For nearly two years, wealthtech firm Orion has been testing and deploying different types of artificial intelligence tools, both to streamline internal workflows and to help their advisor clients with smarter data analytics.
Orion, which services more than $4 trillion in assets under administration and $65 billion of wealth management platform assets, has been on a busy tech streak. Last month, the Omaha, Nebraska-based firm launched an AI-backed investor behavioral tracking tool called PulseCheck, as well as an app that helps users more easily streamline different Orion services and platforms to see data and portfolio changes on one dashboard. The firm also launched an advanced estate planning reader tool and an enhanced compliance tool for post-trade supervision during its flagship conference in March.
Yet even Brian McLaughlin, the Tesla-driving, AI-optimistic president of Orion Advisor Technology, warns that AI isn't ready for a full embrace from financial advisors.
McLaughlin, the founder and former CEO of Redtail Technology — maker of the CRM software popular with advisors — joined Orion after it acquired his firm in 2022.
"The hallucinations are what's scary," he said about AI. "There'll be a bad recommendation, or a poor choice or decision made in the use of AI, and that's something I actually don't think advisors should get on board with yet because it's just too early."
Yet there are areas in which AI has great potential for advisors, he said. McLaughlin sat down with Financial Planning to discuss the various ways Orion uses AI, where he sees AI headed in the wealth space — and the lessons he's learned about it so far.
This interview has been lightly edited for clarity and length.
Financial Planning: You mentioned that the independent advisor probably shouldn't be dabbling in AI immediately. Can you dive into why?
Brian McLaughlin: I am very much a believer in choosing your own adventure model, which is: You're going to have to take your own risk. We're all going to have different levels of risk and what we're willing to do and not do. I do believe, though, that everybody should lean in.
There's nothing to fear about it. It's more about learning how to use it, learning how to engage it in a way that makes sense for you.
And there's so many simple ways to engage AI. I use it in my email system to summarize emails. I use it for things like creating actionable items and to-do lists. You can build tasks and calendar items that can automatically be sent and followed up for the client. So, we can use the advantages of these AIs' technologies just to make our lives easier.
FP: How exactly is Orion practicing with AI?
We've embedded AI in three different ways in the Orion system. We have one in Redtail Speak. That was the first one we released, which allows a suggested auto-reply based on the conversational history. It takes previous history items of text messages, and with one click, it will auto suggest the next reply [so] that you can go faster.
READ MORE:
Now, what it does not do is auto send that reply, which, to me, was where that boundary was. You can modify it and then hit send, or you can just hit send. But it's not an auto fire.
And there's another system as well where we do summaries of data, where we actually take account data and look at it for commonalities and model what we think is the next step. That's still experimental, we haven't released that in the wild yet. So again, it's still all done with an opt-in, the user has to want to use this data.
The other place I'm looking at right now is utilizing AI from a compliance lens. So, looking at trades, looking at patterns and trying to understand what the trends are inside the data — it's almost like fraud alerting. But can we do that at a superhuman scale? We're very, very experimental playing with that.
FP: Why haven't we seen AI used in more of a predictive mindset in the financial space?
BM: I think it's just too early. The easy wins in the generative AI work are more so providing narratives, task generation. Those are the easier wins. When you get into this, you get closer and closer to that regulatory risk.
You've got to be careful because we don't have the [regulatory] guidance on what we can and can't do exactly.
But I would say, the [incoming regulation] would probably look at something around personal privacy and setting guardrails up. Because the concerns we hear are about using OpenAI, and there [are] certain levels of it that are more private than others. But for the general consumer version, anything you put in [OpenAI] is going into the public domain.
That, I think, is the biggest "gotcha" that we have. Until there is an isolation model, or we have guidance on it, that's the only option.
FP: How is Orion containing the data through OpenAI?
BM: We use a private license of it, and we scrub it for sensitive data. We don't do anything that involves Social Security numbers or anything with privacy data. None of that is sent in. And we filter it out with data detectors. So that's the only way we can do it today.
I'm really optimistic though. If we can get to some place where every advisory firm has their own LLM model that we're working off of, I have a lot more isolation to make sure my accuracy is better and I'm not getting that hallucination effect.
FP: It's been roughly 18 months since Orion started testing different AI tools. What have you learned?
BM: We actually did some testing around building financial plans. A whole group of us did this a while back. Just inputting really general terms for a typical investor, and comparing the output of a financial plan AI-model suggestion to the software we built for 20 years.
And to be honest, I was surprised. I was actually pretty shocked at how close it came out. That, to me, was really interesting because we're constantly trying to give the advisors more time with their clients. We want them to have more engagement. We take away from that every time we spend four hours working on a plan, eight hours working on those estate documents.
If we can [give them more time with clients] and have that accuracy, that's the win. And so that's what I keep experimenting with.
I think we'll solve it as an industry. We will solve it; it's just a matter of more time. Heck, my [Tesla] drives itself. If my car could drive itself down city streets in Sacramento, I think we can figure out this problem for financial services.
