Pricing and generative AI
Steven Forth is a Managing Partner at Ibbaka. See his Skill Profile on Ibbaka Talio.
Copywriting and text generation is one of the early use cases for generative AI. There have been a number of product announcements in this area in the late spring of 2023. A lot of money is being invested. How will the value being created get captured so that these innovations are sustainable.
Could I write this post using generative AI? Or at least take a first pass. This blog is hosted on Squarespace, and the following icon now shows up in the toolbar.
Prompt: How are large software companies using generative AI and large language models?
”Major software vendors are eagerly adopting Generative AI and Large Language Models to enhance and streamline their products and services. By leveraging Generative AI, these vendors are able to create advanced systems that can generate content, designs, and code, automating repetitive tasks and saving valuable time for developers. They harness the power of Large Language Models to improve natural language processing capabilities, allowing users to interact more seamlessly with software through voice commands or written requests. Moreover, software vendors are utilizing these technologies to strengthen their recommendation engines, enabling them to provide personalized suggestions for users based on their preferences and behavior patterns. Ultimately, the use of Generative AI and Large Language Models empowers major software vendors to unlock new levels of productivity, efficiency, and user experience within their offerings.”
OK, that was a bit naive. I know that GPT means Generative Pretrained Transformer. Generative in that it is able to generate content (as it did above), pretrained in that it has to ingest the content and then build a model. The specific AI architecture is that of a transformer, the standard technology for LLMs. There will always be some lag between the training and availability. The above is fine as far as it goes, but it is a bit too generic for my purposes.
People following the generative AI space (which is a great many, I suspect millions) are likely aware that in the past few weeks, in addition to Squarespace putting generative AI directly into its content authoring tools,
Hubspot launched ChatSpot
Zendesk AI came out with an AI that automates customer support
Salesforce made available AI Cloud.
Adobe updated Firefly with AI and introduced Adobe Sensei GenAI
This is just the tip of the iceberg. Hundreds of generative AI companies are getting funded (see CB Insights the AI 100) and many other enterprise software companies are adding generative AI to their existing applications or are building new applications with this technology.
Generative AI is not free, or even low cost. Open.ai’s GPT-4 has about one trillion parameters. It takes serious money to develop a Large Language Model like GPT-4 and we can expect to pay to use it. We described the Open.ai’s published API pricing in the post How Open.ai prices inputs and outputs for GPT-4.
Most LLM companies charge by some combination of input and output tokens (a token being a word or part of a compound word). The little snippet above would have cost Squarespace only a tiny amount to generate. Input was about 20 tokens and output about 230 tokens. If they are using Open.ai and paying the standard price of 3 cents per 1,000 for input tokens and 6 cents per thousand for output tokens the cost would only be about US$0.0144. Not much but with massive adoption, it would add up.
Of course there are already cheaper alternatives to GPT-4. Cohere.com charges only $0.000015 per token. They charge equally for inputs and outputs so at this pricing Squarespace would be charged US$0.00375 for the prompt and output.
The cost of implementing generative AI is high enough that most B2B solutions will need to capture it one way or another.
How do you capture value? Let’s ask the Generative AI again.
Prompt: Write a paragraph about ‘How can B2B SaaS companies capture the value of innovations based on Generative AI?’
B2B SaaS companies have a significant opportunity to capture the value of innovations based on Generative AI. By leveraging the power of Generative AI, these companies can enhance their products and services, optimize internal processes, and stay ahead of the competition. Generative AI can enable B2B SaaS companies to create personalized and unique user experiences, allowing them to tailor their offerings to the specific needs of their clients. Additionally, by incorporating Generative AI into their product development cycle, these companies can accelerate the speed of innovation, generate novel ideas, and create cutting-edge solutions. By embracing the potential of Generative AI, B2B SaaS companies can unlock new revenue streams, improve customer satisfaction, and ultimately strengthen their position in the market.
Hmmm, let’s narrow this down to pricing.
Prompt: Write a paragraph about ‘How can B2B SaaS companies price the value of innovations from generative AI?’
B2B SaaS companies can effectively price the value of innovations from generative AI by focusing on the outcomes and benefits it brings to their customers' businesses. Firstly, they should conduct a thorough analysis of the specific pain points that their generative AI solution addresses, such as enhanced productivity, improved decision-making, or cost savings. By quantifying these benefits, either in terms of time saved, revenue generated, or resources conserved, companies can demonstrate the value of their innovation. Additionally, it is essential to consider the market's demand elasticity and competitive landscape to ensure the pricing is both competitive and aligned with the perceived value. Finally, adopting a pricing model that allows for flexibility, such as a tiered pricing structure or pay-per-use, can cater to customer needs while capturing the true value of the generative AI solution. With a customer-centric approach and a clear understanding of the impact generative AI brings, B2B SaaS companies can price their innovations effectively, ensuring a strong value proposition for their customers.
I may need some training with this. Fortunately, Squarespace is offering me some help in an article called “How to Write Website Copy with AI.”
I like the focus on ‘outcomes’ and the ‘customers business.’ These lead back to value-based pricing, where price is based on the impact on the customer’s business (and not on something like willingness to pay or the cost of delivering the solution).
8 Ways to Capture Value of Generative AI
Let me try. The ways to capture the value of a B2B SaaS innovation are by …
Growing the market by bringing in users who would not have used the application without the new functionality
Capturing market share by winning over customers using competing solutions
Keeping customers who might have otherwise left
Increasing use which if you have usage based pricing will increase revenue
Increasing the price of the existing service, in other words, as the justification for a price increase
Offering the functionality as an optional extension that users must pay for
Using the functionality as a fence that pushes people to buy a more expensive package in a Good Better Best style package design
Packaging the functionality as a new offer and taking a new product to market
Ibbaka is planning to layer generative AI into Valio and we are still considering all of these value capture options.
A Simple Framework for Generative AI Pricing Strategy
Choice of pricing metric for generative AI will depend on:
Overall Strategy
General Product Strategy
Product Strategy for the Specific Product
By working through this cascade, you can start to think through how you will approach pricing AI. There are predictable patterns here. Most companies will be conservative.
Establish a defensive positioning to buy time (I think this is what most of the large companies are actually doing)
Enhance existing value drivers for existing customers
Bundle AI with other functionality
That is what a lot of the announcements we are seeing amount to.
Some companies will go farther, Microsoft and its full throated adoption of OpenAI is an example.
Enhance current positioning
Create new value drivers for existing customers
Offer AI as a priced product enhancement
A few companies will be focussed on category creation. Many of these will be startups, but a few large companies will also play.
Create a new positioning
Create new value drivers for new customers
Offer AI as a separate product
Of course there are other ways to align general strategy with general product strategy and specific products, but I suspect these will be the three most common patterns.
One can then map the 8 Value Capture Tactics to the three strategies.
At this point, Microsoft, Hubspot, Adobe, Salesforce, Zendesk are taking the extension strategy. Category creation is happening more with start-ups. Let’s look at two content generation startups that have been getting some attention.
Two Generative AI Startups and their Pricing
Jasper and Copy.ai are early entrants into the rapidly emerging Brand and Copywriting category. Let’s take a look at their value propositions and pricing.
Jasper helps you create on-brand AI content and is not limited to text generation. Images are also included. The key differentiator here is that Jasper will speak in your brand’s voice (and implicitly, if you don’t have a brand voice, then the AI will develop one). You can set the tone: Cheeky, Formal, Bold, Pirate. More importantly, you can train the AI on your own content so that it will learn how to write about your brand.
Jasper pricing has two priced tiers and a business tier. It combines product-led growth (for Teams and Creator) and sales-led growth (for Business). Packages are fenced by the number of users, number of brands, knowledge assets, campaigns and integrations. Seems a bit complex to me.
Interestingly, there is no limit in the number of words generated by the AI. Contrast this with LLM model vendors like Open.ai and Cohere, which charge for the number of tokens input and output (we are working on post comparing Open.ai and Cohere’s pricing).
Jasper also makes value claims on its website.
It is hard to evaluate these claims without more data. The 3.5X return on investment is interesting (though low for a new innovation where 10X is more common) but one wants to know what factors are driving this ROI. A value management system like Ibbaka Valio can do this.
If one can trace the connection from content download to actions that generate revenue you could build a value driver from content downloads. It is this sort of value driver that is likely to drive Jasper’s adoption. Revenue value drivers are more powerful than cost value drivers in most use cases.
Copy.ai is also part of the emerging category of AI copywriting. There key value promises are
These are very similar to those for Jasper. One of the signs that a new category is cohering is that different vendors have similar value propositions. The difference here is that Copy.ai has mapped value propositions to personas: ‘Write blogs 10X faster’ for Blog Writers (I am skeptical of this claim) and ‘Write higher converting posts’ for Social Media Managers. If one can quantify the value of conversions this can be a compelling value driver.
Let’s look at the Copy.ai pricing page. The pricing metric is opaque and has to be constructed from the different fences. The fences are ‘words per month,’ ‘user seats,’ and various functionality points.
The Enterprise package seems almost like a different application with its focus on workflows, SOC-2 compliance (a lot of companies are concerned about the security implications of AI adoption). Note that ‘Custom brand voices’ comes in as a special feature of the enterprise package; for Jasper this was a core value proposition across all packages.
Copy.ai offers unlimited words for the Pro package. I wonder if this is also true for Enterprise. Given the cost of operating a LLM it may not be. I looked for a chatbot where I could ask this question, but did not see one. It is going to be interesting to see how ‘unlimited words’ gets reconciled with the pricing models of the LLM vendors. Something’s gotta give.
Should copywriters be worried?
One of the current concerns about Generative AI is that it will take away jobs. Visual Capitalist has a good infographic on this based on research from Open.ai GPTs are GPTs: The Labor Market Impact Potential of Large Language Models. Writers and authors are one of the groups with high exposure.
In their current state, Jasper and Copy.ai are meant to complement skilled humans and not to replace them. One would not want to rely on these systems to write finished copy in the brand’s voice. But they are likely to get there. And this will reduce the number of jobs for writers as writers. They may have new jobs as prompt engineers, but that is a different skill set with different emotional qualities.
Who is going to read all the content that these systems are generating. If a writer becomes 10X more efficient are they going to write 10X more content? I hope not. Who would read all that content? Other AIs I suppose.
Writing is one of those skilled trades where the best writers are orders of magnitude more impactful than the average writer. Perhaps AI will raise the quality of writing rather than lead to the generation of more writing. It would be nice to read more well written articles and I would like to write better myself.
From a business point of view, this sounds like a Red Queen game. A Red Queen game is one in which everyone in an industry improves, the improvement becomes table stakes. Not much changes when it comes to differentiation or positioning. The idea comes from Lewis Carroll’s Alice in Wonderland. ““My dear, here we must run as fast as we can, just to stay in place. And if you wish to go anywhere you must run twice as fast as that.”
Red Queen games are good for providers. We will all have to adopt these tools. But for the rest of us, it is going to be a lot of work to keep up!
LeadLander