What are the emerging value drivers for generative AI agents and how will they be priced?
Steven Forth is CEO of Ibbaka. See his Skill Profile on Ibbaka Talio.
Companies are struggling to get returns on their massive investments in AI. These investments have been led by the largest AI companies, Google, Microsoft, Open.ai, Meta, Anthropic, and so on, but VCs have also been active and the vast majority of B2B SaaS companies are investing in one form or another of AI functionality or platform.
How different companies are approaching monetization will be the subject of my talk at the Professional Pricing Society spring conference in Chicago on Friday, April 26. Pricing expert Mark Stiving and I will then work through this together in a webinar on May 23. Please sign up for the webinar. Even if you can’t attend you will get the recording and some additional resources.
Pricing of AI is an important theme, but in order to price one must be creating something of value. One of the standard plays is to create an ‘agent’ but what kind of agent and how to price?
The Information had an interesting thought piece on this last week.
To Unlock AI Spending, Microsoft, OpenAI and Google Prep ‘Agents’
As many businesses remain cautious about spending on conversational artificial intelligence, AI providers such as Microsoft, OpenAI, and Google are racing to make the technology more of a must-have—by introducing new features that can handle complex tasks with little guidance from the customer.
The Information organizes these into three types of agents.
Note that Adept is positioning itself as a general RPA (Robotic Platform Automation) or maybe IPA (not the beer, Intelligent Process Automation).
Computer-using agents - acts on the user’s computer and the data it contains
Multi-step application agents - acts on an application or set of applications as RPA agents (Robotic Process Application)
Web-based task agents - acts in the open web, across many data sources and applications
Note that this agent taxonomy is organized around where the agent acts and what data it acts on.
Does the type of agent impact how one would price and package?
Yes.
Computer-using agents - these agents will be seen as high risk as they need direct access to the user’s computer. This requires trust. Many companies will not have this trust and will not be able to sell this type of agent at any price. But if the user does trust the vendor, that trust will be rewarded by significant pricing power. These applications are likely to be priced per device and will often use tiered GBB (Good Better Best) style packaging.
Multi-step application agents - these agents will need to be easy to manipulate applications. Technically, there are several ways to do this but the most effective will require access to application APIs. In other words, they will provide a compelling use case for APIs, beyond basic integration, and be a value multiplier. There are likely to be two common pricing patterns.
As extensions to the existing pricing metric for the application (sadly, this is still per user in too many cases)
As a multiple on the API pricing, which is often a combination of endpoints, calls, and throughput
Web-based task agents - these agents are the most general and independent. They are not tied to devices or other applications. Vendors will have a wide range of pricing options open (let’s hope they do not gravitate to per-user pricing). Large vendors with complex suites (Microsoft, Google) will need to position these agents within their overall suite, and will likely be constrained by past decisions (at this point more of a problem for Microsoft than Google, but only because Microsoft is ahead of Google). Smaller more focussed companies will be able to focus on value-based metrics and the use of value-path-based pricing.
These agents may even force a change in the definition of ‘value path.’
Old definition: A series of actions taken by a user that ends with something of value.
Example: Planning a trip to Chicago for PPS and finding flight and hotel options.
Definition for AI agents: A series of actions that an agent takes for me that ends with something of value.
Example: Planning a trip to Chicago for PPS and purchasing flights and hotels; suggesting meetings and making appointments. In many cases, this will use a conversational interface, but generative AI is opening new ways to think about configuration and UX (User Experience).