What is inarguable is the fact that enterprise SaaS is incredibly profitable and, by default, incredibly expensive to use across time. To be clear, I am not against SaaS. My own company, Deep Analysis, uses many SaaS products, and I have recom- mended many firms to do the same over the years. SaaS can bring many advantages, especially to small and mid-sized firms. But you do have to go into these agreements with your eyes open, and, increasingly today, you have no choice, as the only option available is SaaS, since on-prem, perpetual license options have been systematically phased out.
Agentic AI
In what might seem like a hard pivot in this issue’s column, I want to talk about agentic AI and why it is so hot right now. Beyond all the technical innovations that have brought about this era of agentic AI, from robotic process automation to large language models and beyond, there is another innovation that gets little attention—the emerging pricing models for agentic AI. With agentic, we’re looking at hybrid pricing as vendors try to work out which formulation sets the right balance of low friction for clients with enough margin for them (considering the likely higher cost of delivery, because AI is hungry for all and any resources you offer it).
Familiar SaaS pricing elements such as seat licenses will give way largely to consumptive models, based on how much you’re using and—more controversially—the contribution the agent is providing you. Did it speed up the time to close out a transaction or reduce the need to escalate calls to a human agent? Outcome-based pricing models, based on whether the agent is able to meet agreed outcome criteria, are novel. It needs to be calculated as the value to the client for completing the task with an agent versus completing it in a more established way, then weighed against the per successful transaction fee the vendor wants to levy.
If this all sounds confusing, that’s because it is confusing. I won’t go quite as far as to say deliberately confusing, but there is a real concern here that enterprise costs will not only be unpredictable to manage, but also have the potential to skyrocket based on the calculations of an algorithm. If SaaS pricing caught many buyers out over the long term, how many more are in for sticker shock in the forthcoming agentic AI era?
Without a doubt, agentic AI will automate many currently inefficient processes and improve them beyond measure. But the claims coming from the giant IT vendors need to be taken with more than a grain of salt. Agents can’t automate all processes. In many cases, when deployed, they will not do the job well. Once in place, they may be very difficult to remove, and you may well find yourself stuck with them permanently—and paying for it. It’s worth noting that it took a decade for people to start to figure out exactly how cloud infrastructure pricing worked. Indeed, entire careers have been made on puzzling out those models, so don’t expect any of this to suddenly become crystal clear without assistance.
The name of this column is Ethical Innovation, and hence this topic on pricing fits squarely under the banner. We are certainly going to be seeing a variety of innovative pricing structures for agentic AI coming to market during the next few years. I have no doubt they will be adhering to legalities in our free, and likely, soon to be less regulated, market. Whether they will be ethical is another question altogether. As the legalese Latin phrase goes: “Caveat emptor”—“buyer beware.”