Can AI Solve the Customer Service Problem?

Customer service has a familiar problem: everyone wants immediate, knowledgeable support, but delivering that experience at scale is expensive and difficult.

Call centers face long queues. After-hours support is limited. Customers are often forced to navigate FAQ pages for relatively simple questions, while highly trained representatives spend much of their day answering repetitive requests.

In the inaugural episode of AI Discovery, host Patrick Lothian explores whether artificial intelligence can fundamentally change that equation.

He is joined by our own Jim Swartwout, CEO of Prosperum Fintech Holdings, and Dennis Evans, VP of Customer Operations at NinjaTrader, for a conversation about how their organizations are already deploying AI agents in financial services, and where the technology could go next.

Bringing Full-Service Support to the Digital Brokerage

For Swartwout, the opportunity starts with recreating something that has largely disappeared from modern discount brokerage: personalized, knowledgeable service.

Traditional full-service brokers knew their customers, their portfolios, goals, experience levels and investment strategies. Modern call centers offer tremendous scale, but customers may speak with a different representative every time they call.

Prosperum’s answer is TacoAI, which uses voice and chat agents trained on years of broker conversations, policies, procedures and FAQs.

Today, those agents can handle questions ranging from how to fund an account to how a particular options strategy works, including during nights, weekends and holidays. The longer-term vision is more ambitious: enabling customers to complete routine account tasks and eventually place customer-directed trades conversationally.

Swartwout sees this less as replacing brokers than as the next evolution of brokerage technology. Early in his career, brokers spent hours reading stock quotes to customers over the phone because there was no practical alternative. The internet automated that job. AI could now automate another layer of repetitive service while leaving experienced brokers available for situations where their expertise matters most.

Scaling Customer Service Without Sacrificing Service

At NinjaTrader, the initial challenge was scale.

As the company grew, an AI agent offered something a traditional support organization could not easily replicate: the ability to handle many conversations simultaneously.

But Dennis Evans argues that the more important benefit may be self-service.

Many customers don’t necessarily want to speak with a representative. They simply want an accurate answer immediately. Instead of directing someone to an article or FAQ page, an AI agent can understand the question and provide the relevant information conversationally.

That is particularly valuable during onboarding, when new traders often encounter many of the same questions.

The result isn’t necessarily fewer humans. Instead, AI can handle the “low-hanging fruit” while experienced representatives concentrate on complicated situations that genuinely require human judgment and attention.

The Next Step: AI That Understands Your Account

The conversation also highlights a much larger opportunity: moving from AI that simply answers questions to AI that understands the context surrounding an individual customer.

Prosperum envisions an agent that could greet a returning customer and immediately surface relevant information: recent dividends, interest received, changes in the account or an options position approaching expiration.

Combined with research and analytics capabilities, the same conversational interface could potentially become a central point for understanding both a customer’s portfolio and the broader market.

That changes the role of the chatbot considerably.

Instead of being a digital FAQ, it starts to resemble an always-available financial interface.

Trust, Accuracy and the Hallucination Problem

Of course, financial services leaves little room for an AI system that confidently gives customers incorrect information.

Evans describes an approach built around controlled knowledge sources rather than unrestricted internet generation. NinjaTrader uses its own product and public-facing documentation as the agent’s knowledge base, while customer-service experts continuously review conversations and improve the system.

When the team identifies situations where an AI agent might make assumptions, it can create playbooks, introduce additional data requirements or force the system to ask clarifying questions.

The objective is to move interactions from nondeterministic guessing toward deterministic workflows, and to escalate to a human when the system doesn’t have enough information to answer reliably.

Interestingly, AI also changes the work performed by customer-service teams. Subject-matter experts who previously spent their time handling repetitive interactions can instead become QA analysts, reviewing AI conversations and improving the system.

What Happens When AI Can Take Action?

Answering questions is one thing. Executing transactions is another.

Swartwout argues that conversational trading can ultimately follow safeguards similar to those customers already encounter online. A customer specifies the trade, the system repeats the details, the customer confirms them and the transaction is submitted.

Importantly, that’s different from giving an autonomous agent broad authority to select investments and trade independently.

The conversation also addresses the regulatory implications. Customer instructions and conversations can be recorded and archived, just as communications with human brokers are today.

From Screens to Agents

Perhaps the most interesting part of the discussion is what happens beyond customer support.

Evans sees the possibility of another major transition in trading technology.

Financial markets once moved from physical trading pits to electronic screens. The next transition could increasingly move interaction from screens to agents.

Rather than spending the entire day clicking through interfaces, traders could eventually communicate their strategies, levels and parameters to intelligent agents. Voice could become an increasingly important interface as well: a trader might simply open an app and ask for an analysis of the day’s markets rather than manually navigating through charts and data.

AI could also become an educational layer for retail traders, identifying patterns in their own behavior and surfacing lessons earlier—potentially helping people learn without having to discover every mistake through expensive trial and error.

Better AI Could Mean Better Human Service

There is an understandable tendency to frame AI customer service as a story about replacing people.

This conversation presents a more nuanced possibility.

If AI can instantly resolve routine questions, customers with genuinely difficult or time-sensitive problems may reach experienced representatives faster. Those representatives, in turn, can spend more time actually solving complicated problems rather than racing through a queue.

The end state may therefore combine two things that traditionally seemed difficult to reconcile: instant service at massive scale and more focused human attention when it actually matters.

And financial services may only be one example.

The underlying model, AI agents that understand institutional knowledge, know the customer, complete routine tasks and intelligently escalate exceptions, could reshape customer service across industries.

The full conversation goes considerably deeper into how TacoAI and NinjaTrader are approaching these systems today, the safeguards required to make them trustworthy, and what agentic trading could eventually look like.

Watch the full episode of AI Discovery on YouTube to hear Patrick Lothian’s complete conversation with Jim Swartwout and Dennis Evans and subscribe to John Lothian News for upcoming episodes exploring how AI is changing the way people work, trade and interact with technology.