Copilot, Gemini, Amazon Q. Priced side by side for 2026.

Microsoft 365 Copilot and Gemini for Workspace both anchor near 30 dollars per user per month while Amazon Q undercuts them at 20, so the real decision is suite fit, data exposure, and how hard you push the pilot math.

How do Copilot, Gemini, and Amazon Q price per user in 2026?

Microsoft 365 Copilot and Gemini for Workspace both anchor near 30 dollars per user per month on annual commitment, while Amazon Q Business lists near 20 dollars with a lighter lite tier below it.

Microsoft 365 Copilot requires a qualifying Microsoft 365 base license. Gemini for Workspace has folded into Workspace editions with AI included at higher per seat rates.

The sticker is only the anchor. Every proposal we benchmarked opened at list, and every committed deal closed below it.

Why does the base license requirement matter?

Because it reprices the assistant. A Copilot seat on top of E3 is a different total cost than the same seat on top of E5 under the Microsoft 365 enterprise plans, and Microsoft uses Copilot interest to pull E5 upgrades through. Price the stack, not the add on.

Where does each assistant actually win?

Suite gravity decides most deals: Copilot wins Microsoft estates, Gemini wins Google estates, and Amazon Q wins developer and AWS operations workloads. Cross suite bets underperform because adoption depends on the assistant living where people already work.

• Copilot: strongest in Teams meeting summaries, Outlook drafting, and Excel analysis on M365 data.
• Gemini: strongest in Gmail, Docs collaboration, and Meet, with simpler packaging after the 2025 edition changes.
• Amazon Q: strongest for developers, AWS operations, and Q in Connect contact centers, weakest as a general office assistant.

Should you run two assistants side by side?

Usually no. Split estates double the data governance work and halve the negotiation leverage. The exception is a developer population on Amazon Q while the office runs Copilot or Gemini, because the workloads barely overlap.

What are the lock in and data exposure risks?

Each assistant inherits the permission sprawl of its own suite, and oversharing surfaces on day one. Copilot reads everything a user can touch in SharePoint and OneDrive, Gemini reads Drive, and Q reads the connectors you grant.

• Permission debt: years of open shares become visible the week the assistant ships.
• Renewal lock: assistant seats deepen suite dependence, which compounds at the next EA or Workspace renewal.
• Usage data asymmetry: the vendor sees your adoption telemetry; demand the same dashboards in the contract.

What contract terms limit the damage?

Three clauses matter most: a mid term seat reduction right tied to measured usage, a price hold covering at least one renewal, and a commitment that your prompts and grounding data never train shared models.

How should you run the pilot and the negotiation?

Run a 90 day pilot on 5 to 10 percent of the target population, measure weekly active use, and let that number size the order. Buying for everyone and hoping for adoption is the single most expensive pattern in our file.

1. Pick pilot users from the teams with the heaviest document and meeting load.
2. Define weekly active use as the success metric before the pilot starts.
3. Fix permission sprawl in SharePoint or Drive before seat one ships.
4. Take the measured active rate into the seat negotiation as your volume case.
5. Stage the ramp: quarterly seat tranches priced at the committed discount.

What does a good first order look like?

Seats for measured adopters plus 20 percent headroom, a locked unit price for the term, and a written option to add tranches at the same rate. That structure paid for itself within two quarters in every engagement where it landed.

Where the common advice on enterprise GenAI assistants is wrong

The standard advice is to standardize early on one assistant for the whole company so nobody falls behind. We disagree. In roughly 18 to 25 GenAI seat negotiations Morten Andersen advised in 2024 to 2025, enterprise wide day one rollouts produced weekly active rates of 25 to 45 percent, which means most seats were pure waste at 30 dollars per user per month. The buyer side move is to pilot narrow, measure weekly active use, and buy in staged tranches at a committed discount. The vendors price on fear of missing out. Your usage data is the antidote.

What the engagement data shows

Three cuts of our advisory engagement file frame the size of the opportunity.

What does each assistant really cost once you add the prerequisites?

The sticker price is never the invoice. Every one of these tools is an add-on that assumes you already own — or must buy — an underlying suite, and that prerequisite is the largest hidden line item. Microsoft 365 Copilot lists at about $30 per user per month on an annual commitment, but it requires M365 E3 or E5 (or Business Standard/Premium). An E3 seat at roughly $36 plus Copilot means the true all-in is $66-plus per user per month, not $30. Google flipped its model in January 2025: it discontinued the standalone Gemini add-ons and baked Gemini into Workspace with a roughly 17 percent suite price rise, while the newer Gemini Enterprise platform sits on top at $30 (12-month commit) or $35 without. Amazon Q Business Pro is $20 but you also pay for the retrieval index by the unit-hour.

Buyer move: normalise every quote to a fully-loaded per-seat-per-month figure that includes the prerequisite suite delta and any mandatory index or consumption floor. The vendor with the lowest add-on price is frequently not the lowest all-in.

Which assistant fits which workload, and where does each actually win?

Fit should drive the SKU, not the incumbent relationship. Microsoft 365 Copilot wins where the work lives in Outlook, Teams, Word, Excel and SharePoint, grounded in the Microsoft Graph, with Copilot Studio for custom agents. Gemini wins for Workspace-native organisations, because it is now embedded in Gmail, Docs, Sheets and Meet at no separate add-on, so the marginal cost of basic AI is effectively zero at the suite level. Amazon Q wins in two distinct lanes: Q Developer ($19) is a coding and cloud-operations assistant competing with GitHub Copilot, while Q Business ($20) is a permission-aware enterprise-search layer over sources like S3, Salesforce, SharePoint and ServiceNow.

Buyer move: these three are not fully substitutable. Q Developer and M365 Copilot solve different problems, so a "pick one" framing set by a single vendor’s account team is a negotiation trap. Run the fit matrix against your top three workloads first — the right answer is often two narrow deployments, not one broad one, which also weakens any single vendor’s leverage.

How does consumption pricing turn a fixed cost into a variable one?

The per-seat license is the predictable part. The consumption layer — where agents, messages and credits are metered — is where budgets break. Microsoft Copilot Studio is the clearest example: as of 1 September 2025, Microsoft replaced "messages" with Copilot Credits, sold in prepaid packs of 25,000 credits a month (widely around $200 a pack), with pay-as-you-go billed to Azure. Credit burn scales with agent complexity, not headcount — a generative answer costs about 2 credits, an agent action about 5, tenant-graph grounding about 10, and premium tools about 100 per ten responses — so a richly grounded answer can cost 10 to 15 times a basic reply, and unused prepaid credits do not roll over. Amazon Q has parallel meters (index-unit hours independent of seats; Q Developer overage on code transformation), and Gemini Enterprise layers Google Cloud consumption on top of the seat price.

Buyer move: before signing, demand a modelled monthly credit or index-hour forecast at your target volume, a hard cap or budget alert in the contract, and written rates locked for the term — Microsoft already changed the billing unit once in 2025. Treat consumption as a separate, capped budget line from seats; never assume it is "included."

How do data residency, IP indemnity and lock-in compare?

For a buyer the risk clauses are worth more than a discount point. All three offer IP/copyright indemnification on paid tiers — Microsoft’s Copilot Copyright Commitment, Google’s Gemini/Workspace indemnification, and Amazon Q Developer’s indemnity — but Q’s applies only on the Pro tier; the free tier explicitly excludes it. On data residency, Microsoft offers the EU Data Boundary, Google offers data regions plus Assured Controls, and AWS inherits region selection; all three state they do not train foundation models on enterprise prompts on paid tiers, but that belongs in your order form, not just a trust-centre page. Lock-in differs by vendor: Microsoft through the E3/E5 prerequisite and Graph-grounded agents, Google through suite embedding, and AWS through index architecture and connector configuration that do not port out.

Buyer move: put four things in writing — IP indemnity confirmed for the exact SKU you are buying (do not let a free pilot hide a Pro requirement), a named data-residency region, a no-training clause, and an exit/data-portability provision covering agents and indexes. Never rely on a public trust page as your contractual protection.

What to do next

Five moves turn this analysis into a lower invoice on the next renewal.

A sequence you can run this quarter

1. Map which productivity suite dominates each user population before shortlisting.
2. Audit SharePoint or Drive permission sprawl before any assistant pilot.
3. Run a 90 day pilot on 5 to 10 percent of users with weekly active use as the metric.
4. Price the full stack including base license upgrades, not the add on alone.
5. Negotiate staged seat tranches with a locked unit price and a reduction right.
6. Demand adoption dashboards and a no training commitment in the order form.

White Paper · Microsoft

Copilot vs Gemini vs Amazon Q. The negotiation framework

Microsoft 365 Copilot vs Google Gemini vs Amazon Q in 2026: the per user metric, the prerequisite stack, and the EA, EDP, and CUD bundle math. Read it free.

Read the white paper

Frequently asked questions

Is Amazon Q a direct replacement for Microsoft 365 Copilot?

No — they solve different problems: Q Developer ($19) targets coding and AWS operations while Q Business ($20) is enterprise search, neither replacing Copilot’s Office-document authoring inside M365. Treat a vendor’s "pick one" framing as a negotiation tactic and match the tool to the workload instead.

Now that Gemini is bundled into Google Workspace, is it "free"?

Not free — Google removed the separate ~$20 to $30 add-ons in January 2025 but raised Workspace prices about 17 percent to embed Gemini, so you pay through the suite, and the newer Gemini Enterprise platform ($30-plus per user per month) is still a paid layer on top.

Is Microsoft 365 Copilot or Gemini for Workspace cheaper in 2026?

They anchor at similar levels, near 30 dollars per user per month on annual commitment, so the real cost difference comes from the required base licenses and the discount you negotiate. Copilot needs a qualifying Microsoft 365 plan, while Gemini now ships inside Workspace AI editions.

How much cheaper is Amazon Q than Copilot or Gemini?

Amazon Q Business lists near 20 dollars per user per month, roughly a third below the other two. It is strongest for developer and AWS operations use, and weakest as a general office assistant inside documents and email.

Can you negotiate Copilot, Gemini, or Amazon Q pricing?

Yes. Committed seat deals closed 20 to 30 percent below list across all three vendors in our 2024 to 2025 engagement file. Volume, multi year terms, and staged tranches at a locked unit price are the levers that move the number.

What adoption rate should we expect from a GenAI assistant rollout?

Plan for 25 to 45 percent weekly active use after the novelty quarter, with a median near 38 percent in our engagements. Size the order on measured pilot adoption, not on headcount.

Do these assistants train on our company data?

All three vendors state that enterprise customer content is not used to train shared models under business terms, but the commitment belongs in your contract. Put a no training clause in the order form rather than relying on policy pages.

Should we wait for prices to fall before buying?

Waiting has a real option value because per seat list prices are under pressure from bundling. If you buy now, lock a price hold through renewal and a seat reduction right so a falling market cannot strand you above it.