Going by the reports from Gartner, by 2026, nearly 80% of enterprises would have used generative AI APIs or deployed generative AI-enabled applications.
While this is good news for founders of AI companies, turning AI hype into profits can be tricky.
Alexandru Costin, VP of Generative AI at Adobe echoes this thought. He says there are too many levers to consider in pricing AI products — “finding the efficient frontier between quality and interactivity” is crucial.
Pricing right initially is important to avoid disruptions caused by frequent price adjustments.
This is why we want to make sure you are on the right track. We did the groundwork for you and we’ve got some interesting insights on how founders of AI companies are pricing their offerings, what’s working for them (what’s not), how all this ties up with your billing system, and more. First up, the key things to consider in AI Pricing.
How to price AI?
"The single most important decision in evaluating a business is pricing power. If you've got the power to raise prices without losing business to a competitor, you've got a very good business. And if you have to have a prayer session before raising the price by a tenth of a cent, then you've got a terrible business.”
- Warren Buffett
Pricing can either drive or destroy your company’s operating margins.
You can’t go wrong with pricing based on value
Pricing strategy is the approach or model used to set the right price for your product or service. It is the primary driver of profitability for most businesses.
While there are many approaches to setting prices, a more thoughtful, value-based approach for your AI product can pay big dividends.
According to Jake Saper, companies find their true north in tying value creation to pricing structures.
“The goal here is to tie your pricing mechanism to value creation. That is where value is aligned. You did add value, therefore you should get paid. But you don't want to disincentive usage.”
- Jake Saper, Emergence Capital
When you price your products based on the value your customer gets, your customers’ needs take center stage.
So, how do you go about it?
The first step in value-based pricing involves determining how customers perceive the value offered by a product or service. The challenging part here is — different customers perceive value differently.
We’ll understand this better with an example.
Take for instance a secure data warehouse company. This company caters to various clients, providing customized data warehouse solutions, including the option to host data in a specific country based on individual client requirements. In this scenario, for a typical SaaS company, where geographical constraints matter less, the value derived is only for storing their data in a warehouse. However, for a healthcare company, the value they derive is much higher because not only can they store data in a warehouse, but it is also securely stored within regulatory geographic constraints. This added value is significantly higher for the healthcare company than the SaaS company.
We have now identified two segments of customers, where the value received for both are different.
Once you segment your customers, you can price the products differently for these customers.
Going back to our example, the cost of data warehouse services for the two segments can be considered as seen below.
These are fictional numbers for illustration purposes.
To break it down:
Y could be as high as 2-5X the price you charge customers in segment 1, although your cost of offering customers services in a specific location (segment 2) is just 10% higher than offering standard data warehouse services. Consequently, by segmenting customer value and pricing differently, you have now successfully boosted your profit margins (this of course doesn’t cover the more important Return on Capital Employed, but that’s a topic for another day)
Pricing based on value for AI companies
With AI, some products are very niche where you know exactly the value that you provide your customers. This makes pricing based on value easy.
Consider the case of a freelance blogger who utilizes an AI tool to generate articles.
Now, let's say that the freelance blogger can produce three times as many articles as usual, all while maintaining or even improving the quality with the help of the AI tool. Owing to the significant value that the AI tool is adding, it’s not a stretch for us to assume that the freelance blogger might be willing to pay 20% of the amount earned on each article to the AI tool. The cost for the AI company would be typically based on tokens which could be significantly less than 20% of a writer’s revenue. For instance, ChatGPT prices API usage based on tokens. The cost of an API call is the sum of the tokens used in the API call (referred to as prompts) and the resulting output generated by ChatGPT (referred to as completions).
You can maximize the value capture by aligning pricing with crucial metrics that resonate with your customers.
In the example above, for the AI tool, the cost is based on tokens — but the value that the freelancer derives is on the articles written. In this case, it could make sense for the AI company to price based on each article instead of per token.
The initial and most important step for a company is to identify the various ways in which your customer received value from your product and then segment your customers based on the way that they derive value. AI companies can also benefit from augmenting their product with specific features for each segment, to justify the price difference. For instance, to a customer, token-based pricing looks cheaper and they will end up opting for that pricing model. However, adding a feature such as a Google Doc integration or a Grammar Checker will help the AI tool get more customers to opt for a per-article-based pricing model.
This helps in two ways:
- For the customer, understanding pricing based on articles written is easier compared to comprehending pricing based on tokens for each article.
- The AI tool has the potential to charge more. The per token pricing is lesser, compared to earning 20% of an article that, let’s say, is charged at $500 by the freelancer.
By pricing closer to customer value, the AI tool is able to increase revenue significantly. Moreover, if pricing based on value is done right, you can also overcome cost and competition constraints.
That seemed easy, so what’s the challenge?
A majority of AI companies today lack insights into how customers derive value from their products. For an AI tool that caters to a large number of customers, segmenting them based on the value derived can be a challenge. Also, we’re still at a nascent stage of the AI boom. Every day, new ways of deriving value from AI are invented. All of which makes it very hard to know how people will derive value from AI.
Against this backdrop, where customer value is hard to determine, pricing that is closer to cost than value (like tokens) might be the most convenient model for a company like ChatGPT.
At times, companies fail to recognize the full value of their AI product and they end up underpricing — missing out on potential revenue. Take the case of Microsoft. They introduced GitHub Copilot, an AI tool designed to assist programmers in generating, correcting, and translating code. They are reportedly experiencing an average monthly deficit of $20 per user for the GitHub Copilot offering.
When judging how customers extract value from your product is not easy, you can opt to price based on the potential costs. You can also consider the pricing strategies of competitors and other similar products in the market. Adding more functionalities to your AI offering will help you stand out from the competition and enhance the value you provide.
While cost and competition-based pricing may get you off to a good start, in the long run, pricing based on value is the winning formula.
When you make the mistake of not pricing your AI products based on value, you most certainly end up leaving money on the table. If you are unsure of how customers are getting value, you need to get started on identifying this today.
- Find the customer value.
- Identify segments of customers based on the value they derive.
- Capture the value that you provide through your pricing.
A pricing strategy consulting firm, Ibbaka, breaks this down by sharing the basic pricing components that you need to take into account. These components are based on the role played by your product in the AI ecology.
- Value Driver: An equation quantifying one aspect of how a solution delivers value to a specific customer or narrowly defined customer segment.
- Value Metric: The unit of consumption by which a user gets value.
- Pricing Metric: The unit of consumption for which a buyer pays.
Take a look at the different approaches to pricing seen in different parts of the AI ecology, here.
Real-world examples of pricing AI products
Deciding on a pricing plan for AI requires a lot of research, which is why we want to make this easier for you. We've taken real-world AI pricing examples and have dissected various pricing models for you.
- Intercom has done a great job in mapping pricing to value. Their AI chatbot, Fin only charges clients $0.99 for resolved conversations. This means that, regardless of the number of answers provided, you are only charged when a customer either confirms the answer provided is helpful or leaves the conversation without seeking additional help.
- IBM watsonx Assistant is a conversational AI platform powered by Large Language Models (LLMs) and designed to overcome the hassles of traditional customer support. The platform empowers organizations to build AI-powered voice agents and chatbots that deliver automated self-service support across multiple touch-points. Their pricing starts at $140 and includes 1,000 monthly active users. As a value-add to the user, the pricing page showcases the costs it saves organizations that use traditional customer support channels.
- Midjourney is an AI tool that shot to popularity with the rise of the generative AI wave. This AI tool converts natural language prompts into images. The basic plan is priced at $10 per month and allows approximately 200 limited generations in a month.
- Grammarly, the AI-powered writing assistant, initially experimented with a freemium model to quickly onboard users. As the product matured, they experimented with tiered subscription plans, showcasing a flexible approach to pricing that evolved based on user needs and the product's growing capabilities. They currently charge $15.00 per member every month.
- Amazon Lex is the company’s AI chatbot service with advanced natural language models to design, build, test, and deploy conversational interfaces into any application using text and voice. Amazon Lex’s pricing is based on the number of speech or text API requests processed by your bot. The pricing is currently at $0.004 per speech request and $0.00075 per text request.
- Notion AI strategically has presented AI functionalities as an optional add-on to any paid plan. It is priced at $8 per member per month for annual billing and $10 per member per month for monthly billing. To effectively manage inference costs within the framework of its existing seat-based pricing model, Notion AI has implemented a “fair usage limit”. In instances where a user surpasses 30 AI requests within a 24-hour window, the user will encounter slower responses for that specific period. This measured approach ensures a balance between accessibility and cost control for Notion AI users.
- Clarifai has demonstrated agility in adapting its pricing strategy to remain competitive. The company has made nuanced adjustments in response to the pricing approaches of larger competitors and has scalable pricing plans to offer. While they allow the user to get started with 1,000 free operations in a month, the Essential plan is offered at a monthly credit of $30 worth of inputs and operations.
- Writersonic is an AI tool that creates SEO-friendly content for blogs, Facebook ads, Google ads, and Shopify. Their pricing is based on the number of words used and starts at $13 per month for 200,000 words.
- Salesforce's Einstein Predictions is priced based on the value it brings to sales and marketing teams through advanced analytics, machine learning, and AI-driven insights. Einstein predictions are priced at $75 per user per month when billed annually.
- Claude is an AI assistant for enterprises. It performs various tasks, from writing content to automating some of the more basic activities like document formatting. The pricing is based on tokens. The Clause Instant plan charges $0.80 per million tokens for prompts and $2.40 per million tokens for completion.
- Hugging Face is a platform that helps the machine learning community build and train ML models. The platform provides the infrastructure to demo, run, and deploy AI in live applications. While they have a free plan, the Pro Account offers additional features at $9 per month.
- Google’s Document AI is their solution to parse, split, and analyze unstructured data. They organize document data that can be easily stored, searched, and utilized to automate processes. The pricing is based on the number of pages, where the pricing for Enterprise Document OCR Processor is $1.50 per 1,000 pages for 1 - 5,000,000 pages per month.
- Elicit is an AI tool that helps users automate time-consuming research-based tasks. The tool analyzes research papers, summarizing them, extracting data, and synthesizing the users’ findings. The pricing is based on the papers summarized. The paid plan starts at $10 per month billed annually and the user can summarize 8 papers under this plan.
- Synthesia is an AI video platform that is used to create professional videos using text. The AI video generator lets you create videos without mics, cameras, actors, or studios. The starter pack is priced at $22 per month and can be used to create 120 minutes of video in a year.
- Cohere is another AI-powered text generation tool. The tool can be used to produce content for emails, landing pages, and product descriptions. While the user can get started for free, the paid plans are priced on the basis of tokens, starting at $0.30 for an input of 1 million tokens and $0.60 for an output of 1 million tokens.
AI pricing model analysis
As companies shape their AI pricing strategies, factors such as finding the balance between usage flexibility and subscription predictability as well as selecting the right meters (API requests, tokens, etc.) further contribute to the complexity.
Based on the above real-world examples, let’s quickly look at the preferred pricing models and commonly utilized usage meters.
Out of the 15 companies, 10 (66.67%) have usage-based pricing models.
Usage-based models offer pricing flexibility but call for more evolved infrastructure needs when it comes to billing capabilities, primarily in handling usage data efficiently. Subscription models provide predictability but may not suit all user behaviors. The choice depends on the nature of the AI product and its users. Various metrics like API calls, tokens, data volume, number of documents processed, transactions, etc., are examples of meters in usage-based pricing models. AI companies must carefully define and communicate these metrics to customers. The goal is to clearly communicate pricing structures, justify the value delivered, and proactively inform customers about any changes.
The role of flexible billing for AI companies
AI companies require a robust billing system that can work seamlessly with usage data, adapt to evolving usage metrics, and accommodate changes in pricing strategies.
Ability to iterate on usage meter
+ Pro Tip: “If pricing is tied to a usage meter that users don't see value from, users won't pay and they will churn.”
The ability to iterate on usage meters is particularly relevant to AI companies due to the nature of their usage-based deep tech products that provide tangible value to the end user (for example - images, customer support, summarization, search, etc.). As costs often scale with usage, the billing system becomes a critical tool for experimenting with pricing based on usage meters that are closer to customer value, such as content length, number of articles, and number of support requests, etc.
Flexibility to experiment with pricing models
The billing tool's ability to work with usage data should empower the sales teams to experiment freely with different pricing levers. The enhanced level of flexibility enables AI companies to iterate on their pricing models and adapt to changing market dynamics. This includes structuring pricing models and terms to have more predictability in cash flow and capital efficiency, which can be achieved by experimenting with various methods of frontloading cash inflow like charging upfront for prepaid usage, having subscription plans with included usage and overage in arrears, and using provisioned balances (like credits) to track usage.
Over to you
Strike the right balance between pricing products effectively and ensuring long-term success in the market by continually refining your pricing approach and exploring ways to deliver unique value to the customer.
There’s no denying that AI companies require a flexible and comprehensive billing system to navigate the complexities of pricing.
Your search for an AI billing tool ends now. Zenskar empowers AI companies to optimize pricing models, seamlessly manage usage data, and automate billing processes. By decoupling metering from pricing, you can easily incorporate raw events or product usage into the billing tool, freeing your engineers from being dependent on specific customer contract details. Zenskar employs robust ways to pull data from the client’s infrastructure, by plugging into their data warehouses and fetching the data from their data store, or by having clients send data using APIs (or even upload CSV files to get started with). We simplify usage data aggregation using SQL and enable our clients and end users to track usage aggregates in real-time. While other billing tools store prices linearly as rows in a table, Zenskar stores prices using a graphical data model — effortlessly handling the nuanced terms of contracts within the dynamic AI industry. With Zenskar, you can configure complex contracts without writing code by just using a drag-and-drop builder. It doesn’t get better than this — if you can describe your contracts in simple English, we can encode them to automate your billing.
Whether dealing with sophisticated product details or usage-based agreements, Zenskar revolutionizes billing practices to help AI companies price products based on the value provided.
Book a demo to revolutionize your billing operations and take your AI business to the next level.