On Thursday I held a 1-hour fireside chat with Azeem Azhar, the founder of Exponential View and an extremely smart AI futurist. (Watch the recording here).

Azeem runs his own AI business, and he talks daily with Fortune 500 CEOs, advising them on how to implement AI. So he’s got an expansive view of where the money will be made (and what’s a waste of time).

Here are my top 10 takeaways from our conversation.‍

1. You have to spend the $20 per month.

‍If you’re not getting reimbursed by your company (and you should be), you have to spend the $20/month for ChatGPT Plus to understand what’s going on. 

“You may as well have the toolkit, even if you can’t apply it to every aspect of your work,” Azeem said. “You should be familiar with the products and how they’re working, because this will start to be the infrastructure and interface of the future. Why not invest in yourself?”

‍2. The early adopters won’t necessarily be the AI bosses of the future.

‍Yes, it pays to be an early adopter right now – but that doesn’t mean you’ll be in charge of the next AOL in five years.

“Just like with the internet, people who use AI tools now will have an advantage – but success in business is also dependent on judgment and social capital,” Azeem said. “The people who got into the internet early didn’t all end up being the bosses of the future, because you needed more than that – you needed human skills.”

3. Start internally to mitigate costs.

‍When you’re adopting AI, start internally for two reasons: one, your employees will tolerate much more failure than your customers, and two, the costs of delivering AI to your customers will be a lot higher.

“If you’re a company with 100,000 customers, the costs of delivering AI to them can ratchet up pretty quickly,” Azeem said. “But if you’ve got 500 employees, your costs are going to be containable – so it helps on the risk side and it levels up your collective intelligence.”

4. If you’re founding a startup, go vertical and solve a specific problem.

You probably can’t found the next OpenAI. Instead, you should find a well-defined customer set that you know really, really well, and where you can get access to good data. 

“The Perplexity or OpenAI model is going to be a very tough ask for most companies,” said Azeem. “I’m not sure how much room there will be for that – but I suspect the vertical applications will create lots of opportunities. For example, if you’ve built an LLM that helps dental surgeries, and you have that customer relationship and that data, that’s the source of your value.” 

5. … Or get into the ‘picks and shovels’ of AI.

‍There will be a massive build-out of infrastructure around AI. AI requires a huge amount of chips, and chips need data centers and they need power. 

“Construction will do well out of this, and renewables and batteries and installers,” said Azeem. “People who train HVAC engineers will do well, because data centers are full of HVAC. Don’t look at the technology – look at the problem and the market size and the customer set.” 

‍6. If you’re a large incumbent, look to your existing problems first.

‍Large companies get a bad rap about their ability to survive technology waves – but you’ll succeed if you listen to your customer relationships and isolate things to fix first.

“Cable companies have a big issue with customer service,” Azeem said. “I don’t think they should be thinking about new products – they should be thinking, ‘How do we improve our NPS or our customer service with AI?’” 

‍7. If you own your own business, use AI to sell better.

‍If you run a company and you depend on clients for your revenue, use AI to market and sell your product. For example: Use ChatGPt to research your prospects and have better, more empathetic conversations with them, or feed your customer feedback into an LLM to get you data on pain points to tackle.

“People who are great at relationships have had an advantage, but someone like me – more of a nerd – can use the help to build customer relationships and analyze feedback,” Azeem said. “Using AI [to get to know your customers] is saving you time and getting you to the next level of performance far more quickly.”

8. Don’t be seduced by ‘the dirty sea of mediocrity.’

In between the routine tasks and the high-stakes tasks that AI excels at, there’s a ‘dirty sea of mediocrity’ where AI can be deployed to little or negative effect.

“The danger is that you say, ‘We’ll halve our social media marketing team because we can produce this content through ChatGPT, and it’s really terrible quality,” Azeem said. “Come back to what your customer cares about. Does your customer really care about getting five marketing emails from you a week instead of one?”

9. You don’t need to know every technical skill or piece of jargon – you just need to get your hands dirty.

‍If you’re overwhelmed by the AI terminology buzzing around, that’s intentional on the part of AI companies. They want you to feel stupid so that you’ll buy something from them. You don’t need to learn Python or master every buzzword you hear, but you do need to learn the basics.

“The most important thing right now, in working with AI, is being specific and thoughtful about your line of thinking,” said Azeem. “The difference between people who use it well and people who don’t, is that those who use it well can structure their question and give direction. That’s the skill you need – not Python.”

10. Start small, start personal.

‍If you’re not using AI yet, use it for personal use cases and experiment for fun. 

“I walked around my garden taking pictures of weeds, then asking what weed killer to use for them,” Azeem said. “Don’t ask for it to give you the date Kennedy died or the score of the basketball game you’re interested in – you’ve got Google for that. Instead, do some experimentation on open-ended questions and slowly build your confidence.”