Alan Gurung, CEO and co-founder of SIFA, a generative AI startup based in London, has a singular goal for how his technology will help his financial advisor customers.
“Our hope is that by the end of this year our advisors will be able to look after three times the number of clients they do today,” said Gurung, and do so without additional stress or long hours.
Gurung and his co-founder, Roshan Tamil Sellvan (an engineer and serial entrepreneur), started SIFA, which stands for super independent financial advisor, along with a small team of two other machine learning engineers and two student interns.
Gurung, for his part, worked as an advisor and partner for the huge U.K. advisory firm St. James’s Place and sold his book of business to start SIFA. (Sellvan had just done the same with his own startup.)
Their new technology was built specifically to automate ongoing communications between advisors and their clients, a pain point Gurung specifically struggled with during his time as an active advisor.
“I was constantly frustrated with how best to communicate with my clients, what would matter most to them, how often and how to keep ahead of their needs and life events,” he said.
SIFA was in stealth for four months before officially launching in January. According to Gurung, 210 advisory firms in the U.K. and U.S. already have adopted it.
He credits tech-savvy XYPN advisors with discovering and bringing the application stateside.
“We built this originally for the U.K. advisor community but found that American advisors have been a lot more responsive in terms of feedback and so, we are going where the demand is, and it’s in the States,” said Gurung.
The AUM, on average, of the firms using SIFA is $100 million to $150 million, he said, but added that there are five user-firms that have more than $1 billion in AUM.
“We see the biggest advantage, though for the smaller firms right now,” he said; the technology can serve as a sort of paraplanner or project manager for a small practice.
Data is brought in from a variety of document types, whether PDFs, Microsoft Word documents or text files—that contain both structured data (like client account statements) and unstructured data. It’s also common for advisors to provide the application with iPhone or Android voice memo recordings of client meetings, or post-meeting notes.
“Sit back and relax, while I work my magic,” a popup tells the user while SIFA is ingesting information or performing another task, such as answering an advisor’s hypothetical question about their client “Dave”: “Who are the beneficiaries in Dave’s will?”
The startup already has integrations with advisor CRM application Wealthbox and financial planning software eMoney; further integrations with two other popular advisor technology products are in the works, but Gurung said he could not yet make those names public.
The SIFA dashboard is made up of tiles, including “Insights”—things the AI has uncovered for the advisor as possible talking points or ideas for a particular client’s To-Do list.
For example, “Dave has a dog that is having surgery next month …” is an insight pulled out of the last meeting with “Dave,” and the advisor can assign SIFA the “Task” of writing to “Dave” preemptively.
With the application’s “Email Assistant,” a dropdown lets you select from “Friendly,” “Creative,” “Formal,” “Funny” or “Professional” and a length of “Short,” “Medium” or “Long.”
These categories weren’t pulled from thin air. Gurung said that they came from trial and error and ongoing tuning, talking to advisors, some of whom said that their communications were coming out as too stiff or overly friendly, etc.
And SIFA also learns your own tone from both what it has ingested from previous communications, as well as how you customize ongoing correspondence.
“What we find now is that once you have created or customized and sent 50 emails, [SIFA] is eerily similar to your own tone of voice,” Gurung said.
Unlike many other generative AI-focused startups, Gurung said SIFA’s technology is built on his team’s own machine learning, large language models and algorithms.
Unlike OpenAI, which trains on massive amounts of centralized data, SIFA’s models, he said, are trained on each individual financial advisor’s client data and other information specific to their firm, which provides deeper personalization to each client. None of this collected data is shared with any third parties.
“What we are working toward is that every single day advisors come on here and go ‘these are my tasks for today’; we believe with generative AI you can do that and scale to have not only more personal relationships but allow you ultimately to work with 10 times the clients.”
