AI Adoption: Workers vs. Management

The Wall Street Journal recently published an article addressing surveys on AI adoption and perceived value in Management vs "Workers." (links below) There was a lot of information in the graphs, but...

1)  It should surprise no one that AI is more popular with management than workers. They are incentivized to be Pollyanna-ish about reducing labor and improving productivity. Delivering more value at lower costs is their ostensible job.

2)  Unfortunately, many of these managers are the same folks who have de-prioritized pipeline and tool improvements that would drive the same results as they're hoping for with AI. Workers remember.

3)  Given general change management within orgs, it's unlikely that there is a detailed implementation plan, or that the workers are given training, or time to see whether AI might improve their workflow. Meanwhile, workers are still required to deliver at the same (or increased pace) as they always have been. If I have a workflow that is so engrained I don't have to think about it, I will absolutely prioritize that over learning some new system, even if it is objectively less efficient. It takes time and cognitive energy to build new habits. Does management have a clear idea what "AI adoption" means or are they just hoping not to get caught flat-footed in another mass extinction event?

4)  AI is currently like early OCR in many ways; i.e. it's right some percentage of the time (60-99% given the task, the query, and whatever happens within the LLM at the moment). For the vast majority of tasks, it's not trustworthy. Which means it's potentially more work to use AI, review, and correct it. If you're an exec, you can say, "we're all reasonable humans here; you get my drift." But, fairness aside, you should not be doing this as a "worker."

5)  The vast majority of AI-integrated office work is geared toward the sort of summarize this, sketch out a deck, and handle the CYA tasks that management deals with. AI is specious at making anything, except writing code for known domains (which it is quite good at).

6)  Does management understand the difference between a prototype and shipping? AI is great for prototypes, but creating shippable anything requires more thought and effort, and this includes reports and presentations and considered strategies. AI can often help with these things, but it's not magic; it is, however, magical.

Everything is going to look much rosier to management than to someone who actually has to implement something and deal with all the cross-department politics and roadmaps, etc. All this may change over time, but we're not there yet, and we're racing toward the trough of disillusionment. AI is Not a Product. We need to stop focusing on AI, and start focusing more on delivering real benefits; less about the electric motor, and more about the vibrator.