AI’s Plateau is a Launchpad for Action

The repetitiveness of a topic can get boring.

The lack of relevant, actionable next steps makes topics less desirable.

And the “fad” state seems to have dissipated.

AI is still roaring with excitement, investment, and interest. Many in the industry are struggling to figure out what to do with all this activity.

And the speed at which new information emerges makes it exhausting to keep up.

So, we revert to what we know, our business and serving families.

And we put this other stuff on the shelf until we can do something with it.

At our firm, we also felt that way, but we decided to take action.

With that, we discovered some real, tangible opportunities through our exploration of AI.

Despite the interest in AI remaining high, the hockey stick interest has diminished. Interest has flatlined, albeit at a much-elevated level.

*Showing the trend of search interest for both the term “AI” and the field of “Artificial Intelligence” since December 2021

Despite the leveling of excitement, I urge you not to rest on the impact AI will have on the future of our business. Many reporters are shying away from talking about AI because it is nothing new, and people are saying the same thing, making it boring for news outlets that want to always create new headlines. However, consistent headlines keep us aware and allow us to grasp what is going on. By having consistent headlines, we can slow things down and digest what is happening instead of keeping up with something new all the time. Here is a reminder. Interest in Tesla peaked in 2019, then it flattened. At the time of interest peaking, Tesla shares were trading at a low of $20 per share. The growth and impact of the investment happened during the leveling out of interest. Leveling out is when things become mainstream.

*The red line being the search interest for the term “Tesla”

*Tesla stock performance since the peak of interest in searching for “Tesla”

Companies worldwide are investing billions in figuring out how to leverage AI's advancements.

State Street is leveraging generative AI to digitize and automate approximately 85% of bank loan settlements.

Hershey uses generative AI to optimize the talent profile for specific key jobs.

Kroger also uses generative AI to deliver the appropriate promotional offers and discounts to customers at the right time.

Bloomberg expects spending on generative AI to grow by over 2,500% between now and 2032.

It is not time to wait and see how things will play out. We must dig in and figure out how to leverage such technology.

Let's start with the basics before we get to what we learned in our process.

Much of the news and information has been around the advancements of OpenAI's ChatGPT.

ChatGPT has brought light to the possibilities of AI for three main reasons:

• Accessibility
• Applicability
• Cost

ChatGPT's interface has made it easy for everyday folks like you and me to gain the benefits of AI. Prior to the launch of ChatGPT, you needed the expertise of specialized computer engineers to realize the benefits of AI.

With this accessibility, the masses identified applications they could benefit from using ChatGPT. Whether it be content generation or creative work, ChatGPT's applicability was vast across industries and generations.

On top of this, people can access ChatGPT for free.

This combination led to the explosion of interest and relevance of AI across industries.

At its core, ChatGPT benefits from having access to incredible (actually unimaginable) amounts of information and data. The information is broad in terms of areas of focus but deep within each area.

The technology understands the information, speaks in natural language about that information, and brings together different sets of information to formulate a cohesive response to an unprepared prompt or request from us, the user.

It's incredible what OpenAI did to create ChatGPT.

Since the launch of ChatGPT, many other versions and iterations have emerged. Microsoft has jumped into AI by developing and launching its AI solution, Copilot, and is a major investor and strategic partner of OpenAI.

Microsoft is now allowing all employees who use Microsoft products to use AI in their daily work with Copilot as part of the Microsoft software package.

In a recent survey from some early adopters of Copilot, we started to see AI's impact on a broad set of industries. It shows the general value that is immediately available for any business. The data shows that 60% of Fortune 500 companies are adopting Copilot.

Users of Copilot responded in this way:

• 70% felt they were more productive using Copilot.
• 68% believed Copilot improved their quality of work.
• 77% of employees want to go back to work using Copilot.
• 71% felt rescued from mundane tasks.
• 64% felt relief from the burden of email.

We continue to see the more apparent reality that AI is not here to replace us. AI will be used as an agent to augment us in our jobs.

The standard view with new technology is to jump to the fear side of the ledger. But with time, we will continue to move more to the partner side of the ledger. Technology should partner with us, not replace us.

In developing our own AI agent, we stumbled upon the concept of a Hierarchical Multi-Agent Model (a pretty techy phrase that will surely impress your buddies at your next cocktail party).

Hierarchical Multi-Agent Models are the next step in AI and will push AI over the edge for its impact on each industry.

The simple way to understand this is by relating it to today's business structure.

In business, we have a hierarchical organization. The CEO is focused on pushing the overall vision, and a set of C-Suite executives are focused on functional verticals. These executives take the firm's vision and relate it to the functional vertical.

Within each vertical, there are layers and different subject matter experts. A well-run organization has great communication and processes implemented to ensure that information from one subject matter expert gets to another in an effort to help the next person with their task based on the relayed information.

And this continues until a significant initiative is executed and pushed forward.

That is what is now happening in AI.

ChatGPT is built to be the organization and has information on every role and responsibility. The hierarchical multi-agent model has specialized agents and leverages the power of ChatGPT to communicate effectively between agents.

A boss AI agent oversees multiple specialized AI agents, receiving requests from humans and orchestrating each specialized AI agent to provide the necessary information.

For example, a multi-agent model in healthcare could improve patient care and streamline hospital operations.

A boss agent would oversee the following agents that are narrow and specific to needs within healthcare.

• Patient agent: trained to monitor patient health metrics and communicate with healthcare providers
• Doctor agent: trained to analyze patient data, generate treatment recommendations, and coordinate with other agents to manage patient care.
• Nurse agent: Assists with monitoring patients, administers medications, and provides the doctor agent with feedback.
• Administrative agent: Manages hospital resources, scheduling, and compliance

These agents use ChatGPT to help with natural language processing (understanding what we say and interpreting it into computer language), data analysis, and communication (putting thoughts together and responding to us in an understandable way).

Here is an example of these agents working together.

The patient agent continuously monitors health metrics and has a deep understanding of the appropriate levels. Suddenly, they identify an anomaly in the health metrics and send it to the doctor agent.

The doctor agent uses ChatGPT to understand the information, analyze patient data, and generate a treatment recommendation, which is sent to the nurse agent.

The nurse agent administers the treatment plan and provides feedback on its effectiveness to the doctor, who adjusts the treatment plan as necessary.

Why is this process more beneficial?

This process allows agents to be hyper-specialized and trained, with communication between agents enabling AI to solve specific and nuanced challenges.

Instead of just using ChatGPT, which has expansive knowledge but not a specific niche knowledge set, this model leverages ChatGPT's benefits and pairs it with industry-specific knowledge to solve and execute complex tasks.

That is game-changing.

And we stumbled upon this when building a proof of concept.

In our scenario, we were looking to leverage AI to help advisors rebalance and reallocate portfolios more seamlessly, leveraging the insights from our investment committee.

We built a hierarchical multi-agent model with a boss agent and then three sub-agents:

• Portfolio agent - an expert at reading and interpreting the information from a PDF portfolio statement.
• Financial agent - an expert who knows our current market and economic views.
• Client agent - an expert who understands the history of the client's relationship.

This allowed us to build a model in which the boss agent could coordinate with the three agents to get back a proposed reallocation that fits the desired allocation target, aligns with the current views of the investment committee, and ensures we account for the client's needs.

This response provided specific investments that our investment committee had proposed. It also provided recommendations on areas to trim within the current portfolio based on both the client's goals and the current investment committee's views.

Although not launched to production, this proof of concept allowed us to uncover the hierarchical multi-agent model, solve a complex business challenge, and demonstrate the direct application power of AI in our business.

We built two additional scenarios using the hierarchical multi-agent model and saw similar success from this.

Only some people should be expected to build their own AI proof of concept.

And many struggle with where to start.

The goal of this post was to remind us:

• AI is not going anywhere.
• Instead of trying to predict how AI will impact us, start gaining greater awareness and keep learning about it.
• Don't get lulled to sleep by the same headlines. Dive deeper, connect with others, and explore other industries to learn how they are incorporating AI.

The moment we wait for AI to be brought to us will be when we instantly start behind the eight ball.

AI will change our business. It's a matter of when.

Prepare by gaining a deep understanding of what is happening in the space today.