The year 2024 marked the moment generative AI became a premium line item in enterprise software contracts. Unlike cloud infrastructure, where vendors compete on per-unit pricing efficiency, GenAI add-ons are locked into existing vendor relationships. If you use Salesforce, you do not have the option to replace Salesforce Einstein with Claude or Llama. If you use Microsoft 365, you cannot replace Copilot with a third-party alternative without renegotiating your entire Microsoft agreement.
This analysis dissects GenAI pricing across the major vendors — Microsoft, Salesforce, SAP, and Google — and explains the negotiation positions available to CFOs and procurement teams.
The GenAI Add-On Land Grab: Why Now?
Overpaying for Enterprise Software? We handle software and cloud contract negotiation on a 25% gainshare basis — you keep 75% of every dollar saved. No retainer. No risk.
Get a free Enterprise Software savings estimate →Between 2022 and mid-2024, generative AI infrastructure was expensive to build and maintain in-house. Enterprise software vendors recognised they could package LLM API access into high-margin add-ons and sell them to existing customers without competitive risk. Switching costs in enterprise software are enormous. If your entire sales team runs on Salesforce, adding a $25/user/month Einstein tier is a semi-captive revenue increase.
The Three Pricing Tactics Vendors Are Using
Tactic 1: Pay-Per-User Add-Ons. Microsoft Copilot for Microsoft 365 is priced at $30/user/month as a standalone add-on to existing Microsoft 365 licensing. Salesforce Einstein Agents cost $10–$50/user/month depending on the feature tier. Google Duet AI adds $10–$30/user/month to Workspace.
Tactic 2: Bundling Into Existing Tiers. SAP and Oracle bundle GenAI features into higher-tier licenses and enterprise agreements, then discount the additional GenAI cost as a "bundled benefit" while raising the overall contract value. The effective GenAI cost is opaque but typically 15–30% of the platform increase.
Tactic 3: Pay-Per-API-Call or Usage-Based Pricing. Microsoft Copilot Studio and Salesforce Platform AI layers charge on message volume, requests, or API calls, with costs compounding as deployments scale. Google Vertex AI charges per 1M tokens processed.
The common thread: GenAI is sold at vendor list price, with minimal discount leverage available during renewals. The negotiation window is narrow, and most organisations do not recognise the AI add-on cost until it appears on a renewal invoice.
Microsoft Copilot for Microsoft 365: The $30 Per-User-Per-Month Add-On
Microsoft's GenAI strategy is the most transparent: Copilot is an add-on to Microsoft 365, priced separately from the base Microsoft 365 E3 ($20/user/month) and E5 ($35/user/month) licenses.
List Pricing and What's Included
Copilot for Microsoft 365: $30/user/month, on top of E3 or E5. Includes Copilot integration across Word, Excel, PowerPoint, Teams, Outlook, and Viva. Minimum deployment of 300 seats required, though enterprises can negotiate smaller pilot groups.
What "Copilot for Microsoft 365" actually delivers: Copilot assists with document composition, email drafting, meeting summaries, and spreadsheet analysis. It is Claude 3.5 Sonnet-equivalent in capability, though Microsoft's integration sometimes throttles throughput to manage API costs. Many enterprises report that Copilot performance lags in production environments with high concurrency, requiring organisations to maintain a mix of Copilot and non-Copilot users.
The Hidden Seat Counting Problem: Microsoft counts Copilot seats on active users, not licensed users. If you have 10,000 E3 licenses but only 5,000 active users in a month, Microsoft charges you for approximately 5,000 Copilot seats (plus a small buffer). This creates strong incentive to remove inactive licenses — but also opens negotiations around defining "active" users.
Copilot Studio and Power Platform AI: The Pay-Per-Message Trap
Beyond Copilot for Microsoft 365, Microsoft offers Copilot Studio (no-code generative AI deployment) and Power Platform AI (native AI capabilities in Power Apps, Power Automate, Power Pages). These operate on pay-per-message pricing models, where costs scale with usage.
Copilot Studio: $100 per 1,000 messages per month (minimum 1,000 messages/month, ~$100/month baseline). A deployed AI assistant that handles 500 conversations/month with 2–3 message exchanges per conversation reaches the 1,000–1,500 message threshold quickly.
Power Platform AI: Charges vary by capability ($3–$10 per 1M tokens for language models, additional fees for vision/search). A mid-scale Power Automate workflow using generative AI to summarise customer interactions can easily consume $500–$2,000/month in token charges.
These pricing models are where Microsoft's GenAI strategy becomes expensive at scale. A typical enterprise with 1,000 Power Automate flows and 10% involving generative AI is looking at $10K–$50K/month in incremental Platform AI charges.
Microsoft Copilot and Power Platform Costs Scaling Unexpectedly?
NoSaveNoPay helps enterprises renegotiate Microsoft 365 and Power Platform agreements to limit Copilot seat expansion, lock in minimum fees, and achieve volume discounts. 25% gainshare on every dollar saved — you keep 75%.
Explore Microsoft Negotiation Services →Salesforce Einstein AI and Agentforce: Layered Pricing by Use Case
Salesforce's GenAI strategy is multi-tiered and sold as distinct products rather than a single add-on. This obscures actual costs.
Einstein AI (Built-In to CRM Core)
Einstein Copilot (formerly Einstein for CRM) is bundled into Salesforce CRM Cloud licenses as a base capability, available at no additional charge. However, this version is limited: 50 CRM Analytics predictions per month, basic email recommendations, and limited generative response capabilities.
Einstein Pro (Add-On Tier)
Einstein Pro layers advanced AI capabilities on top of base CRM: $10/user/month (Essentials edition) to $50/user/month (Enterprise Edition). Includes unlimited CRM Analytics predictions, advanced copilot interactions, and Data Cloud AI features. Most enterprises adopting advanced AI features migrate to Einstein Pro during renewals.
Agentforce (Autonomous Agent Pricing)
Salesforce Agentforce (announced 2024) is a separate conversation-based AI product tier with usage-based pricing: $1–$2 per conversation (defined as one customer interaction, often spanning multiple turns). High-volume contact centres can run $50K–$500K/month in Agentforce fees depending on conversation volume and complexity.
For example, a 500-person customer service contact centre handling 10,000 customer interactions/month at an average cost of $1.50/conversation incurs $15,000/month in Agentforce fees, on top of core CRM and Einstein Pro.
The Bundling Problem: Enterprise Agreements and AI Cost Inflation
Most Salesforce customers are on Enterprise Agreements (multi-year, usage-based pricing with volume discounts). When renewing an Enterprise Agreement, Salesforce often increases base CRM pricing by 15–25%, then positions this as including "enhanced Einstein capabilities." The incremental Einstein cost is never stated explicitly; it is buried in the overall contract increase.
During renegotiation, Salesforce will rarely reduce the contract value — but will often accept removal of Agentforce or reduction in Einstein Pro seat commitments in exchange for a lower per-unit CRM price.
SAP Joule: Bundled Into RISE and BTP With Opaque Pricing
SAP's GenAI strategy is the most opaque of major vendors. Joule (SAP's GenAI assistant) is positioned as a "benefit" of RISE (SAP's cloud migration programme) and Business Technology Platform (BTP), but the actual cost allocation is variable.
SAP Joule Baseline
SAP Joule is available within RISE at no stated additional cost as long as you consume Joule within the RISE entitlements (typically limited to 100–500 queries/month depending on contract). Additional consumption is billed at usage rates negotiated as part of each RISE agreement.
For enterprise customers deploying Joule across Finance, Supply Chain, and HCM modules, the hidden usage-based charges often become substantial ($10K–$100K/month depending on deployment scale).
BTP and Generative AI Hub Costs
SAP Business Technology Platform, which underpins Joule, charges separately for AI/ML capabilities. A typical BTP AI Service subscription for Generative AI Hub costs approximately $5,000–$15,000/month depending on model usage and prompt volume.
Why SAP's Pricing Is Difficult to Negotiate
SAP has not published clear GenAI pricing tiers or per-unit costs. RISE contracts are custom-negotiated, and Joule costs are often embedded as a line item without itemisation. The result: enterprises cannot easily benchmark their Joule spend or compare it to competitors.
Negotiation leverage exists only if you can demonstrate that Joule costs are exceeding industry benchmarks (typically $2–$8 per query for enterprise deployments). SAP will sometimes bundle Joule into a revised RISE cost structure in exchange for locking in a higher overall RISE commitment.
Google Duet AI and Gemini for Workspace: The $10–30 Add-On
Google's Duet AI (rebranded as Gemini for Workspace in 2024) is positioned as a Workspace add-on rather than a base feature upgrade.
Gemini for Workspace Pricing
Gemini Business: $10/user/month on top of Workspace Standard/Plus/Business licenses. Includes Gemini integration in Gmail, Docs, Sheets, Slides, and Meet. Performance is generally strong but capability is narrower than Microsoft Copilot for Microsoft 365 (limited to productivity and communication workflows; lacks deep platform integration).
Gemini Enterprise: $20/user/month with advanced features (priority API access, extended context windows, broader integration). Minimum 10-seat commitment required.
Google Cloud Gemini and Vertex AI: Developer-Centric Pricing
For enterprises using Google Cloud AI services (Vertex AI, Gemini API), costs are measured in tokens per 1M, with pricing varying by model: approximately $0.01–$0.10 per 1M input tokens and $0.03–$0.30 per 1M output tokens depending on model and service tier.
Vertex AI Search and Recommendation (powered by Gemini) charges $0.15 per search query, with high-volume implementations easily reaching $10K–$100K/month.
Negotiability: Low
Google Duet/Gemini is the least negotiable of the major vendor GenAI add-ons. Google does not offer volume discounts on per-user pricing and rarely customises terms outside enterprise cloud contracts. Unlike Microsoft and Salesforce, Google has fewer switching costs and less motivation to lock in GenAI revenue.
GenAI Add-On Pricing Comparison: All Vendors
Below is a standardised comparison of GenAI add-on pricing across Microsoft, Salesforce, SAP, and Google as of March 2026.
| Vendor & Product | List Price | Metric | Min. Commitment | Negotiability | Typical Enterprise Cost |
|---|---|---|---|---|---|
| Microsoft Copilot (Microsoft 365) | $30 | per user/month | 300 seats | Medium (10–20% discounts available) | $8K–$15K/month (1K users) |
| Microsoft Copilot Studio | $100 | per 1K messages/month | $100/month minimum | Low (usage-based, minimal discounts) | $500–$3K/month |
| Microsoft Power Platform AI | $3–$10 | per 1M tokens | None | Low (pay-per-usage) | $5K–$30K/month |
| Salesforce Einstein Pro | $10–$50 | per user/month | 50 seats | Medium (bundled into contract) | $3K–$15K/month (1K users) |
| Salesforce Agentforce | $1–$2 | per conversation | None | Medium (can be excluded from renewals) | $10K–$100K/month (10K conversations) |
| SAP Joule (RISE-bundled) | Custom | per query/month (usage) | Bundled into RISE | Medium (custom negotiation) | $5K–$50K/month |
| SAP BTP Generative AI Hub | Custom | per BTP subscription | $5K/month minimum | Low (platform pricing) | $5K–$20K/month |
| Google Gemini for Workspace | $10–$20 | per user/month | 10 seats (Ent. only) | Very Low (no volume discounts) | $3K–$12K/month (1K users) |
| Google Vertex AI / Gemini API | $0.01–$0.30 | per 1M tokens | None | Very Low (published pricing) | $5K–$100K/month (high-volume use) |
Cost Compounding: The Stacking Problem
Most organisations deploying multiple GenAI features end up stacking costs across vendors. A typical enterprise with Microsoft 365, Salesforce CRM, and Power Platform automation might pay: $10K (Copilot) + $15K (Einstein Pro) + $20K (Power Platform AI) + $10K (Copilot Studio) = $55K/month in incremental GenAI add-ons. That's $660K annually in addition to base platform costs — and it is rarely visible in budget planning until the contract renews.
What Enterprises Actually Pay vs. List Price
List price is meaningless without context. The real negotiation question is: what discounts are available on GenAI add-on pricing?
Microsoft Copilot: 10–20% Discounts Possible
Microsoft offers modest volume discounts on Copilot ($30/user/month list price). Volume commitments of 3,000+ seats can achieve $24–$26/user/month pricing. Enterprises with existing Microsoft enterprise agreements can sometimes lock in a combined E3+Copilot or E5+Copilot bundle at a slight discount versus buying Copilot as a separate add-on.
Actual range: $24–$30/user/month depending on seat volume and contract negotiation.
Salesforce Einstein: 0–15% Discounts, But Often Bundled Into Contract Increases
Salesforce Einstein Pro ($10–$50/user/month) rarely receives explicit discounts. Instead, during Enterprise Agreement renewals, Salesforce will increase base CRM pricing by 15–25% and claim the increase includes "enhanced Einstein capabilities." The net effect is that you pay more for Einstein, but the cost is hidden in the overall contract increase rather than appearing as a separate line item.
Negotiation leverage: excluding Agentforce or specific Einstein features in exchange for a lower-than-proposed CRM price increase can sometimes save 10–15% on the incremental GenAI cost.
Actual range: $10–$50/user/month (minimal standalone discounts, but can be excluded from renewals).
SAP Joule: 0–10% Discounts, Often Buried in RISE
SAP does not publish GenAI add-on discounts. Joule is typically bundled into RISE pricing, making it impossible to benchmark. Custom negotiations can sometimes result in lower-than-standard RISE costs if you demonstrate you are using Joule at lower-than-assumed volumes.
Actual range: Custom pricing, typically $2–$8 per query depending on deployment scale.
Google Gemini: 0% Discounts, Published Pricing Only
Google does not negotiate on per-unit Gemini pricing. Published pricing is final. The only negotiation lever is volume commitments for Vertex AI usage, which can achieve 10–20% discounts on token pricing for multi-year commitments.
Actual range: No negotiation on per-user Workspace add-ons; Vertex AI discounts 10–20% for multi-year usage commitments.
⚠ GenAI Add-Ons Auto-Included in Contract Renewals
The most common enterprise mistake: assuming GenAI add-ons are optional during renewals. In practice, vendors include GenAI features in the renewal quote without explicitly calling them out. Many organisations discover they have been charged for 12 months of Copilot, Einstein Pro, or Joule only when the renewal invoice arrives. By that point, the contract has already been executed and GenAI features are locked in. Explicitly exclude GenAI add-ons from your renewal RFP unless you have already approved a business case for them.
Negotiation Strategy: When to Bundle vs. Standalone
GenAI add-ons present different negotiation opportunities at different contract stages.
Bundling GenAI Into Broader Contract Negotiations (Low Risk)
The least risky approach: negotiate GenAI add-on costs as part of a larger platform contract renewal. If you are renegotiating a Microsoft 365 agreement that covers 5,000 users, bundling Copilot into that negotiation as a line-item discount (10–15% off list price) is straightforward. The vendor incentive aligns: they win the overall contract and accept a modest GenAI discount in exchange for larger seat commitments.
Typical outcome: 10–15% discount on per-user GenAI pricing in exchange for 3–5 year commitment and volume growth targets.
Excluding GenAI Add-Ons from Renewals (Medium Risk)
If you have not yet deployed a GenAI feature at scale, the easiest negotiation is to exclude it from your renewal entirely. Salesforce and Microsoft will include Agentforce and Copilot in renewal quotes by default. Explicitly tell them in your RFP: "Do not include GenAI add-ons unless we have approved them in writing." Vendors often accept this condition, particularly if your base platform contract is large.
Typical outcome: GenAI add-ons are excluded from the renewal quote, deferring the decision to a future pilot or expansion.
Pilot-to-Production Terms (Higher Leverage)
If you want to trial a GenAI feature, negotiate a pilot clause explicitly into your contract: "First 6 months, 200 seats of Copilot at 50% discount, then renegotiation based on usage data." This prevents vendors from auto-including GenAI in your renewal after the pilot concludes. Vendors generally accept pilot clauses because they assume most pilots will convert to larger deployments.
Typical outcome: 50% discount on pilot seats for 6–12 months, then negotiated production pricing based on actual utilisation.
Renegotiating Usage-Based Pricing (Highest Leverage)
For usage-based GenAI pricing (Copilot Studio, Power Platform AI, Agentforce, Joule), the negotiation lever is usage audits. Request that the vendor share 12 months of historical usage data and cap future pricing at a multiple of average historical usage (e.g., "We will pay for up to 150% of historical average monthly usage, with any overages invoiced at 50% of list price"). This prevents bill shock and gives you predictability.
Typical outcome: 20–30% savings on expected usage costs by capping growth and negotiating overage rates.
NoSaveNoPay GenAI Negotiation Approach
NoSaveNoPay works with enterprises to manage GenAI add-on costs across three tracks: (1) excluding GenAI add-ons from renewals entirely unless there is an approved business case, (2) bundling GenAI pricing into broader platform negotiations to achieve 10–20% discounts, and (3) renegotiating usage-based pricing by capping growth and establishing overage rate discounts. We work on a 25% gainshare basis — all the savings you achieve, you keep 75% of.
Why Your IT Team Is Not Preventing GenAI Cost Overruns
If GenAI add-ons are so expensive and so negotiable, why do organisations pay list price?
Reason 1: AI Is Not Budgeted Centrally. GenAI features are often requested by business units (Sales wants Agentforce, Finance wants AI summarisation tools) and funded through departmental budgets rather than centralised IT procurement. By the time procurement is involved, the business case has already been approved and the vendor is included in the renewal quote.
Reason 2: Vendors Hide GenAI Costs in Contract Increases. Rather than presenting GenAI as a separate add-on, Salesforce and SAP bundle it into overall contract increases. A 20% contract increase might be 10% base platform growth, 5% GenAI, and 5% currency/volume effects. The GenAI component is never called out explicitly, so procurement does not negotiate it separately.
Reason 3: No Benchmarking Data. Most organisations do not know what GenAI add-ons cost at competitors. Without benchmarks, there is no baseline for negotiation. "Is $30/user/month for Copilot expensive?" Most procurement teams have no way to answer that question.
Reason 4: Time Pressure at Renewal. Enterprise software renewals often happen on compressed timelines. Procurement discovers GenAI add-ons in the renewal quote with 30–60 days to contract close. By that point, the business has already committed to using the AI feature, and vendors know exclusion is not an option.
Three Immediate Actions for Your GenAI Contracts
1. Conduct a GenAI Spend Audit
Pull 12 months of invoices for Microsoft 365, Salesforce, SAP, Google Cloud, and any other platform with GenAI add-ons. Create a line-item breakdown of GenAI costs. Most organisations discover they are paying for GenAI features they did not know they were consuming. This data is your negotiation baseline.
2. Establish an "AI Add-On Approval Process"
Before accepting any new GenAI add-on in a contract renewal, require business unit sign-off on: (1) the business case for the specific AI feature, (2) expected usage volume, (3) owner and budget. This prevents vendors from auto-including GenAI in renewals and gives you leverage to exclude features without business justification.
3. Benchmark GenAI Pricing in Your Industry
Ask your peers: "What are you paying for Copilot, Einstein Pro, Joule?" (Many organisations will not answer, but some will.) Compare your per-user costs and per-usage costs against the ranges in this analysis. If you are paying $35/user/month for Copilot, you should be paying $24–$28 based on your seat volume.