Focused Chaos
AI business models · Updated regularly

AI Business Models.

Lean Analytics described six business models. AI products don't fit them neatly, because every user costs compute and pricing has moved from seats toward usage, output and outcomes. Each tab describes a distinct AI business model, with its lifecycle, loops, economics and the metrics that matter. Real companies usually combine several.

Last updated October 18, 2026

Model 5

AI add-on to SaaS

The customers are already there. Getting them to pay extra is the hard part.

In this model, an established software company sells AI to the customers it already has, as a paid add-on, a higher tier or a credit meter. The distribution is the advantage: there's no need to find new customers, because AI shows up inside a product they already use. Microsoft 365 Copilot shows both the scale and the challenge. It passed 30 million paid seats by July 20262, yet in January 2026 its 15 million seats were only about 3% of Microsoft 365's more than 450 million commercial seats1,3. Customers try AI easily. Getting them to pay extra for it is much harder, so many vendors end up bundling AI and raising prices for everyone.

The installed base is the funnel

Unlike the other models, this one starts with customers who already pay. The map's acquisition row is about selling AI to them: prompts inside the product, a free taste of AI, account managers, launch events and renewal conversations. The main path runs from noticing AI to trying it, building a habit and finally getting someone to approve the spend. That last step matters because the buyer is usually an admin or a budget owner, not the person using the feature.

Pilots have to prove value before seats expand

Enterprise AI add-ons are often bought for a small group first. Expansion comes when that group proves the value, which is the seat expansion loop on the map. Proving it has been slow. In a 2025 survey of 187 IT leaders by the research firm Gartner, presented by Gartner analyst Dan Wilson, nearly half of the companies that had tried Microsoft 365 Copilot rated it "some value, shows promise." Only 5% of those that finished a pilot were moving to a larger deployment4.

Buyers expect AI to be included

Customers increasingly see AI as part of the product, not an extra. In interviews with 350 software buyers, G2, the software review site, found that 81% expect core AI in the base product and only 6% favor selling it as a separate add-on. Roughly one in four called a 20% AI premium a dealbreaker5. The consulting firm SBI reports that 58% of software vendors already include AI in their core packages at no extra charge, though it hasn't published how it measured that6.

Adoption is rising, but it won't fix the add-on

Use of AI at work is climbing. Microsoft 365 Copilot went from 15 million paid seats in January 2026 to more than 30 million by July1,2. A Gallup survey of more than 22,000 US workers in late 2025 found that nearly half use AI at least a few times a year, about a quarter use it a few times a week, and 12% use it daily, up from 21% who used it at all in 2023. Gallup also saw signs that growth may be starting to level off11. Even so, more usage doesn't mean more customers will pay for a separate add-on, since buyers expect AI to be included. Our view, an inference rather than a finding, is that the likely end state is AI bundled into every plan with a price increase, plus credits or a meter for heavy use.

So vendors bundle AI and raise prices

When few customers pay for an add-on, the common answer is to put AI in every plan and charge more for the plan. That's the bundling loop. Google dropped its separate Gemini add-on in January 2025 and put Gemini into every Workspace business plan, at a higher price. Before, a Business Standard customer paid $12 per user per month for Workspace, plus $20 for Gemini. After, the plan cost $14 with Gemini included. That was a big cut for the few who had bought Gemini, and a $2 increase for everyone who hadn't7. Notion retired its standalone $10 AI add-on in May 2025 and made Business, at $20 per seat, its AI tier8. Slack stopped selling its AI add-on that summer and moved AI features into its plans9. Canva raised some Teams prices by up to 300%, citing new AI features10. Salesforce raised prices 6% on its main Enterprise and Unlimited editions in August 2025, retired its standalone AI add-ons and moved customers to Agentforce bundles priced from $125 to $550 per user per month, with Flex Credits for usage12. Microsoft went the other way, bundling Copilot into a new $99 E7 suite, where E5 at $60 plus Copilot at $30 would cost about $90 separately3.

AI costs drag on a high-margin business

In classic software, serving one more user costs the vendor almost nothing. AI features cost compute every time they're used, including when they're bundled in for free. SBI says companies that absorb those costs miss revenue targets at twice the rate of companies that charge for AI separately6. That's why the third option on the map is a meter. Notion added credits for its AI agents in August 20258, and Microsoft CEO Satya Nadella has suggested pricing that combines subscriptions with usage limits and a meter3.

How it flows

  1. Existing customers see AI through in-product prompts, a free taste, account managers, launch events and renewal conversations.
  2. Some try the AI features. Many ignore them.
  3. Users who get value build a habit. Others stall in a pilot with no clear return4.
  4. An admin or budget owner approves the spend.
  5. The customer pays for AI as a per-seat add-on, a higher tier or metered credits.
  6. A pilot group that proves value leads to a wider rollout, which is the seat expansion loop.
  7. When few customers pay, the vendor bundles AI into every plan and raises prices7,8, which restarts the loop for every customer.
  8. Customers who drop AI usually keep the core product, and better AI features can bring them back.

Metrics that matter

MetricDefinitionWhy it matters here
AI trial rateShare of active customers who use an AI feature at least once in a month.The installed base is the funnel. Customers who never try AI can't build a case to pay for it.
Weekly AI usageShare of users with AI access who use it in a given week.Habit is the proof. A feature people try once won't survive a budget review.
Attach rateShare of customers, or seats, paying for AI.The headline number for this model. Microsoft 365 Copilot reached about 3% of commercial seats by January 20261,3.
Pilot-to-rollout rateShare of accounts that expand AI from a pilot group to a wider rollout.Enterprise AI is often bought in small pilots. In a 2025 Gartner survey, only 5% of companies that finished a Copilot pilot were moving to a larger deployment4.
AI seat penetrationAI seats as a share of all seats at accounts that bought AI.Measures how far the seat expansion loop has run inside each customer.
Gross margin with AIGross margin of the whole product after AI compute costs.AI adds real cost to a high-margin product, and bundled AI adds cost without adding a line item.
Retention, AI vs. non-AINet revenue retention of customers who pay for AI compared with those who don't.Shows whether AI is protecting the core product or just adding a line to the invoice.

Examples

Microsoft 365 CopilotA $30 per user per month add-on, now also bundled into the $99 E7 suite3. More than 30 million paid seats by July 20262.
Google Workspace with GeminiDropped the Gemini add-on in January 2025 and raised plan prices instead7.
NotionMoved from a $10 AI add-on to an AI tier in May 2025, then added credits for agents8.
SlackStopped selling its AI add-on in 2025 and spread AI features across its plans9.
SalesforceRetired standalone AI add-ons, raised core prices 6% and moved customers to Agentforce bundles with Flex Credits for usage12.
CanvaRaised some Teams prices by up to 300% in 2024, pointing to its AI tools10.

Sources

  1. Tony Redmond, “Microsoft 365 exceeds 450 million commercial paid seats,” Office 365 for IT Pros, January 30, 2026, on Microsoft's FY26 Q2 results.
  2. UC Today, “Microsoft 365 Copilot passes 30 million paid seats as cloud and AI growth power record quarter,” July 30, 2026.
  3. Mary Jo Foley, “M365 E7 to launch May 1 for $99 per user per month,” Directions on Microsoft, March 9, 2026.
  4. Ben Moore, “Gartner: Microsoft Copilot hype offset by ROI and readiness realities,” TechPartner News, June 26, 2025. Survey of 187 IT leaders.
  5. G2 Research, “AI Pricing After the Hype,” June 11, 2026. Based on 350 interviews with software buyers.
  6. SBI, “The State of SaaS Pricing Report 2026” preview. Sample size and method not published.
  7. Constellation Research, “Google Workspace drops Gemini add-on charge, raises business, enterprise plan prices,” January 15, 2025.
  8. Usage Pricing, Notion AI pricing blueprint (pricing history), 2026.
  9. Slack Help Center, “Updates to feature availability and pricing for Slack plans,” June 2025.
  10. Fortune, report on Canva's Teams price increases and its AI features, September 3, 2024.
  11. Matt O'Brien and Linley Sanders, “How Americans are using AI at work, according to a new Gallup poll,” Associated Press, January 25, 2026. Gallup survey of 22,368 US workers, fielded October 30 to November 13, 2025.
  12. Mastufa Ahmed, “Salesforce implements a 6% price increase while mandating AI bundle adoption,” CIO, June 18, 2025.
The PPTX is built from editable shapes. It also opens in Google Slides (File › Import slides).