Copilot, Gemini, Amazon Q, the buyer side comparison
Microsoft 365 Copilot and Gemini for Workspace both anchor near $30 per user per month while Amazon Q undercuts them at $20, and the real decision is suite fit, data exposure, and how hard you push the pilot math, because the commercial facts, low active use and discountable list prices, are identical across all three.
Prepared by Redress Compliance · August 6, 2026 · GenAI negotiation advisory. Based on 18 to 25 enterprise seat negotiations advised 2024 to 2026.
Executive summary
The list prices anchor, the suites decide. Copilot and Gemini for Workspace anchor near $30 per user per month, Amazon Q Business near $20, and the rate difference matters less than suite gravity: the assistant living inside your dominant productivity suite beats a better model in a foreign one, because the assistant's value is the context it inherits and the workflow it interrupts least.
The defining commercial fact is low active use. Across our seat negotiations, weekly active rates settled at 25 to 45 percent of licensed seats after the novelty quarter, median near 38 percent, which converts every enterprise wide commitment into a majority payment for empty seats. The gap between licensed and active is the negotiation, whoever the vendor is.
Staging beats committing. Buyers who staged seats against measured weekly active use paid 30 to 50 percent less in year one than buyers who licensed entire divisions upfront, and the staged posture loses nothing: the seats are always available to add, and never available to remove.
The no discount posture is a posture. Committed multi year seat deals closed 20 to 30 percent below list across all three vendors despite list price anchoring in every first proposal. The remaining differentiator is data exposure: each assistant inherits its own suite's permission sprawl, and that risk lands on day one, before any productivity gain does.
The three assistants, compared where the money is
| Microsoft 365 Copilot | Gemini for Workspace | Amazon Q Business | |
|---|---|---|---|
| List anchor | Near $30 per user per month | Near $30, on eligible Workspace tiers | Near $20 |
| Where it wins | Microsoft 365 estates, which is most enterprises | Workspace estates, and dual vendor leverage plays | AWS centric estates and price led pilots |
| The commercial lever | Attach defense against the 100 percent default | Eligibility tiers and the dual vendor posture | The price wedge that disciplines the other two |
| The data exposure | Inherits SharePoint and Teams permission sprawl | Inherits Drive sharing sprawl | Inherits the connected repository permissions |
The active use gap, the fact that prices everything
The novelty quarter flatters every pilot: usage spikes, screenshots circulate, and the enterprise wide proposal follows. Then the weekly active rate settles, at 25 to 45 percent of licensed seats in our file, and the estate pays list for the majority who opened the assistant twice. The persona logic the Microsoft leverage guide applies to the Copilot attach, right sizing to 35 to 65 percent of actives on evidence, is the same logic all three vendors' seats need, applied before the commitment rather than regretted after it.
The staged ramp operationalizes it: an initial cohort sized to demonstrated demand, expansion gates tied to measured weekly active thresholds, and pricing locked for the expansion seats at signature. The construction cost buyers nothing and saved 30 to 50 percent in year one, because the vendor's alternative, seats that might activate, priced at list, is not a thing a prepared buyer pays for.
The assistant comparison analysis
The three assistants priced and compared end to end: suite fit, data exposure, the staging constructions, and the discount benchmarks by commitment shape.
Get the white paper →The negotiation, identical mechanics at three tables
The first proposal anchors at list with no discount posturing; the closed deals sat 20 to 30 percent below list on committed multi year shapes, at all three vendors. The levers are the seat estate's standard set: staged ramps with locked expansion pricing, caps on the renewal uplift, and the consumption lines, Copilot Credits against the MACC per the credits brief, Gemini's Vertex side per the Gemini licensing guide, negotiated as their own lines before commitment sizing.
The dual vendor posture is real money for the estates that can hold it: parallel Microsoft 365 and Workspace footprints carried 8 to 15 incremental discount points into assistant negotiations, and the multi year exclusivity clauses that surrender that posture are priced accordingly, by exactly one side of the table.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across seat negotiations, 2024 to 2026
Across roughly 18 to 25 enterprise GenAI seat negotiations Morten Andersen advised between 2024 and 2026, the gap between licensed seats and weekly active users was the defining commercial fact:
Where usage settled against licensed seats after the novelty quarter, across vendors and industries.
Year one savings for buyers who ramped seats against measured use instead of licensing divisions upfront.
The data exposure finding deserves its standing: each assistant surfaces whatever its suite's permissions allow, which means permission sprawl becomes answer sprawl on day one. The remediation, permission audits before rollout, is a security project with a licensing consequence: it defines the population that can safely activate, which is the honest ceiling on any seat commitment.
Your first five moves
- Let suite gravity pick the primary, and keep the other two priced as discipline, because the negotiation is with your incumbent either way.
- Measure weekly actives past the novelty quarter before any commitment; the median settles near 38 percent of seats.
- Stage the ramp: cohorts on demonstrated demand, expansion gates on active thresholds, expansion pricing locked at signature.
- Refuse the exclusivity clause where a dual vendor posture exists; it is worth 8 to 15 points, annually, to the side that keeps it.
- Audit permissions before rollout, because sprawl becomes answers on day one, and the safe population is the commitment ceiling. The GenAI practice runs the table with you.
Frequently asked questions
How much do Copilot, Gemini, and Amazon Q cost?
Microsoft 365 Copilot and Gemini for Workspace anchor near $30 per user per month at list, Amazon Q Business near $20. Committed multi year deals closed 20 to 30 percent below list across all three in our negotiations, despite uniform no discount posturing in first proposals.
Which AI assistant should our enterprise choose?
Usually the one inside your dominant productivity suite: the assistant's value is the context and permissions it inherits, so suite gravity beats model comparisons in practice. The other two earn their place as priced alternatives that discipline the incumbent's negotiation, worth 8 to 15 points where the posture is credible.
How many licensed AI assistant seats actually get used?
Weekly active rates settled at 25 to 45 percent of licensed seats after the novelty quarter in our engagement file, with the median near 38 percent. The gap between licensed and active is the central commercial fact, and it is why staged ramps beat enterprise wide commitments by 30 to 50 percent in year one.
Can you negotiate discounts on Copilot or Gemini?
Yes: committed multi year seat deals closed 20 to 30 percent below list across all three vendors. The levers are the standard seat estate set, staged ramps with locked expansion pricing, uplift caps, and the consumption lines negotiated separately, opened before the commitment is sized rather than after.
What is the data risk of rolling out an AI assistant?
Each assistant surfaces whatever its suite's permissions allow, so existing permission sprawl, over shared drives, open sites, stale access, becomes answer sprawl the day the assistant activates. The permission audit belongs before the rollout, and it defines the population that can safely hold a seat.
Should we commit to enterprise wide AI assistant seats?
Not against a 38 percent median active rate. The staged construction, initial cohorts on demonstrated demand, expansion gates on measured thresholds, pricing locked for the expansions, kept every option and saved 30 to 50 percent in year one. Seats can always be added; they can almost never be removed.
