You’re likely wrestling with how to price an AI feature (or service) in your product or maybe wondering whether you should shift to outcome-based pricing. We’ve thought about this a lot, worked with clients, and tested models. Let’s walk through both approaches, share stories, and show trade-offs. Then you can decide what makes sense for your business.

What these two models are (in simple terms)

AI Add-On Pricing

This is when you have a SaaS product, and you add AI features as an optional extra. Think of it as a side dish you can order in addition to your main course. 

• You have your base plan (or plans) 

• You build AI capabilities — maybe automation, smart suggestions, predictions, agents, whatever 

• Then you price that capability as an add-on: extra cost per user, per usage, per seat, or per feature. 

• If someone wants the AI, they pay more; if not, they stick with the old plan.

Examples: 

• “AI-advanced” module costs +$50/user/month 

• More storage or more “smart searches” as an extra 

• Basic plan + “AI agent” add-on

Outcome-Based Pricing

Here, you tie the price to the outcome. Instead of paying because a user exists, or because a feature is “on,” the customer pays when you deliver specific results. You both share risk: we deliver value → you pay.

• Success metrics must be clearly defined (resolved tickets, leads generated, cost savings, etc.) 

• If the outcome doesn’t happen, either you don’t charge (or charge less), or there's some fallback. 

• It aligns you and your customer: their success = your success.

Examples:

• Pay per resolved support ticket via AI agent. Intercom’s “Fin” is a famous example: about ~$0.99 per resolved ticket.  

• Fraud prevention tool charging only for successfully prevented fraud or approved identities.

Why outcomes matter more nowadays (and what’s pushing the shift)

Because of AI. Because AI features tend to:

1. Have variable performance — sometimes the AI does great, sometimes less so. 

2. Increase expectations — customers expect that if they're paying extra for AI, it should do something measurable. 

3. Make “seat-based” or feature-based models somewhat lopsided: you could have many users barely using the AI, but still paying; or intense usage where infrastructure costs balloon.

Also, market signals:

• Companies that do outcome-based pricing tend to see more aligned incentives, better retention. When your customer sees you sweating results, trust goes up. 

• Analysts predict outcome-based models will grow in SaaS, especially in AI-first tools.

Pros & Cons: What you gain and what you risk

To help you decide whether to stick with (or adopt) AI add-on vs move toward outcomes, here are the trade-offs.

Hybrid: Mixing both so you get the best of each

Most companies we talk to do hybrid models, because pure outcome-based or pure add-on has downsides. Some ideas:

• A base subscription + smaller outcome-based component 

• Guarantee minimal performance, with bonus payments if outcome thresholds exceed. 

• Tiered models: low cost for add-on, but outcome-based pricing for big customers or for customers who want more guarantees.

What to check BEFORE you decide to go outcome-based

Because outcome-based is seductive, but risky if done poorly. Ask yourself:

1. Is the outcome clearly measurable AND under your control (or jointly under your control)? 
   If it's something heavily influenced by the customer’s input (poor data, no usage), you’ll always be battling excuses.

2. Can you monitor and report reliably? 
   Transparent metrics: customers must trust your numbers. If you say, “resolved ticket,” defines resolution clearly. Time windows, re-opens, etc.

3. Do you understand the costs and risks? 
   If the AI model sometimes fails, or usage surges, your cost of supporting/maintaining/fixing could be eaten into margins. Run the worst-case scenarios.

4. Are customers comfortable with risk? 
   Some customers want predictability: “I pay X/month, I know my budget.” Outcome-based might seem like a wildcard. You might need to be educated.

5. How competitive is your landscape? Will your customers see this as added value (not added complexity)?

Our View: When to Lean One Way or the Other

Since you’re running a SaaS or building one, here’s our take on when which model tends to work: 

• If your AI feature is support / automation / agent type (support ticket resolution, chatbots), outcome-based often shines. Users care about solved problems. 

• If AI is more exploratory, or experimental, or high variance (e.g. content generation, “creative” stuff), an add-on plus usage-based metrics works better initially. 

• For enterprise clients, you can negotiate outcome-based deals more easily because they often have big budgets and care about ROI. For SMBs, simpler, predictable pricing tends to lower friction. 

• If your product’s infrastructure cost for AI (computer, data, inference) is substantial, you may need a hybrid: base fee to cover fixed costs + outcome-based for the rest.

Stories & Real Examples

• Intercom’s Fin: Their AI agent charges about $0.99 per resolved ticket — no resolution, no charge. That forced Intercom to focus intensely on improving resolution rates.  

• Zendesk has layered AI add-ons and also outcome-based components (automated resolution counts, etc.). But their users complained about unpredictability: when usage spikes, costs go up in surprising ways. 

• Also, there are cases where outcome-based pricing flopped because the metric wasn’t clear, or the customer’s side of the work (data quality, behavior) wasn't considered. Risky if you don’t define things well. (We’ve seen this happen in our advisory work.)

How We’d Approach It

Here’s what we’d do if we were adding an AI feature to our SaaS product:

1. Start with an add-on (because that’s fastest to get to the market). Gauge usage, see what customers value. 

2. Track outcomes: how many users use it, how often, where value shows up (time saved, tickets solved, leads improved). 

3. Once we have enough data, build a pilot outcome-based pricing version for selecting customers (maybe enterprise or loyal ones). Use that to refine metrics. 

4. Use a hybrid model: basic subscription + option to pay for better outcomes. Maybe even “performance bonus” fees when certain thresholds are exceeded. 

5. Be super transparent with customers: explain what “success” means, what we will measure, what happens if outcome is not achieved.

Final Thoughts

If you’re building a SaaS product (or already have one with AI features), outcome-based pricing is powerful. It aligns incentives, reduces friction, and can help you stand out. But it’s not a magic wand — it demands clarity, reliability, and sometimes more upfront effort. 

AI add-ons are safer, more predictable, and easier to sell initially. They let you test the water. But over time, customers will expect more — more value, more clarity. Outcome-based models are the path forward in many cases. 

So, if I were you: don’t force outcome-based from day one unless you’re confident in the metrics. But build toward it. And always listen to your customer: what they perceive as value, what they’re willing to pay for.