Table of Contents
The Embedded AI Licensing Problem
Enterprise software vendors discovered something important in 2024 and 2025: AI is the most effective licence fee expansion mechanism they have seen in a decade. The pattern is consistent across every major vendor: embed AI capabilities into existing products, position them as materially new functionality, create premium edition tiers that include AI, and charge existing customers significantly more to access the product they already use.
Microsoft charges $30 per user per month for Copilot on top of M365 E3/E5 — a 50–70% premium on the existing licence cost. Salesforce charges $50–75 per user per month for Einstein and Agentforce capabilities. SAP requires upgraded S/4HANA Cloud editions to access Joule AI features. Oracle bundles AI features into higher licence tiers priced 30–40% above standard editions.
The cumulative impact is significant. An enterprise with 10,000 M365 users, 2,000 Salesforce users, and 1,500 SAP users could face an additional $8–15M per year in AI add-on costs from vendors whose base products are already consuming $30–50M annually. These increases are being positioned as optional — but they are not, in practice, optional if your competitors are deploying AI and you are not.
How Each Major Vendor Monetises Embedded AI
Microsoft: Copilot as Universal Add-On
Microsoft's embedded AI strategy centres on Copilot — a branded AI layer applied across the entire Microsoft product portfolio. Microsoft 365 Copilot charges $30/user/month on top of existing M365 licensing. Microsoft Dynamics 365 Copilot is priced at $50/user/month. GitHub Copilot charges $19–39/user/month for developers. Azure AI Services are consumption-priced with significant variance by service tier.
The critical question for existing EA customers is whether Microsoft Copilot constitutes a new product (requiring separate purchase) or an enhancement to an existing licensed product (potentially covered by your existing agreement). Microsoft's position is uniformly that it is a new product. But enterprise contracts signed before 2022 often contain "product enhancements" language that can be argued to cover functionality extensions to licensed products.
Microsoft's EA renewal process is the primary leverage point for Copilot pricing. At renewal, Copilot seats are typically offered at 15–20% below list price for large enterprise agreements. Significant additional discounts — up to 35% — are available for multi-year Copilot commitments structured alongside the core EA renewal.
Salesforce: Einstein and Agentforce
Salesforce has pursued the most aggressive AI monetisation strategy of any CRM vendor. Einstein AI features are distributed across multiple pricing tiers, with the most capable features requiring Einstein 1 Platform editions that cost 60–100% more than Professional or Enterprise editions. Agentforce — Salesforce's autonomous AI agent platform — is priced separately at $2 per conversation, with enterprise commitments negotiated in conversation volume bundles.
Salesforce's bundling strategy means that many organisations are paying for Einstein features they do not use, embedded in edition upgrades that were sold primarily on the basis of other capabilities. The first step in Salesforce AI cost management is determining which AI features your organisation actually uses — the answer is often far fewer than the Salesforce account team would suggest.
SAP: Joule and BTP AI
SAP's AI embedding strategy operates at two levels. Joule — SAP's AI copilot — is progressively embedded across SAP SuccessFactors, SAP Ariba, and SAP S/4HANA Cloud, and access typically requires the current cloud edition with AI capabilities enabled. SAP Business Technology Platform (BTP) contains the AI Foundation services that power these features, and BTP consumption costs are a significant and often underestimated component of total SAP AI spend.
SAP on-premise customers are in a particularly exposed position. Joule is a cloud-native capability that does not exist in on-premise S/4HANA deployments. SAP is using AI as a migration forcing function — presenting AI capabilities as a reason to migrate to S/4HANA Cloud at significant additional cost. This context should inform your S/4HANA migration negotiation strategy.
Oracle: AI in the Autonomous Everything Strategy
Oracle has embedded AI throughout its product portfolio under the "autonomous" branding — Autonomous Database, AI-powered ERP analytics, and Oracle Digital Assistant. Oracle's monetisation approach typically involves: upgrading existing database licences to Autonomous variants at 30–40% premium, requiring Oracle Cloud subscriptions for AI analytics capabilities that previously ran on-premise, and bundling AI features into higher APEX and Analytics Cloud tiers.
Oracle's strong audit posture creates a specific risk around AI feature adoption. Enabling AI features in Oracle products — even features marketed as included in your licence — can create audit exposure if Oracle subsequently reclassifies the feature as requiring a separate licence metric. Contract review before enabling Oracle AI features is essential.
| Vendor | AI Product | Typical Cost | Licence Model | Negotiation Window |
|---|---|---|---|---|
| Microsoft | M365 Copilot | $30/user/month | Per user add-on | EA renewal |
| Salesforce | Einstein / Agentforce | $50-75/user or $2/conversation | Edition upgrade or consumption | Annual renewal |
| SAP | Joule / BTP AI | BTP consumption + edition premium | Cloud edition + consumption | Maintenance renewal |
| Oracle | Autonomous DB / AI Analytics | 30-40% over standard licence | Edition upgrade or cloud SKU | Support renewal or ULA |
| ServiceNow | Now Intelligence / AI Pro | 20-35% premium on platform | Platform tier upgrade | Annual ACV renewal |
What Your Existing Contract May Entitle You To
Before accepting a vendor's position that embedded AI requires new licensing, review your existing agreement for several provisions that may entitle you to access AI features as part of your current licence.
Product Updates and Enhancements Language
Enterprise software contracts typically contain language addressing the customer's rights to product updates, upgrades, and enhancements made available during the support term. The critical distinction is between an "upgrade" (a new product version) and a "new product" (a separately licensed offering). Vendors characterise embedded AI as new products; customers with broad enhancement rights may have grounds to characterise them as product updates.
Review your support and maintenance agreement for language such as: "includes access to all product enhancements released during the support term" or "licensee is entitled to all generally available versions of the licensed software." These provisions were written before AI was a consideration, and the ambiguity they create is a legitimate basis for negotiation.
Bundle and Suite Rights
If you licence a suite of products rather than individual products, your suite rights may be broader than you realise. Oracle Cloud Applications suites, Salesforce platform licences, and SAP RISE contracts often include "roadmap" commitments or "platform" rights that can be argued to encompass subsequently released AI capabilities on the platform.
Data Rights Risks with Embedded AI
Beyond cost, embedded AI creates data governance risks that many organisations have not addressed. When Salesforce Einstein analyses your CRM data, or SAP Joule accesses your HR records, or Microsoft Copilot reads your email and documents — your business data is being processed by AI systems under terms that were not negotiated with AI in mind.
The three primary risks are: your data being used to train or improve the vendor's AI models; your data being processed by third-party model providers (OpenAI, Anthropic, Google) not disclosed in your original vendor contract; and AI-generated insights derived from your proprietary data being used to benefit other customers indirectly through model improvement.
Each of these risks requires explicit contractual address. Do not assume your existing data processing agreement covers AI processing — it almost certainly does not. For guidance on the specific contract language required, see our article on AI governance contract requirements and our AI Contract Red Flags white paper.
Negotiating Embedded AI Licensing Costs
The most important principle in embedded AI licensing negotiation is: do not negotiate AI add-ons in isolation from the underlying platform renewal. Vendors structure AI pricing proposals separately to prevent you from using the base platform relationship as leverage. Insist on negotiating the total commercial relationship — base licences, support, and AI add-ons — together.
Pilot Pricing and Phased Adoption
For AI capabilities where adoption is uncertain, negotiate pilot pricing before committing to enterprise-wide deployment. A 90-180 day pilot at 20–30% of the enterprise licence count, with an option to expand at negotiated rates, protects you against paying for AI features that do not deliver the value the vendor's account team projects.
Consumption-Based Caps and Credits
Where AI features are consumption-priced (Agentforce conversations, Azure OpenAI API calls, Oracle AI service credits), negotiate caps and pre-purchased credit structures that set a ceiling on unplanned spend. AI consumption can escalate rapidly as users discover new use cases. Without consumption governance, embedded AI features can generate significant budget overruns within quarters of activation.
Competitive Alternatives as Leverage
The strongest negotiating position for embedded AI is demonstrating that standalone alternatives are viable. Microsoft Copilot for M365 is an $30/user/month add-on. Standalone AI writing and productivity tools exist at $15–20/user/month. Salesforce Einstein can, for many use cases, be partially replicated by integrating a general-purpose AI API with your CRM data through middleware. The threat need not be fully credible to create leverage — but it must be specific enough that the vendor's account team has to work to refute it.
Governance Requirements for Embedded AI
Activating embedded AI features in enterprise software platforms requires governance decisions that many organisations are deferring — at increasing regulatory risk. Before enabling Salesforce Einstein AI decisions in your hiring workflows, Oracle AI analytics in your compliance processes, or Microsoft Copilot in your financial reporting, consider:
- Which embedded AI features require EU AI Act high-risk classification review?
- Do the vendor's standard terms for embedded AI meet your data processing obligations?
- Has your legal team reviewed the AI-specific terms added to your vendor agreements in the last 18 months?
- Have you conducted an algorithmic impact assessment for AI features making decisions about individuals?
- Do you have audit trail capabilities to explain AI-generated decisions if challenged by regulators?
The governance framework for embedded AI is covered in detail in our article on AI governance contract requirements. For enterprise-scale AI contract review and negotiation, see our AI Procurement Advisory service.