I'm sure you've all read the sheer echo chamber of posts on AI be it from this forum, LinkedIn or God knows how many internal corporate communications.
I thought to take a temperature of the room and share my own thoughts on what I've seen from my vantage point.
- Productivity - AI has objectively made specific tasks or even workflows easier. This is especially true for research aggregation and first-class corporate messaging as well as aiding in structuring or brain storming large problems. On the flip side, it also regurgitates stale or inaccurate content which is an issue in SaaS and especially given how quickly products or processes change. It also does not scrutinize the source and sometimes uses local documents or vague email threads as proof.
- ROI - this one has been evident for some time. AI is not leading to ROI gains as touted. This is seen through an uptick in employee growth and documented exponential costs (explicit and implicit) from use and implementation of AI.
- Externalities - the first unintended consequence of over-reliance on AI is how it fundamentally weakens our ability to think, parse through complex bodies of text and distill nuance. I'm guilty of this myself at times. And being cognizant of it has made me realize that there may need to be an age restriction to AI so kids who haven't yet developed the critical thinking don't outsource it at that stage. The second is that we have yet to finalize the commercial models in place for AI as an industry with every firm in a trial and error phase to see what's worth paying for and what balance to strike. This will likely lead to a push for and pull back from AI in the continued few years as use cases are refined with greater granularity.
Where I stand: I've seen AI do some really impressive stuff for clients including building firm-defined financial models and slide decks to ~70-80% accuracy. I've also seen the cost of those models and implementation breach the cost of the people that has managed the workload prior. I also think the overwhelming consensus is that AI cannot handle nuance or complexity often not prioritizing relevance or brevity in text. This all ignores or pauses the errors that can be small from a misinterpreted prompt to a legal or security issue.
All that said, I envision AI adoption will peel back and be used with pre-defined prompts or agents where the usage is known and output is proven to be high ROI. Until then, a few firms will rise and fall.
