Banking on Information with Alec Crawford CEO and Founder of Artificial Intelligence Risk Inc

In this episode of Banking on Information, Rutger van Faassen interviews Alec Crawford, CEO and Founder of Artificial Intelligence Risk Inc. They discuss the importance of managing AI risks, the need for cybersecurity in AI applications, and the role of governance and compliance in ensuring safe AI usage. Alec shares insights on the future of AI in customer service and emphasizes the necessity for individuals and organizations to embrace AI technology to stay relevant in an evolving landscape.

Takeaways

• Alec Crawford has a long history in AI and finance.
  
  
• The need for AI risk management is critical for companies.
  
  
• Cybersecurity is essential for the safe deployment of AI.
  
  
• Governance in AI involves setting rules for its use.
  
  
• AI can significantly improve efficiency in customer service.
  
  
• Companies must encrypt sensitive data to prevent breaches.
  
  
• AI agents can help reduce customer service call times.
  
  
• The future of AI should focus on inclusivity and job repurposing.
  
  
• Organizations need to adopt AI tools to remain competitive.
  
  
• The next generation of AI will surpass current capabilities.
  
  

Chapters
00:00 Introduction to AI Risk Management
02:56 The Importance of AI Governance
06:01 Enhancing Efficiency with AI Agents
08:48 Cybersecurity in the Age of AI
11:49 Future of AI and Its Societal Impact

Keywords

AI, cybersecurity, risk management, governance, compliance, future of AI, customer service, technology, innovation, banking

Transcript

Rutger van Faassen (00:01.537)
Hello and welcome to another episode of Banking on Information. Today, my guest is Alec Crawford, who is a founder and CEO of Artificial Intelligence Risk Inc. Welcome to the podcast.

Alec Crawford (00:15.65)
Thanks, Rutger great to be here and great to see you.

Rutger van Faassen (00:18.519)
Yeah, it almost feels like yesterday that we were in a similar recording studio. It's like this is kind of interesting. But let's dive right in with the first question that I always ask everyone. Why do you do what you do?

Alec Crawford (00:33.998)
Yeah, well, it really goes back to a little bit to my background and then the origin story of the company. Like my background is, as first as an AI person, I was developing neural networks from scratch in 1987. So fast forward through three decades of finance, doing various things like training and managing risk and working at big banks and doing technology and AI things.

I retired from Lord Abbett and, playing a lot of golf and looking around. And then I just see these giant companies and banks onboarding AI with no guardrails and just said, this is crazy. It's not just bad for them. It's bad for their customers. And if things go wrong, it's bad for society, right? Like if all the banks melt down or get hacked or whatever. So I decided to start a company.

to help not just banks, but all companies deal with managing the risks of AI and doing it safely. And that morphed into accelerating the adoption of GEN.AI safely is our motto, because we can deploy our software and the models and connect the bank data and all that kind of stuff in a day now, which is obviously unheard of. are like, what? A day?

Rutger van Faassen (01:56.458)
Yeah. So you are a techie at heart and were for a long time, but you're passionate about making sure that we don't go off the rails with this new AI wave that we're in and that we're doing it safely. So that is a great why. Now, how does that translate into what you do? Can you describe kind of like what the main use cases are of what you do for your customers?

Alec Crawford (02:20.812)
Yeah. So, so think about, what would happen at your company or your bank if all of a sudden you said, you know, we're just not going to do any of cybersecurity stuff. Who needs that? We don't need firewalls and virus detectors or that. We're fine. We'll be fine. Right? Like, no, that's a disaster, right? You'd be out of business in less than a day. Right? So that's how important it is. We're not quite there yet with gen AI. In other words, if you don't have cybersecurity, yeah, you're not dead yet. Maybe soon.

Rutger van Faassen (02:36.757)
Alright, yep.

Rutger van Faassen (02:48.596)
Yep. Yep.

Alec Crawford (02:49.902)
Right? You're going to need security, cybersecurity for Gen.ai. So how we do it is we take the models themselves, whether it's Lama 3 or OpenAI or whatever, deploy them on-prem or on a company's private cloud, deploy our software, connect that together, create the compliance database, hook into Outlook and Teams and...

Rutger van Faassen (02:59.35)
Mm-hmm. Yep.

Yep.

Alec Crawford (03:16.204)
document storage and SQL databases and third party APIs. And you're kind of often running with a front end and an agent building the system. And all of it gets covered by our kind of governance, risk management, compliance, and cybersecurity platform. Think about it as like the operating system for AI that will keep you safe and allow people to do.

Certain people do certain things with AI and others to do different things with AI

Rutger van Faassen (03:46.976)
Yeah, yeah, so making sure that you have one kind of central location where you keep tabs on everything that's happening and making sure that there's no big risks getting out of hand. So you want to have the management center, that's your platform.

Alec Crawford (04:03.584)
Exactly. The operating system for AI. And if you think about governance, like what does governance mean? Right? So, so part of it is what are you going to allow people to do with AI? Right? Cause there may just simply be something at a company where it's like, we're never going to do that with AI. for example, maybe you want a rule that says we're never going to write Rutger's performance review with AI. Right? That could be a role. Right. And, but how do you, you know, how do you enforce that? Right? Like, how do you make sure that's going to happen?

Rutger van Faassen (04:05.897)
Yeah. Yeah.

Rutger van Faassen (04:23.753)
Right. Yep. Mmm.

Alec Crawford (04:32.376)
Well, partly it's upfront. You create these AI agents that do different things. And you might have one that's a, you know, social media content agent, but you can also tell that agent, eh, don't write performance reviews. Right. And then by the way, you're tracking everything that happens in AI with our centralized compliance and audit tool. And in that you can, you can have, on a monthly basis, the compliance team go in and say, Hey, has anybody been writing performance reviews with AI? And hopefully the answer is no. If that's.

Rutger van Faassen (04:52.062)
Mm-hmm.

Rutger van Faassen (05:00.244)
You're right. Yeah.

Alec Crawford (05:02.2)
something that's not allowed in your system.

Rutger van Faassen (05:04.935)
is like that's a great example. Now when you talk to your customers, how do they describe the value that they get from working with you and from working with the platform?

Alec Crawford (05:14.37)
Yeah. So part of it's hidden value, right? Meaning they're getting the governance and the risk management and cybersecurity, which is always kind of hidden, right? And also the compliance stuff is kind of hidden too, right? Cause most people are never going to see that. The compliance team, the admin team, the legal team, your CEO, maybe getting reports or whatever. So there's value to them in that kind of being able to observe what's happening and do things like, Hey, what are the top 10 questions being asked? Like,

Rutger van Faassen (05:26.577)
Mm-hmm. Yep.

Rutger van Faassen (05:36.04)
Mm-mm.

Alec Crawford (05:41.688)
How much money am I spending on tokens? What's the top agent? Who's the top users? Like that's all important information for management to have, but most people are never going to see it. So I think a lot of the value comes from our agent building capabilities and also the cybersecurity around those agents. Everyone is saying agentic AI and AI agents and things like that. We've been doing this for years at this point.

And the agent part's important, the security part's super important. Because if all of a sudden, let's take an example, I create an agent that can make a loan at my bank. Well, what happens if that gets out in the wilds? Or everybody at my bank can use that agent. That's a disaster, right? So you need to have security around that. Who can use that agent? What's the agent allowed to do? What data can it access? Things like that. Those are all things that we do. So I think...

Rutger van Faassen (06:18.621)
Yep. Yep.

Alec Crawford (06:40.152)
Part of the value also is the, effectively call it the agent app store, right? So our clients can choose from among a whole bunch of preset agents based on what business they're in. right? A great example is some customer service agents. They obviously need to be customized. right? For the customer service team, they can answer questions, they can review a call, makes it, you know, give helpful tips.

Rutger van Faassen (06:48.755)
Mm-hmm.

Alec Crawford (07:08.558)
And I'll tell you story about one of our clients where they built their FAQ agent for new customer service people, super helpful, basically saved them, know, something between 40 and 50 % of the time on a call. So call used to be 10 minutes. Now it's five minutes because the agents can look up the questions, the answers to questions super quickly. But then even more interesting, we created another agent where it goes back and looks at the transcripts of all the calls.

customer service and says, what are customers confused about? What are the top three questions they're asking? They went through that, figured that out, and then they changed their bank website to address those questions on the website. Now they get half as many phone calls. So overall, 75 % reduction of time on the phone, half from reducing the time on a call, and they've also reduced the number of calls by half by discovering

Rutger van Faassen (07:42.268)
Mmm. Yeah.

Alec Crawford (08:06.926)
of what are customers confused about in addressing it upfront.

Rutger van Faassen (08:08.967)
Yeah. So good governance translates into more efficiency, but also reducing risk. right? And obviously that's in the name of your company, making sure that you manage those risks. And in banking, banks are very well aware of that. That's most of what they do. right? Managing risks and you help them manage the risks on AI. Now that's very valuable.

Alec Crawford (08:28.014)
Absolutely.

Absolutely. Yeah. And a lot of it's around customer information, confidential information. Think about what's a ransomware attack. In the end, it's we've got your confidential information and we're going to tell everybody unless you pay us. And one of the things that we help our customers with is encrypting everything. So normally a graph database and Gen AI is not encrypted.

Rutger van Faassen (08:47.931)
Yep. Yeah, yeah.

Rutger van Faassen (08:55.75)
Right. Yep.

Alec Crawford (09:00.738)
You type something in there, it's out there in the graph database. You got your whole set of customer information there. Someone steals it. You're kind of hosed, right? We encrypt everything in motion and at rest in our system. right? So that's very helpful. The other thing which people don't think about is, all right, well, what if the customer service agent types someone's social security number or other confidential data in? That's got to get encrypted too, right? So you have to detect.

and encrypt that on the fly. It's not just data that you're storing in a database somewhere. It's also things happening on the fly. We do all of that. It really allows the bank executives to sleep at night, right? They know everything's encrypted and safe and you've got cybersecurity everywhere. So everything that comes into the model or leaves or comes into the bank or leaves is being

Rutger van Faassen (09:42.256)
Right, now that's very good. Yep.

Alec Crawford (09:55.102)
screen for Gen AI cybersecurity to prevent prompt injections and other types of hacks.

Rutger van Faassen (10:01.69)
Yeah, so you provide good sleep at night. This sounds very futuristic. I like to do this thing called future stinking. You're already very much on the cusp of that. But let's think even further out what a possible future could look like 10 years from now. What do you think the world looks like with this new wave of AI? Where are we 10 years from now?

Alec Crawford (10:22.284)
Yeah. 10 years is super hard. but I'll tell you, an experience I had recently where I'm on the phone with a credit card company and I'm, I need to send them a legal document, right? for a relative of mine. And I literally can't get a human. It's not going to be on one of the touch tone choices. It's not, want to pay my bill. Right? So it's basically figure out how do I get a human? It took me like five minutes to figure out how do I get a human on the phone?

The good news is over time, that's going to get vastly better to the point where the good news, bad news is that, yeah, you're still going to need human customer service agents 10 years from now, but not as many as today. And what I hope from a social standpoint is we as a society figure out like, how do we make everybody's job better? Get them a co-pilot? How do we repurpose people so they're doing things that are kind of more interesting?

I want the future to be inclusive and where AI is doing all these great things, especially stuff that we don't want to do, but not a situation where we end up with 14 % unemployment. So I think that's what we have to be careful about as a society as we develop AI and use it more.

Rutger van Faassen (11:41.413)
Yeah, no, That makes sense. a future where humans have a lot more time for themselves to do social things, AI will take over a lot of the workload, but not so much that the humans have nothing to do. So that's an interesting kind of possible future. So if you think about that and you do what we call backcasting, which is, so what should you do today to get ready for a future like that? What would you, what would you say that that is?

Alec Crawford (12:07.203)
Yeah.

Well, the most important thing, if you work in the knowledge space is you better be using AI now, right? So the classic line is AI is not gonna take your job. It's the people that know how to use AI are gonna take your job, right? So if you're not using at a minimum chat GPT and there are literally hundreds of tools out there, a great website.

there's, there's an AI for that. So you could be like, I need an AI that creates a presentation deck. I need neither creates a presentation deck. will find 20 for you. Right. Like, and it may be hard to figure out which the best one is that's going to take some work. but it's not just chat GPT. Right. Like you do have to look around. there are definitely some specialized AI tools out there that are.

Rutger van Faassen (12:43.693)
up right

Rutger van Faassen (12:53.092)
Right?

Alec Crawford (13:01.038)
Pretty cool. And obviously, look, everybody on the planet is going to embed AI in their software as fast as they can too. So it's going to be ubiquitous quite soon.

Rutger van Faassen (13:05.562)
Yeah. Yeah.

So that's great advice. So if you're still out there and I know who you are, you're still on the sideline, get off the sideline, right? Go do something with AI, start experimenting with it at a minimum, but you got to catch up because this is going to change everything.

Alec Crawford (13:27.554)
Totally agree it's gonna change everything. yeah, it is gonna be interesting to see the advancements. I think that, well, I do have one prediction for 10 years from now, and that is that we will have something better than large language models by then, right? Now, I don't know what it's gonna be. I don't know if it's gonna be composite AI or new kind of AI or whatever it's gonna be, but it's gonna be better and smarter than gen AI.

Rutger van Faassen (13:40.058)
Hmm.

Rutger van Faassen (13:44.377)
Right.

Rutger van Faassen (13:49.687)
Mm-hmm. Yeah.

Alec Crawford (13:55.842)
I think it's going to take a little longer than that to get something that's conscious, shall we say, or, or AGI. but we'll almost certainly, over the next two years has something that looks like AGI, even though it's not really AGI. We've already fooled Google engineers with Chat GPT, right? So it's not going to be hard to fool other people into thinking, we've got something that's actually thinking here.

Rutger van Faassen (13:58.957)
Right? Yep. Yep.

Rutger van Faassen (14:13.507)
Yep. Yep.

Alec Crawford (14:19.854)
Whereas what we have today is called statistical AI. It is not conscious, to be clear.

Rutger van Faassen (14:26.362)
Now think that's a great point to wrap this up on. A little bit of a cliffhanger, right? There's gonna be something else. We don't know what it is, but there's gonna be something else. This is great. Thank you so much for being on the show, Alec. And yeah, glad to have all your inputs.

Alec Crawford (14:40.824)
Great to be here, Rutger Thanks for having me on the show.

Rutger van Faassen (14:43.321)
Perfect. And until next time, choose to be curious.