planning worksheet for a small U.S. business. The dollar amounts in the worked example are hypothetical, not a Microsoft quote or a promise of savings. Use your actual subscription offer before making a purchase.
Start with the work, then choose the license
Write down the task you want to improve before choosing a plan. Preparing a weekly account update, assembling a first presentation draft, and summarizing a meeting are different jobs. Pick one repeatable task with an owner, a known input, and a clear definition of an acceptable result.
Microsoft's current business plan page distinguishes bundles, base subscriptions, billing terms, and options with or without Teams. Some bundles include work-grounded Copilot in Microsoft 365 apps. Compare the exact plan name and included features rather than assuming every product carrying the Copilot name provides the same access. Availability and prices should be confirmed at checkout.
Separate existing costs from new costs
If your team already pays for Microsoft 365, do not count that entire bill as a new AI expense. Calculate the difference between the present arrangement and the proposed arrangement. Conversely, if a different base plan is required for your intended workflow, include the upgrade rather than hiding it in the general IT budget.
Ask whoever manages the account to record the renewal date, commitment period, eligible users, and cancellation terms. A one-month evaluation window is not automatically a one-month financial commitment. Annual billing can leave you paying after the experiment ends.
A five-person budget you can change
Illustrative scenario: assume five incremental licenses at $30 per person per month, two training hours per participant, six administrator hours to prepare the pilot, and labor valued at $40 per hour. These are planning inputs, not verified vendor rates.
| Item | Calculation | First-month cost |
|---|---|---|
| Incremental licenses | 5 × $30 | $150 |
| Participant training | 5 × 2 hours × $40 | $400 |
| Setup and administration | 6 hours × $40 | $240 |
| Total | Before ongoing review and taxes | $790 |
The immediate subscription expense is $150 in this example, while $640 represents internal capacity used. Keeping those categories separate prevents the budget from pretending that salaried time is either free or an additional cash payment in every case.
Measure finished work, not impressive drafts
Take a small sample of the selected task before introducing AI. Record the time from opening the source material to getting an approved result. During the pilot, measure the same interval, including corrections, missing facts, and formatting. Do not stop the timer as soon as a draft appears.
For example, a report that used to take 35 minutes might take 12 minutes to draft with assistance and another 18 minutes to check. The net gain would be five minutes, not 23. That is still potentially useful, but it supports a very different purchasing decision.
Make the break-even assumption visible
At the illustrative $40 hourly labor value, recovering $150 of monthly license cost would require 3.75 hours of useful capacity each month across the five-person group. Recovering the full $790 first-month budget would require 19.75 hours. Neither calculation means payroll actually falls by that amount.
Training may have benefits beyond the first month, so show both the first-month result and an ongoing result. Use our AI cost calculator to change seats, setup, and review assumptions. Keep a separate note for any contractual commitment that lasts longer than the pilot.
Decide in advance what happens next
Give the pilot a decision date and three possible outcomes: expand, continue with a narrower use case, or stop. An example expansion rule might require four of five participants to use the workflow weekly, a measurable reduction in completion time, and no unresolved accuracy problems in the evaluated sample. These are suggested thresholds; choose yours before seeing results.
If only one person benefits, consider retaining a small deployment rather than buying seats for everyone. If the task was poorly chosen, change the task before concluding that every AI tool is useless. A pilot should make the next decision easier, even when the answer is to wait.
Sources and next steps
Product packaging reference: Microsoft's official business plan comparison, checked September 22, 2026. All budget assumptions and decision thresholds above are StackBrief's illustrative analysis. For the broader evaluation, read our Microsoft Copilot buyer's guide.