Very Interesting. More and more rigorous third party studies are pointing out the major flaws with LLMs and agents. I believe these flaws will not be eliminated with major architectural changes which could take a long time given the relentless determinism of the current AI vendors. The real issue for me as you mention is that they are useful so how do we adapt culturally , process-wise and technologically to get the best of them while minimizing their significant deficits. Predictable unpredictability, like their failure at arithmetic, is easy to deal with. It is their unpredictable unpredictability that is difficult for humans to adapt to, in my opinion. I am reading this 12 page report over - and it is interrupting me - where should I be looking for the mistakes. Unfortunately the answer right now is everywhere
Yes, I believe Agentic AI will eventually replace SaaS except the SSO and Security/Privacy SaaS. Not sure how Agentic AI can cover that role. Might come up in the future.
Excellent write up. The hype and disinformation around agents is ridiculous. I think companies like Zapier have been smart to bolt-on an LLM (Zapier agents) to their process to get some easy money out of FOMO fuelled CEOs. Maybe their software engineers never told them these things have existed for years? 😅
Haha, in fact, Zapier has also filled all their marketing messages with AI
'The most connected AI orchestration platform'
'Build and ship AI workflows in minutes—no IT bottlenecks, no complexity. Just results.'
'240,000,000 AI tasks automated on Zapier (and counting)'...
An interesting trend is that (you probably knew) enterprises want to automate and get rid of engineers since the start of engineering (no code, Zapier,... now AI agents); BUT the irony is, even some engineers themselves worried that they'd be replace whenever there's a new wave of hype.
I suspect that most engineers will find it difficult to compete with consulting firms in convincing their CEOs about the technology directions.
Very Interesting. More and more rigorous third party studies are pointing out the major flaws with LLMs and agents. I believe these flaws will not be eliminated with major architectural changes which could take a long time given the relentless determinism of the current AI vendors. The real issue for me as you mention is that they are useful so how do we adapt culturally , process-wise and technologically to get the best of them while minimizing their significant deficits. Predictable unpredictability, like their failure at arithmetic, is easy to deal with. It is their unpredictable unpredictability that is difficult for humans to adapt to, in my opinion. I am reading this 12 page report over - and it is interrupting me - where should I be looking for the mistakes. Unfortunately the answer right now is everywhere
In case you've read my other comment, in short, it takes time... and a lot of companies will struggle their way out (as always).
BTW, I know I still owe you the draft idea about the livestream :D coming soon!
Yes, I believe Agentic AI will eventually replace SaaS except the SSO and Security/Privacy SaaS. Not sure how Agentic AI can cover that role. Might come up in the future.
There's more about the future of Saas in part 2, maybe you can find more ideas there :)
Excellent write up. The hype and disinformation around agents is ridiculous. I think companies like Zapier have been smart to bolt-on an LLM (Zapier agents) to their process to get some easy money out of FOMO fuelled CEOs. Maybe their software engineers never told them these things have existed for years? 😅
Haha, in fact, Zapier has also filled all their marketing messages with AI
'The most connected AI orchestration platform'
'Build and ship AI workflows in minutes—no IT bottlenecks, no complexity. Just results.'
'240,000,000 AI tasks automated on Zapier (and counting)'...
An interesting trend is that (you probably knew) enterprises want to automate and get rid of engineers since the start of engineering (no code, Zapier,... now AI agents); BUT the irony is, even some engineers themselves worried that they'd be replace whenever there's a new wave of hype.
I suspect that most engineers will find it difficult to compete with consulting firms in convincing their CEOs about the technology directions.