AI. In search of value, in search of price
Now that we are on our way to spending billions of dollars on AI, the question of making at least some of the money back is coming to the forefront.
Most recently, Microsoft announced the launch of Copilot for business with free AI chat and pay-as-you-go agents. Originally, Microsoft announced that the new Copilot features will be priced at $30/user/month.
As a side note, Microsoft is trying really hard to confuse its customer base by naming its products and, shortly after, renaming them (Bing Chat and Bing Chat Enterprise will now simply become Copilot). But that's old news and every big software company is doing that. IBM would be another glorious example of that.
It would appear that at $30/user/month this new amazing thing is not flying off the shelves as hyped, hoped and expected. Microsoft came up with another idea: how to allow people to try it without committing to a large spend. Pay-as-you-go is another popular business model where, as the term suggests, you pay only for services you use. In this case ... I am truly lost. In this announcement Microsoft goes into great detail about the pricing, but it is only obvious to people who put that together how it really works.
After the mandatory preamble that describes the wonders of Copilot Chat and the future ability to run agents within the Copilot Chat (still no idea what it means), we learn that agents are the future. And it is the agent we are going to pay to use. The unit of usage with these agents is measured in 'messages.' That's how your company will pay for it. But wait, there is more to it (I think these people watched too many Ginsu knives commercials). It will depend on the type of request you send to the agent. And that will determine how many 'messages' you use/consume.
Here is the table.
Web grounded answers = 0 messages
Classic answers = 1 message
Generative answers = 2 messages
Tenant Graph grounding for messages = 30 messages
Autonomous actions = 25 messages
What!?!?!
Web grounded answers are obvious. #irony And don't tell me that you don't understand the obvious difference between Classic and Generative answers!
For the less gifted, Microsoft prepared the explanation:
'Classic answers, used for predefined responses that are manually authored by agent makers. These are static, do not change unless manually updated, and are typically used when precise responses are required…'
'Generative answers, used for dynamically generated responses based on knowledge sources and context. These provide more flexible and natural interactions because they build on a conversation’s context and available knowledge…'
Is the 'Classic answer' a new name for the hopeless search in SharePoint? And is the 'Web grounded answer' just MS Bing hiding in the witness protection program?
And I forgot to mention that one message is $0.01.
This brings us to the main question. Is $0.01/message a lot or too little? I am fairly positive that Microsoft doesn't know either. Some of the pricing is based on the underlying cost for running the infrastructure and the computing cost to answer the question, but other than that, it is just a guess.
Microsoft is not the only company which has no clue how to price 'AI'. Recent confession from Mr. Altman, the CEO of OpenAI where he discussed pricing for the top tier of ChatGPT, the ChatGPT Pro. The Pro version goes for $200/month and during the interview, Mr. Altman shared that a) OpenAI is losing money on it, because he didn't expect to be that popular and b) he said and I quote 'no, i personally chose the price and thought we would make some money.'
And that's coming from the CEO of a company valued at $157 billion. That's a billion with a B. One would think that a man whose company is building the most sophisticated AI that the universe ever saw would just ask the internal ChatGPT Super Secret Ultra Pro version a simple question: 'How much should I charge for you?' And with that answer conquer the world.
I guess AI still doesn't know enough about its own value.
The recurrent pattern in all this. The above examples show how early we are in this and how we still don't know what/where/how the technology will be used. Despite all the marketing hype and product announcements, the tech companies are waiting for the business to find out where we can make money with this new hammer.