Salesforce is still hyped for AI everywhere! Especially its chatbot Agentforce, introduced in December! Agentforce didn’t work so well in outside tests. The reviewer said: “I can’t help but questio…
Here’s the thing about platforms trying to deliver analytics and AI like this:
Once you have like a million business clients, their use cases are way too varied for you to be able to provide personalized business value to them at scale.
So you go ahead and try adding AI, surely it can personalize content well.
Your internal tests looked good, the problems YOU thought were useful worked well.
But your clients try it out basically once.
By the time they’ve explained their problem to the AI they could have gotten it done the old way. And the next time they want to use it they have to bring it up to speed all over again.
So it doesn’t work seamlessly for their problems and they never touch it again. Middle management says maybe they’re just not exposed to it and they’re just unfamiliar with it, in your cherry picked tests it’s 40% faster, so you keep pushing it. But the client already decided it sucks and there’s no clear way to show that it doesn’t anymore.
You’re just shoving it down their throats trying to demonstrate your business impact in a metric so you don’t look like you wasted a ton of money on a bad product.
Which is why I like working for smaller places. We have a specific business problem and a relationship with users. Yes, I can understand their problem and solve that. And yes, I can put in an AI to help automate it too, because the problem is well understood and documented.
Anyways, that’s my rant, I’m bearish on companies like sales force shoving AI into their products in the hope it makes them better. The problem is their scale makes it impossible to do so, and if they can an off the shelf LLM will can it just as well.
And your petabytes of industry data aren’t helpful either, because you’re not actually using them to build a better AI model.
Your internal tests looked good, the problems YOU thought were useful worked well.
The only job AI is great at, is someone else job. YOUR job is complex and nuanced and has challenging aspects, but someone else’s job is easy, AI could do it easily.
The only job AI is great at, is someone else job. YOUR job is complex and nuanced and has challenging aspects, but someone else’s job is easy, AI could do it easily.
I noticed that, too. It’s quite a notable coincidence.
Here’s the thing about platforms trying to deliver analytics and AI like this:
Once you have like a million business clients, their use cases are way too varied for you to be able to provide personalized business value to them at scale.
So you go ahead and try adding AI, surely it can personalize content well.
Your internal tests looked good, the problems YOU thought were useful worked well.
But your clients try it out basically once.
By the time they’ve explained their problem to the AI they could have gotten it done the old way. And the next time they want to use it they have to bring it up to speed all over again.
So it doesn’t work seamlessly for their problems and they never touch it again. Middle management says maybe they’re just not exposed to it and they’re just unfamiliar with it, in your cherry picked tests it’s 40% faster, so you keep pushing it. But the client already decided it sucks and there’s no clear way to show that it doesn’t anymore.
You’re just shoving it down their throats trying to demonstrate your business impact in a metric so you don’t look like you wasted a ton of money on a bad product.
Which is why I like working for smaller places. We have a specific business problem and a relationship with users. Yes, I can understand their problem and solve that. And yes, I can put in an AI to help automate it too, because the problem is well understood and documented.
Anyways, that’s my rant, I’m bearish on companies like sales force shoving AI into their products in the hope it makes them better. The problem is their scale makes it impossible to do so, and if they can an off the shelf LLM will can it just as well.
And your petabytes of industry data aren’t helpful either, because you’re not actually using them to build a better AI model.
The only job AI is great at, is someone else job. YOUR job is complex and nuanced and has challenging aspects, but someone else’s job is easy, AI could do it easily.
I noticed that, too. It’s quite a notable coincidence.