Blog > Which types of AI training should investment managers seek?

Which types of AI training should investment managers seek?

Prompts alone won’t save the day

Uncategorized

For a long time, generative AI wasn’t quite the headline-grabber it is today. Initially OpenAI was very secretive about its products. The company knew it was ahead of everyone else, and wanted to keep it that way. It was therefore very selective about who it gave access to try its models.

Then, in April 2022, they opened up GPT-3 and allowed anyone who wanted to play with it. Already, it was shockingly impressive. As a personal matter, I felt marketers needed to know about this.

Within a month, I had recorded a series of training videos to post on LinkedIn. Here was the first:

This was six months before ChatGPT was unleashed onto the world. Since then, a trickle of companies began approaching me directly to deliver training on using AI in their work.

Over time it developed organically into a service that FinText offers. And in early 2023, in our article Three Marketing Trends Now That AI Can Write, we made the following prediction:

We’re seeing three things happening.
For a start, some companies are burying their heads in the sand, pretending this has nothing to do with them.

Others, currently the minority, are grabbing the bull by its horns and working on how they can safely use generative AI in their day-to-day.

Lastly, and most interestingly, we see employees using it on their own initiative. They do so mostly privately. And they use it to cut down the workload, or tackle urgent, unexpected tasks.

This year, we’ll see some companies move from the first group to the second, and a whole lot of companies –  unwittingly – move to the third.

A year on, companies are more switched on. Investment managers realise the technology can offer a lot, very cheaply, but are frustrated that there’s no clear route to using it productively.

No question, your company can find someone who will agree to provide training on AI. Whether that knowledge actually translates to useful changes is a very different matter.

Now, obviously, we’re speaking from position here, because we’ve been delivering AI training, and believe in our offering. But equally, the need for knowledge far outstrips any single company’s capacity.

So, whether it’s us or someone else, there are several key elements we recommend you insist on in whichever AI training you select:

1. Does it Explain what a Large Language Model is?

You can absolutely use generative AI without knowing how it works. After all, our kitchens are full of appliances we can use but not build – why should this be any different?

But to get the most out of AI tools, it helps to understand how they ‘think’, especially when it comes to generating text. Because what they produce looks so much like English, it’s easy to believe that’s what they’ve learned.

The tools built on top of generative AI will change and improve. But most will likely be built on top of the same foundational technology. Understanding the principles behind it will help make use of whatever tool fits your specific needs.

2. Does it Understand Your Needs and Processes?

Here’s Institutional Investor reporting on a new tool Bloomberg rolled out to equity analysts:

Anyone can copy and paste a public company’s earnings report into a growing list of AI tools and ask them to summarize it. Bloomberg says it is offering something different.

Bloomberg’s First Generative AI Tool Hits the Terminal, Institutional Investor

That something different was a deeper understanding of what equity analysts actually needed, an understanding that was achieved because the project was managed by a former experienced analyst:

Among other things they know investors are focused on, the summaries include context on a company’s guidance, capital allocation, hiring and labor plans, the macro environment, new products, supply chain issues, and consumer demand.

“It’s been an iteration [process] to the point where we got comfortable that this makes a difference, this is unique, this is differentiated,” said Andrew Skala, who spent nearly 10 years as a sell-side researcher before joining Bloomberg.

Iterating and testing will be part of your company’s journey with AI even if the company won’t be training their own models (like Bloomberg has done) and using off-the-shelf products.

But there has to be a method to this iterating and testing, or else it won’t converge to something useful.

3. Does it provide hands-on Testing during the sessions?

If a training session leaves your employees to experiment with the technology after the session, you can safely assume it won’t happen.

There is a ton of resistance to overcome when approaching any new technology service. The appetite is even smaller when the benefits of the service are uncertain and require trial and error.

This pattern is even evident in ChatGPT, which is by far the most popular generative AI service. When you observe the Google search volumes, you can witness a meteoric initial growth. But it stalled. However big, these numbers are nowhere near the total amount of professional-services employees.

The same is true more generally: for any hope of generative AI to take route, the initial experience has to be in a guided environment, where people are comfortable trying and questioning what is happening.

For this reason, incidentally, we no longer do AI training remotely. We did it a couple of times, and the differences between in-person and remote were clear enough.

If you’re considering investing in AI training and wish to learn more, let’s talk!