Here Comes The Meter Man – A Lesson in GenAI Planning

last updated:
An image of a Smart Meter display

Here Comes The Meter Man – A Lesson in GenAI Planning

last updated:

Our expectation is that the future state for the use of GenAI tools in business applications is likely to be billing through forms of metered use, just as we see for existing API-based tools elsewhere.

For many of us in the UK, this week – the first properly chilly one since the summer – has been the week when the heating switches on. If like me, you’re on a Smart Meter, this means we can see our bank balance diminishing in real time, as the little gas flame icon which we have got a little bit used to being pretty much invisible on the display, suddenly gets a bit too bold for everyone’s liking. You can be cold and unhappy because you’re cold, or warmer and unhappy because of the financial consequences of not being cold; those are your current consumer choices.

Right now we’re working on the final preparations for the launch of our new report “Workplace AI Market Analysis: Generative AI and the Desktop (R)Evolution” which will be available for both existing subscribers and additionally as a one-off purchase for non subscribers in the next few days. The primary reason for writing this specific report at this time is to provide some deployable advice to organizations asking themselves – amidst a torrent of generative AI product announcements from their both incumbent and prospective software suppliers – “where do we start?”.

Whereas there has been quite a lot of advice which you could paraphrase as “don’t ask questions, just get started with GenAI or the world will leave you behind in their innovative dust clouds” (and trust me, I’ll be getting to that in a future post), we’ve taken an alternative path with this report. Instead, we’ve drawn up a guide to the most common use cases being suggested by those offering GenAI tools for the desktops of the workforce, how you can evaluate them for your organization and best build project structures that can deliver realistic results. What it lacks in hyperbole, we believe it delivers in practicality and value.

If the days of spring were dominated by the emergence of GenAI everywhere and all at once within business applications, autumn appears to be the season for the fall in expectations. The Wall Street Journal recently published a piece – sadly behind their paywall, but here’s the gist – that this first generation of GenAI integrations are loss leaders for those software companies concerned. The article cites an example of Microsoft Copilot for GitHub being offered for $10 a month, but losing $20 a month on average, with some users costing the company $80 to service. Microsoft wouldn’t be a lone example of using an early low price to encourage use for an early adopter service from which it can learn and adapt future offers, in fact we wouldn’t be surprised if that’s the case for every GenAI application integration currently available. 

However, let’s also not kid ourselves here that what might be a loss leader for our suppliers is somehow close-to-free to use for customers. Looking more toward mainstream desktop applications, Microsoft Copilot for 365 is being pitched at $30 per user, per month. Which means for a 1000 seat organization, a cool $1m over 3 years for a service which might only be just about covering the cost of provision if we take WSJ’s information at face value. That pricepoint looks to be where alternatives from Google and Salesforce are also being pitched (before we get to customer specific discounting, of course).

Image: A excerpt from Chapter 3 of “Managing the (R)Evolution”; Identification and analysis: process

Our expectation is that the future state for the use of GenAI tools in business applications is likely to be billing through forms of metered use, just as we see for existing API-based tools elsewhere. Indeed, within our new report, we mention building around this expectation when examining which processes we might want to bold GenAI capabilities upon to ensure our economic analysis stacks up alongside our process and task analysis. It’s a vital piece of futureproofing, but one which is new and distinct to what has come before.

Where IT departments have become familiar with the realities of dealing with API metering on top of the complex calculations involved in the economics of cloud provisioning, there’s a degree to which this can be anticipated by existing load measurements and architectural patterns. Much of the current use case mechanics around GenAI use in business applications depends upon iterative use based on the human in the loop – the workforce – being happy to proceed having made unpredictably repeated use of the integration. Fine within an “all you can eat” user billing environment, less so when packages of calls within a metered environment becomes the norm and potentially punitive beyond those limits.

This is of course, only one small element amongst a broader range of points of planning we cover within the report (indeed, it’s actually covered pretty briefly) but it’s indicative of our larger point that the same sort of circumspection should be applied to introducing GenAI into your organizations as you would any other where the experience level is limited. There might be potentially large upside, but for it to be realized does not mean leaving everything you know about deploying technology behind to reach it, in the belief that the opportunity to do so is somehow closing. You know better than that.

* Title courtesy of the fabulous The Meters

Leave a Comment

We use cookies to ensure that we give you the best experience on our website. If you continue to use this site we will assume that you are happy with it.

Work Intelligence Market Analysis 2024-2029