​

Today, let's dive into the variety of monetization models available for AI products - whether SaaS or Otherwise.

1. Resource-Based Pricing ala ChatGPT/Claude

ChatGPT and its contenders today charge on a per token basis - why?

Because this is how they measure resources consumed.

The more tokens or chunks of text that have to be read or generated, the more resources are consumed.

The more expensive it is for the model.

So, you pay what you use for - power users pay more, happenstance users pay less.

Pro Tip: Do everyone a favour, don’t make your customers work to understand pricing. It’s a big turn off. AWS gets away with it, because they are mammoths occupying 31% of cloud infrastructure marketshare.

Potential downside: Profits will scale proportional to Customer usage, which means keeping your Customer base narrow is unlikely to be a profitable option without significant numbers of power users.

2. Add-on Pricing

Use this when you have an existing core product with recognized value, and now want to introduce AI features for efficiency.

Classic example is Notion’s model.

There is a core product offering with tiers (including a free one).

You can then add-on ‘AI’ to any of them.

If you’ve already got an existing audience, this can reduce friction to adoption. Additionally, since the add-on is a single $ value, the pricing is simple and might be easier to try for some Customers.

Potential downside: This type of flat add-on pricing means power users will drain resources from your company’s infrastructure quickly. Your pricing does not account for this. Therefore, you are assuming that you have enough low volume usage Customers to balance out the power users.

3. Speed-of-Processing Based Pricing

This model essentially combines a handful of key metrics that drive your product costs into your existing pricing model sans AI/ML.

For example - take a look at Midjourney’s model.

A monthly subscription model that adds restrictions to the tiers based on speed of inference - determined by GPU usage.

This is interesting because they’ve abstracted the concept of tokens away and replaced it with the idea of faster processing (i.e., access to parallelism on faster GPUs).

Pro Tip: Identify a value that your more premium Customers need and crave. Do not sacrifice quality of output though - that remains sacrosanct.

This value may be speed, quantity, variety (think tone of voice, image sizes etc.) and many others depending on your product.

4. Freemium and value-based tiers

Offer a basic version for free and get Customers hooked to your product’s value.

The goal is to make it part of their everyday workflow - a tool they can’t live without.

The paid tiers offer advantages like higher usage limits, newest models etc.

Technically, the Freemium model can be combined with any of the above models.

The idea of a Try-Before-You-Buy is not new, but simply a Customer acquisition channel.

​Notion, again has a great Free tier for the basics Its AI add-on has a few sample trial uses baked in.

Another example is Ideogram. It has a great free tier which is completely usable if your work requires lower volumes and you’re cool with waiting around for a few minutes.

PS - In case you couldn’t tell - I use both these tools extensively!

5. Outcome-Based Pricing

This model is super interesting - it signals that you truly believe in your product! If you don’t deliver the value you promise, you don’t get paid. It has a strong precedent in service businesses e.g., a sales consultant that takes a portion of a new sale they bring to you.

Pro Tip: Find a key value that your Customer is yearning for, build your pricing model around it and double down on making your product awesome at delivering this value.

Potential Downside: Likely obvious, but if your solution doesn’t deliver the promised value - your revenue will be hurting. You can minimize your downside by charging a fixed engagement fee (like in the example below).

For example, Lift AI - a buyer intent detector for your website prices based on meeting forecasted ‘lift’ to converting buyers.

6. Ethical Data Monetization

You could consider reselling data usage from Customers, post anonymization to interested parties.

Understandably, there’s an associated ick factor for the Customer, so

Pro Tip: Lean on Customer trust and be transparent about why you do it and what you sell.

Don’t forget to abide by regulations - GDPR, CCPA.

Make Customer consent obvious, and not buried in size 5 font in a 10 page T&C.

7. Partnerships and Licensing

License your underlying product’s capabilities to larger companies.

This can provide a significant revenue boost and open doors to new markets.

Pro Tip: Negotiate to retain your branding post licensing so that your brand is recognized should you choose to B2C or B2B later.

Licensing is insanely hot for media content right now.

As an example - Elf Labs​

Final Thoughts

Remember, the key is to align your monetization strategy with the value you provide. Don’t be afraid to experiment and iterate. Your pricing model is as crucial as your product – treat it with the same level of innovation and care.

​