Generative AI – the automatic process cop?

Getting your Trinity Audio player ready...

• Generative AI could have a role in process management.
• The technology could help smooth international regulatory issues.
• It could free businesses to focus on their core strengths.

Generative AI seems like it’s almost everywhere, just eight months into its widespread existence. And every day new companies find new ways to embed it into their systems – including financial firms working internationally.

We spoke to Dennis Winter, CTO at Solaris, an embedded finance firm, to ask him what role he saw for generative AI in the future of finance.

DW:

Here’s the thing. We’re a very young company ourselves, so innovation has always been integral to what we do. That includes adding the likes of automation to our systems, so we understand the need to add new technologies.

No-one really identified machine learning as a thing that you should actually invest in. But then generative AI exploded in the last quarter of last year, when ChatGPT became accessible to more and more people.

Generative AI process proxies?

And if we look into the last three months or so, it’s mind-boggling what’s happened out there. In March, we had a hackathon internally, looking at how we could potentially use generative AI ourselves, and our engineers said it was quite interesting, but somehow it was too heavy for us to really use.

Just a couple of weeks later, this whole project was taken up by the open-source community, and we ended up with a solution where this whole thing has shrunk down to a technology where you can deploy a large language model on a laptop, you can train it on a laptop, and you can utilize it in a specific area of your organization, which of course, significantly simplifies the whole question.

Everyone was concerned about how we could use objectivity for our work. When it’s in the United States, we’re sending our data over there, and we have no idea how the training actually works. Which means you run the risk of having data protection officers looking into how you’re using it.

I was at a conference last week, and the first CTO I ran into told me that he’s just founding an AI company which acts as a proxy between GDPR and the United States for AI SaaS organizations. Which means they’re trying to extract PII data out of a data stream, and then enrich it again on the way back.

So there’s a lot happening in this whole regulatory field where AI could be used – data protection will always be an issue, particularly in the European Union.

THQ:

Where do you see the space for generative AI there?

Generative AI: organizational process development.

DW:

You have to think about organizational development. How do you adapt to such things and not fall behind? Because if you’re not doing it, there will be companies out there that right now are just playing around with it, but will come up with strategies to utilize generative AI.

First, you need to come to a conclusion about what you tell your people, what you do with the people working in the organization, because compared to say blockchain, generative AI has a very low entry barrier.

For people to understand and work with blockchain, people had to understand the technology – what a private key is, and what blockchain itself means, and how a general ledger works and so on.

For generative AI, you have this simple entry field that you know from Google, and you enter your question in much the same way as many people also do on Google. And all of a sudden you get a lot of information back.

So helping people understand how to use it in a secure manner is crucial – as is working out how to leverage it internally.

People are waiting for a kind of iPhone moment, where they look at things again and find out that there are many processes internally that you can optimize or make more efficient using technology like generative AI.

Many organizations at the moment are going through this exercise of understanding what the actual use cases are which are really improved by generative AI and need those new language models.

The thing is that you can, for a lot of those processes, just use regular machine learning to get similar results.

But then there are new internal use cases for generative AI, like customer support, when you’re just feeding all your data, like all your tickets, into one language model, and you allow your customer service people to ask questions about the customers you have, what’s happened over the last couple of conversations, to make the process more efficient.

THQ:

All of this sounds like a fairly reasonable approach to the new technology, no?

Generative AI: regulatory process checker?

DW:

If we take a step back, I’m convinced that the whole topic of machine learning and generative AI will show up more and more in regulatory technology companies, like the one proposed by the guy that I told you about, so as to keep other organizations compliant in regards to data protection.

THQ:

So, generative AI operating at machine speed to maintain jurisdictional compliance? Sounds like a good use of the technology.

DW:

Well, yes, because when companies are branching out, they go international, they want to work in different jurisdictions, and you have different regulations in these jurisdictions. And that puts a huge burden on companies, managing all that.

From our side in finance, we have a German banking licence, and just branching out into other countries within central Europe puts a huge burden on us, because there is slightly different reporting regulations, different tax regulations, and so on and so forth. And this just blows up your organization.

You usually go through a discovery phase, where you have a certain assumption, you set things up in a certain manner, then you find out that things need to happen differently in Italy, Spain, or wherever. And this process of bumping into the walls costs you a lot of money.

THQ:

So that’s where you see generative AI coming into its own? Smoothing out costly business processes?

DW:

Yes, my interpretation is that there will be more and more organizations that specialize in setting up processes from a regulatory perspective in the finance sector, in the health sector, and so on, particularly whenever it comes to data protection.

They’ll help organizations to stick to the things they know best, where they can create their value. And at the same time, make their money by making sure that certain things are in check using AI, using efficient processes or machine learning.

In our case, that will include things like fraud detection, which is a global issue. But the background processes, the back office and things like that will be outsourced to organizations that are utilizing AI and machine learning.

 

In Part 2 of this article, we’ll explore what using generative AI as process translators could look like – especially in the wake of the open-source development of smaller, more concise language models.