Most enterprises don't have an AI problem — they have an activation problem. SAP, Microsoft, Oracle and the rest have already shipped more embedded AI than most organisations use. This tool estimates what that unused capability is worth in your stack — from public list prices and published adoption research.
Two components, three scenarios, every assumption on the table. This is an estimate built to be argued with — not a benchmark.
AI you explicitly bought — Microsoft 365 Copilot seats ($30/user/mo list), Salesforce Agentforce add-ons ($125/user/mo), ServiceNow Now Assist uplifts — valued at list price for the share of seats that published research says goes unused.
The anchor: reporting synthesising Microsoft disclosures with Gartner/Forrester survey data puts weekly-active usage at 20–30% of purchased Copilot seats, and under 40% daily-active even in committed programmes. The three scenarios assume 40% / 30% / 20% of paid AI seats are genuinely active.
Formula: seats × list price × 12 × (1 − active rate)
The heart of the activation gap: capability already inside subscriptions you pay for. Oracle ships 100+ embedded agents and Agent Studio at no extra cost with Fusion. SAP bundles base Joule with RISE/GROW cloud contracts. Microsoft includes Copilot Chat with eligible M365 plans. Salesforce Foundations grants 100,000 free Flex Credits. SAP AI Units and Workday Flex Credits expire after 12 months — unused balances are simply forfeited.
Because included features carry no invoice line, they're valued at a shadow price — what comparable capability costs where it is priced (e.g. ServiceNow charges a $50–75/user AI uplift for capability Oracle bundles at €0). Shadow prices are clearly estimates; the conservative scenario sets most of them to zero, so the floor of the range is built almost entirely on invoiced numbers.
Formula: users × shadow price × 12 × (1 − activation rate) + forfeited prepaid AI credits
| Assumption | Conservative | Typical | Aggressive | Basis |
|---|---|---|---|---|
| Paid AI seats active | 40% | 30% | 20% | Copilot weekly-active reporting, 2026 solid |
| Copilot Chat shadow value (per M365 seat/mo) | €0 | €2 | €5 | Included at $0; full Copilot $30 ceiling estimate |
| Oracle embedded agents shadow (per Fusion user/mo) | €0 | €10 | €34 | ServiceNow's $50–75 AI uplift as market comparable estimate |
| SAP base-Joule shadow (per SAP user/mo) | €0 | €2 | €5 | Copilot-Chat analogue estimate |
| Included-AI activation rate | 25% | 10% | 5% | MIT: 95% of GenAI pilots deliver no ROI proxy |
| Prepaid AI credits forfeited (SAP AI Units, Workday) | 25% | 46% | 70% | S&P Global: firms scrapped 46% of AI PoCs proxy |
| Now Assist utilisation | 60% | 30% | 20% | Analogy to Copilot data estimate |
Currency: USD list prices converted at $1 ≈ €0.86 (mid-2026).
Honest footnote: utilisation evidence is strongest for Microsoft Copilot; applying it to other vendors is an analogy. Shadow prices value entitlements at what vendors charge elsewhere — not at realised business value. Both cuts are why the conservative scenario exists.