You're probably using AI wrong

And that's ok.

Everyone is rushing to get something with AI in the title going, so they can tell people they are an "AI First, Data Driven Organisation (TM)". Because of this rush, I've seen a bunch of mistakes crop up over and over again.

Here are the Top 3:

Mistake 1: Creating Generic Content

AI makes it trivial to create generic content.

So everybody does it. But, AI has made making personalised content easier as well, and so few people are doing it. Still showing every visitor to your website the same thing?

Those days are gone.

Show everyone something for them.

It's easier than ever to give people ultra personalised content. It makes people way more likely to engage. There are off the shelf tools that handle this for you too (check out planform.ai))

Don't use AI as an excuse to publish slop, use it to build trust and close sales!

Mistake 2 : Automating the Creative Work

AI is not creative, at all.

What is a Large Language Model(LLM) actually doing when you ask it a question? It's predicting what the next most likely word in the sentence should be. That's it.

LLMs like ChatGPT are trained on huge volumes of text and learn how to predict what words should come next in a passage of text. This means they never come up with anything new. It's always some combination of things it has seen before.

The illusion of novelty is powerful, but definitely an illusion.

Why does this matter?

Outsourcing your creative work to something uncreative is actually fine in the short term. Get out a first draft quickly, iterate from there. Technology helps you move fast.

The problem arises when you automate the creative work to spend more time on the automatable tasks that really should be given to AI.

Yeah, its great that you wrote 100 cold emails in 5 mins with ChatGPT, but firstly: you now have 100 perfectly average emails that are unlikely to convert any clients (they get 100 ChatGPT emails a day); Secondly, you spend the next 55 mins fiddling with the margins on your landing page.

AI lets you spend your time on the highest leverage, creative tasks.

Don't use it to do the opposite!

Mistake 3: Dumping Spreadsheets into an LLM

AI is useful, but not for everything.

Exhibit A: Spreadsheets. Most enterprises live and die on spreadsheets so everyone wants a way to just make AI handle them. Unfortunately, LLMs are massively limited in how they deal with structured data. They often rely on writing code and formulas to interpret data as they have no way of "understanding" tabular data themselves.

The worst part is, the LLMs think that they understand the data and answer confidently and often entirely incorrectly.

I spoke to someone who had been dumping monthly sales data into an LLM and just asking it for "trend analysis".

They got completely reasonable answers and presented the findings to the rest of their company. Except, the numbers were completely made up. The LLM had just blurted out the most likely answers, not the correct ones.

When using LLMs, know their limits!