Technology Tools for Today (T3) is holding its annual T3 Technology Conference at the Hyatt Regency New Orleans March 9 to 12 with the theme: “From Noise to Knowledge: The FinTech Playbook for 2026 and Beyond.” T3 began the event by hosting its first “AI University,” a day of workshops and learning about AI on March 9.
Advisors Building AI Agents
In his presentation titled “From Chatbot to Coworker: Building AI Agents for Advisor Workflows,” Craig Iskowitz, CEO and Founder of Ezra Group, demonstrated how advisors can create AI agents using standard AI large language model (LLM) interfaces, providing as examples a custom integration of wealthtech tool provider PreciseFP with CRM provider Wealthbox (separate from the standard integration these firms provide), as well as a gaps detector in Wealthbox.
Putting aside coding agent Claude Code and the specialized workplace tools of Claude Cowork, Iskowitz opted to use the standard Claude interface to demonstrate what can be accomplished by a single advisor who is not a coding expert.
After walking the audience through how he built and integrated the agents, as well as how they operate, Iskowitz noted that advisors who build their own agents must prepare to debug them on an ongoing basis.
Responding to an audience question, Iskowitz explained that “data lakes,” which are repositories for data, including unstructured data such as meeting notes, photos or audio files, will become more commonplace in the future.
He added that the industry currently struggles with the related issue of who owns the advisor’s data, which is often owned by vendors, generating friction if the advisor wishes to switch vendors. In time, he expects that more advisors will own their data and hold it in data lakes.
Discussing how AI will change the industry, Iskowitz contrasted today’s AI to the robo-advisors of several years ago. Robo-advisors used the same software that a human would use to derive results, while AI is truly robotic, able to think and build financial plans. Though AI financial advisors are in early days, he said these will catch up to human abilities soon, so advisors must think more broadly about how to provide services in ways AI can’t, such as taxes or financial coaching. “Whatever it can do, we have to do more.”
Cold Outbound Is Getting Warmer
In her presentation titled “The AI Growth Playbook: From Signal to Client,” FINNY AI President Victoria Toli presented a short history of AI and discussed how it can be leveraged for prospecting.
She explained that AI proceeded from rules-based computing through machine learning and deep learning, until the “transformer” was developed in 2017, a device to predict the next word in a sentence based on wide sets of training data. At that point, AI “moved from predictor to collaborator.” The next phase, which we are currently in, is agentic AI, in which AI can hold memory and use tools to act by itself for the user.
Toli said that cold outbound prospecting can be effective if an advisor finds the right people (people who they can serve in a relationship) at the right time (often life changes such as layoffs, marriage, divorce, selling a home or having children) with the right method (so the prospect feels “seen, not sold to”).
She described a method by which an advisor can build their own AI agent to scan the news for signals, find people impacted by the events signaled in the news and produce personal outreach sequences, breaking down the process into five sequential agents: signal detection, looking up prospects, scoring prospects, sequencing outreach emails and outputting both prospects lists and email templates.
She also explained the limitations of this approach, with limited coverage when detecting signals, less personalization of outreach and suboptimal content because of missing nuance, as well as hallucinations which can arise from giving the AI system too much freedom rather than precise instruction.
AI Uses And Common Problems
In a session titled “From Prospecting to Annual Review: How AI is Transforming Wealth Management,” John O’Connell, Founder and CEO of The Oasis Group, explored use cases for AI as well as ways to manage common problems.
In response to recent statements circulating on social media that “prompting is dead,” O’Connell countered, “Prompting matters significantly.” He explained that good prompts are required to generate quality outputs.
O’Connell explained the difference between primary models that are based on large sets of training data – such as ChatGPT, Claude, Gemini and Llama – and secondary models like CoPilot, Jasper and Perplexity that add functions and nuance for specific uses onto primary models.
Various sources of bias can be introduced, said O’Connell, by applying non-heterogenous data to a specific use case. For example, if an advisor’s clients are typically ultra-high net worth, then using that data set to generate results for a less wealthy high net worth client may skew the results.
He presented a list of measures advisors can take to reduce AI hallucinations, including good data governance and high data quality; testing new AI models, which requires ongoing, dedicated resources from an advisory team to ensure new models still meet the firm’s needs, such as compliance needs, as the models are rolled out; transparency in decision-making; regular monitoring and adjustment; a compliance framework including an AI use policy for employees; and a “human in the loop.”
O’Connell led the audience through multiple use cases, including text-to-speech conversion for client communications, a prospect intake form and a virtual advisory board for employees to ask questions.
Julius Buchanan, Editor in Chief at Wealth Solutions Report, can be reached at julius.buchanan@wealthsolutionsreport.com.
