I think we are on the cusp of an AI-course correction. I know I'm not alone in thinking this - almost daily, news stories are published about companies doing seemingly implausible things in the name of AI adoption. And quite frankly, I think most of them are wildly optimistic about AI's actual impact on end-to-end workflows.
This type of theatrical AI application is what I like to categorize as "just because we can do it, should we?" CEOs and other C-suite members are typically far removed from the actual work at large companies. That isn't to suggest they don't work - but the work they do isn't what keeps the business running day to day. Their role is to compile information and make strategic decisions in the best interest of the firm. Unfortunately, because of that role, they don't see the processes held together with tape and one legacy employee whose entire operational playbook lives in their head (of which there are inevitably many). They don't see the value middle management creates in knitting employee productivity to strategic goals. And they don't see the disparate systems housing years of questionably accurate data that requires significant massaging before it can be trusted.
Yet, as we've seen all over the news, the C-suite is making decisions to "transition" processes en masse to AI at the expense of workers. The problem with this approach is two-fold:
First, not every process can be automated. And even if it could, the architecting, data cleanup, and refining it would take real time. I tried to help a friend automate a process that seemed simple last week - she wanted AI to take data from a worksheet template and apply it to a services agreement contract. After about two hours, we gave up trying to retrofit it and decided the entire process needed to be rearchitected instead.
Second, AI isn't free. Even with current subsidies, it's becoming cost-prohibitive for companies once a toxic "we do everything with AI" culture takes hold. And the subsidized cost doesn't account for the external costs of computing power and its environmental toll - if markets correct to reflect the full cost, including mitigation, these tools will get expensive fast.
As a services and customer success leader, I find that using AI to augment rather than replace is the real key to gaining efficiency without adding to the noise around AI adoption - though that's not what executive leadership wants to hear. AI is good at standardizing the things humans aren't great at: note-taking, reviewing and retaining large swaths of information, organizing timelines and thoughts. It levels the playing field on a lot of the low-value work teams struggle with. It can also measure adoption in SaaS - but only if customers are expected to use a tool in a fairly standard way. Real analysis still requires domain expertise, and the judgment to gut-check what the AI concludes. AI can synthesize expertise, but I don't believe customer-facing roles should be replaced with AI-first experiences. Instead, I think the course correction comes when customers get exasperated explaining the intricacies of their business to a machine with short-term memory issues - and the cost, both real and in lost logos, outweighs the benefit of using cutting-edge tech.