Microsoft Copilot Cowork Cost Estimator
Estimate your monthly Copilot Cowork consumption by role — Trust Insights
New here? How to estimate your cost
Whether you’re in IT planning a Copilot rollout or in Finance building a budget, this tool estimates what Microsoft Copilot Cowork will cost your organization each month. Cowork is billed by consumption: every task an employee hands to the AI spends Copilot Credits, and a complex job can cost many times more than a simple one. You give it three inputs — who uses it (Step 1), how heavily they use it (Step 2), and the pricing assumptions (Step 3) — and Step 4 returns your estimated monthly and annual cost, all-in and per user. You don’t need exact figures: the defaults reflect real Microsoft “Frontier” customer usage, so you can start there and adjust.
The four roles use Cowork very differently:
- Corporate Knowledge Workers — HR, finance, operations, marketing, internal comms. Mostly emails, summaries, and document work; lots of light and medium tasks.
- Customer-Facing Knowledge Workers — sales, customer success, account management, support. Proposals, briefings, and client-ready materials, so more medium tasks that produce real deliverables.
- Technical Workers — developers, engineers, data and IT analysts. Code review, QA audits, and data analysis — usually the heaviest mix, because multi-file and audit work runs long.
- Managers & Senior Leaders — team leads, directors, executives. Fewer prompts overall, but some very heavy ones such as board packs and scenario models.
Choosing your numbers: In Step 1, enter how many people are in each role. In Step 2, set how many Light / Medium / Heavy prompts each person runs per month — or open “Build from real tasks” to pick concrete tasks and let the tool fill in the mix. Not sure? Keep the defaults. Finance teams: tune the credit price and license cost in Step 3 to match your actual Microsoft agreement, and treat the result as a directional budgeting estimate.
Tip: tasks that look simple are often medium — opening and editing a Word file, for example, because of the document-handling skills the AI runs behind the scenes.
Step 1 — Who will use Cowork?
Example — edit to your org. Seed values show a representative 100-person org.
Step 2 — Prompt mix per role
Enter how many Light, Medium, and Heavy prompts each role runs per user per month — or build from real tasks below.
Step 3 — Assumptions
View / edit assumptions
100 credits = $1 (default, as of June 2026 — verify current Microsoft pricing). Defaults from the Microsoft Customer Cowork Estimator. The $30/user/mo license figure is illustrative — verify your actual Microsoft 365 Copilot license cost and update this field. Adjust all values as needed.
Understanding Light / Medium / Heavy tasks
These task tier definitions are illustrative — based on the Microsoft Copilot Cowork documentation and Trust Insights analysis.
- Light (125 credits default)
- Single-source lookups, short email drafts, one-action approvals. The agent reads one file or queries one system and returns a concise answer. No file writes.
- Medium (500 credits default)
- Multi-step tasks with 2–5 tool calls. Example: opening a Word doc triggers the docx skill, which loads ≈30K tokens of markdown, ≈60K tokens for a simple .docx, or ≈120K tokens total when the full docx editing skill is included — easily 10× a light prompt’s token count before any reasoning happens.
- Heavy (1,200 credits default)
- Agentic, multi-file, or overnight runs. Board packs, territory scoring, whole-repo refactors. 8+ tool calls, multiple deliverables, sustained runtime.
Why is opening a Word doc a medium task? In Trust Insights’ own testing, a plain-text summary runs ≈30K tokens (markdown). The same content as a .docx file is ≈60K tokens. Opening the file with the Copilot docx-editing skill adds another ≈60K tokens of skill metadata — so a single “open and edit” totals ≈120K tokens before any reasoning begins. That token weight places it firmly in the medium tier (and heavy when combined with extended reasoning).
