AI and Your Existing Software Budget: The License Audit You Should Have Run Last Quarter

Executive Summary

• The average enterprise wastes $19.8 million per year on unused SaaS licenses. Only 54% of purchased licenses are actively used (Zylo, 40M+ licenses analyzed, 2026 SaaS Management Index). AI features are now embedded in those licenses — which means organizations are paying for AI capabilities nobody activated, nobody governed, and nobody measured.
• 77% of IT leaders discovered AI-powered features running in their environment without IT’s knowledge (Zylo, 2026). The shadow AI problem has a quieter twin: authorized AI spend that is invisible because it arrived inside a contract renewal, not a purchase order.
• Microsoft, Google, Salesforce, and Zoom have all embedded AI into base pricing in the past 12 months. Microsoft 365 E3 rises from $36 to $39/user/month in July 2026, bundling Copilot Chat. Google Workspace increased 17-22% to absorb Gemini. Salesforce’s Agentforce add-on starts at $125/user/month. These are not optional features — they are line items the CFO is already paying for.
• This worksheet identifies five questions the CIO and CFO should run with the IT team on Monday morning to determine what AI capabilities exist in the current software stack, what they cost, whether anyone uses them, and whether company data is flowing through them.

The Problem: AI You Are Paying For But Not Managing

Most mid-market companies approach AI as a future initiative — something to budget for, pilot, and deploy. The reality is different. AI is already in the building. It arrived through software renewals, not strategy sessions.

Gartner predicts 40% of enterprise applications will feature task-specific AI agents by the end of 2026, up from less than 5% in early 2025 (Gartner, August 2025). The question for a 300-person company is not “should we adopt AI?” but “what AI did we already adopt without knowing it?”

Three categories of hidden AI spend exist in the typical mid-market software stack:

The shadow AI discovery worksheet (the companion to this tool) finds unauthorized tools employees brought in themselves. This audit finds authorized tools the organization brought in but forgot about.

The Five Monday Morning Questions

Question 1: What AI features are active in the tools you already pay for?

How to check: Pull the current software inventory. For each of the five platforms below, determine the AI feature status.

What you are looking for: Features that are on but ungoverned, and features that are off but paid for. Both are problems. The first is a security and compliance exposure. The second is waste.

Question 2: What is the AI component of your current software spend?

How to check: For each platform, calculate the AI-attributable cost per seat per month.

A 300-person company on Microsoft 365 E3 and Google Workspace Business Standard is spending approximately $18,000 per year on embedded AI before buying a single AI tool. That number rises to $126,000+ if 100 users have full Copilot licenses at $30/seat/month. At a company with Salesforce Enterprise plus Agentforce for a 50-person sales team, the AI add-on alone runs $75,000/year.

What this number tells the CFO: This is not new money. It is already in the budget. The question is whether the organization is getting value for it.

Question 3: What is the utilization rate of these AI features?

How to check: Each platform provides usage analytics. Most IT teams have never looked at them for AI features specifically.

Benchmark context: Microsoft 365 Copilot has a 35.8% active usage rate among licensed users (Stackmatix, 2026 analysis). Only 3.3% of Microsoft 365’s 450 million users pay for Copilot at all (Microsoft Q2 FY2026 earnings, February 2026). Salesforce reports 67% of Einstein implementations face significant user adoption challenges (Oliv.ai, 2025 analysis of enterprise deployments).

The mid-market pattern: IT purchases licenses in a batch. A handful of enthusiasts use them. The rest sit idle. At $30/user/month for full Copilot, 100 unused licenses cost $36,000/year. At $125/user/month for Salesforce Agentforce, 20 unused seats cost $30,000/year.

Question 4: Is company data flowing through AI features nobody approved?

How to check: This is the governance question the CISO and GC need answered.

Why this matters: Google’s Gemini features in Workspace are on by default for Business Standard and above. Zoom AI Companion is on by default for all paid plans. If the organization never changed the defaults, AI has been processing company data — emails, documents, meeting transcripts — without a governance review.

This is distinct from the shadow AI problem. Shadow AI involves employees choosing to use unauthorized tools. This is the vendor choosing to activate AI on the organization’s data. The acceptable use policy (the companion to this tool) governs employee behavior. This audit governs vendor behavior.

Question 5: What is the net position — waste, exposure, or opportunity?

How to score: Tally findings from Questions 1-4 into three buckets.

Decision framework:

If the dominant finding is waste: The organization has budget available for AI without a new line item. Redirect existing AI spend toward the functions where the evidence shows measurable returns (see the task selection card for which tasks those are). A company spending $36,000/year on unused Copilot licenses can fund a governed pilot with training and measurement — and get more value from money already allocated.

If the dominant finding is exposure: Pause and govern before expanding. The acceptable use policy template provides the Day 1 framework. The vendor contract red lines card provides the negotiation leverage. Neither is useful without knowing what is already active, which is what this audit just revealed.

If the dominant finding is opportunity: The tools are paid for, the governance is in place, and the gap is adoption. This is a training problem, not a procurement problem. The upskilling pocket guide provides three training models for mid-market companies.

Key Data Points

What This Means for Your Organization

The mid-market company that has not run this audit is in one of two positions, both expensive. Either the organization is paying for AI features nobody uses — shelfware that shows up as cost without value — or AI features are active and processing company data without governance, which shows up as risk without awareness.

A 300-person company on Microsoft 365 E3, Google Workspace Business Standard, and Zoom Business is already spending approximately $24,000-$30,000 per year on embedded AI through price increases alone. If 50 users have full Microsoft 365 Copilot licenses and only the benchmark 36% are active, the organization wastes roughly $11,500 per year on that single line item. Multiply across Salesforce, ServiceNow, and HubSpot, and the hidden AI budget is material — often exceeding what the company planned to spend on its first intentional AI initiative.

The good news: this audit takes two hours with the IT team, not two months with a consultant. The five questions above surface the current state. The decision framework translates findings into action. And the findings feed directly into three companion tools — the shadow AI discovery worksheet (for unauthorized tools), the TCO one-pager (for total cost modeling), and the acceptable use policy (for governance). If this audit raised questions specific to your organization, I would welcome the conversation — brandon@brandonsneider.com.

Sources

1. Zylo, 2026 SaaS Management Index (40M+ licenses analyzed across enterprise customers, 2026). Independent SaaS management platform. Reports 54% license utilization, $19.8M average annual waste, 77% of IT leaders discovering unknown AI features. Credibility: Independent platform data, large sample, consistent methodology year-over-year. https://zylo.com/reports/2026-saas-management-index/

2. Microsoft, Q2 FY2026 Earnings (February 2026). Reports 15 million paid Copilot seats against 450 million total M365 commercial users (3.3% penetration). 160% year-over-year seat growth, 10x daily active user growth. Credibility: Publicly traded company, SEC-filed earnings data. https://www.theregister.com/2026/02/02/microsoft_ai_spend_copilot/

3. Microsoft, “Advancing Microsoft 365: New capabilities and pricing update” (December 2025). Announces E3 increase from $36 to $39/user/month and E5 from $57 to $60/user/month effective July 1, 2026, bundling Copilot Chat into base plans. Credibility: Primary source, vendor announcement. https://www.microsoft.com/en-us/microsoft-365/blog/2025/12/04/advancing-microsoft-365-new-capabilities-and-pricing-update/

4. Google, “The future of AI-powered work for every business” (January 2025). Announces Gemini bundled into all Workspace Business and Enterprise plans at no separate add-on cost, with 17-22% base price increase effective March 2025. Credibility: Primary source, vendor announcement. https://workspace.google.com/blog/product-announcements/empowering-businesses-with-AI

5. Salesforce, “Updated Product Packaging and Pricing” (2025). Agentforce add-on at $125/user/month for Enterprise and Unlimited editions. Agentforce 1 Edition at $550/user/month all-inclusive. Credibility: Primary source, vendor announcement. https://www.salesforce.com/news/stories/pricing-update-2025/

6. Zoom, AI Companion 3.0 launch (December 2025). AI Companion included with all paid plans. Custom AI Companion add-on at $12/user/month. Standalone tier at $10/user/month for free-tier users. Credibility: Primary source, vendor announcement. https://www.zoom.com/en/blog/zoom-ai-companion-3-0-agentic-ai-conversation-to-completion/

7. Stackmatix, “Microsoft Copilot Adoption Statistics & Trends (2026)” (2026). Reports 35.8% active usage rate among licensed Copilot users. Credibility: Industry analysis, methodology not fully disclosed — treat as directional. https://www.stackmatix.com/blog/copilot-market-adoption-trends

8. Oliv.ai, “Salesforce Einstein Features: What Works, What Doesn’t” (2025). Reports 67% of Einstein implementations face significant adoption challenges. Credibility: Vendor analysis (Oliv.ai sells competing product) — flag potential bias, but directional finding consistent with independent reports. https://www.oliv.ai/blog/salesforce-einstein-features

9. Gartner, “40% of Enterprise Apps Will Feature Task-Specific AI Agents by 2026” (August 2025). Predicts rise from <5% to 40% of enterprise applications embedding AI agents. Also warns >40% of agentic AI projects will be canceled by 2027. Credibility: Independent analyst firm, established methodology. https://www.gartner.com/en/newsroom/press-releases/2025-08-26-gartner-predicts-40-percent-of-enterprise-apps-will-feature-task-specific-ai-agents-by-2026-up-from-less-than-5-percent-in-2025

10. ServiceNow Community, “Now Assist License Pricing” (2025-2026). Pro Plus add-on reported at 30-45% premium on base user price. Consumption-based AI pricing with assist pack overage model. Credibility: Community-sourced pricing data — exact numbers vary by contract; directional only. https://www.servicenow.com/community/now-assist-for-creator-forum/now-assist-license-pricing/td-p/3147466

Brandon Sneider | brandon@brandonsneider.com March 2026