OpenNashPodcasts

Person

Andrew Sellers

Andrew Sellers is a Confluent speaker who discussed AI and data governance in financial services on a podcast episode about what data governance means for the industry. He described financial firms as stewards of capital and, ultimately, risk managers, and said he has spent about 20 years researching AI or bringing AI-enabled applications to market.

Sellers argued that data governance is the best investment financial firms can make, and that great data with off-the-shelf models has always beaten limited data with cutting-edge models. He also said large language models are not yet reasoning agents, that they work best when they augment people, and that he has not met anyone who lets one handle data governance without human oversight.

1 episodes · 19 quoted claims

Andrew Sellers describes financial services firms as stewards of capital and therefore ultimately risk managers.

“In financial services, ultimately what we are is we're stewards of capital. And that means that we're ultimately risk managers.”

AI in Financial Services · 0:01:43

Andrew Sellers says that in his career, great data with off-the-shelf models has always outperformed limited data with cutting-edge models.

“One constant theme of my career is really great data with off the shelf models has always outperformed, you know, we'll say limited data with cutting edge, sophisticated models.”

AI in Financial Services · 0:02:55

Andrew Sellers argues the best investment financial firms can make is data governance, meaning they know where their data comes from.

“But the best investments they can make is in data governance, making sure that the provenance of their data is sound. They know where it comes from.”

AI in Financial Services · 0:02:55

Andrew Sellers says data governance matters in financial services partly because regulatory frameworks require transparency and accountability in decision-making.

“governance in FinServe is also very important for regulatory frameworks because the decision-making typically requires some transparency and accountability.”

AI in Financial Services · 0:05:22

Andrew Sellers says a $100 million innovation project creates expectations so large that, trying to be all things to all people, it will probably satisfy no one.

“The problem with spending $100 million on an innovative project is the expectations that it creates are so much that if it tries to be all things to all people, it'll probably satisfy no one.”

AI in Financial Services · 0:06:10

Andrew Sellers says document recognition and summarization have been largely commoditized by current AI technologies.

“That's something as a capability that I would say has been pretty well commoditized by these technologies.”

AI in Financial Services · 0:06:32

Andrew Sellers recommends using document-processing AI internally to augment data entry staff rather than replace them.

“So having document processing, supplant maybe some of the data entry folks that you have and not to replace them, but to augment them.”

AI in Financial Services · 0:06:54

Andrew Sellers says LLM investments have tended to succeed more when they augment human operators than when they try to automate them away.

“the investments have tended to be more successful when we're augmenting human operators, rather than trying to automate them away.”

AI in Financial Services · 0:07:13

Andrew Sellers says he has spent about 20 years researching AI or bringing AI-enabled applications to market.

“I've been, in some sense, either researching artificial intelligence or bringing AI-enabled applications to market in some capacity for 20 years now.”

AI in Financial Services · 0:10:24

Andrew Sellers explains that large language models are foundation models: big general-purpose models built with large investment and reused for many tasks.

“I think with these foundation models, which is what these LLMs are, these big things, we make big investments and we produce something general, and then we can use it for lots of things.”

AI in Financial Services · 0:10:54

Andrew Sellers says transparent insurance underwriting models often use Monte Carlo simulation, such as running a thousand years of simulation for a given peril to see the distribution of loss.

“it's a lot of Monte Carlo stuff. I can run, you know, a thousand, you know, years of simulation for a given peril, and I can see the distribution of loss.”

AI in Financial Services · 0:11:50

Andrew Sellers says LLMs are not yet reasoning agents and cannot yet replace normal predictive models.

“these things are still really good, but they're not yet reasoning agents. They're not yet to the point where they can supplant normal predictive models”

AI in Financial Services · 0:12:12

Andrew Sellers says many Confluent customers, particularly in insurance, successfully pair LLMs with more traditional models.

“one pattern we've seen a lot of our customers use with some success, particularly in insurance and other places, is to actually pair LLMs with more traditional models.”

AI in Financial Services · 0:12:12

Andrew Sellers says LLMs are stochastic, so unless the random seed is exposed their outputs are not even replicable.

“by their nature, they're stochastic, which means that if you don't kind of expose the seed, the outputs aren't even replicable.”

AI in Financial Services · 0:13:14

Andrew Sellers says the best protection against LLM hallucinations, via retrieval augmented generation, is to supply highly contextualized data.

“The best thing that we can do to protect ourselves from hallucinations is to supply highly contextualized data.”

AI in Financial Services · 0:13:44

Andrew Sellers says Confluent presented data governance capabilities augmented with generative AI in a keynote at its last Current event.

“talked a bit about some of the things we're bringing to market and really exciting for us in terms of data governance capabilities that are augmented with generative AI.”

AI in Financial Services · 0:16:28

Andrew Sellers says much of data governance is cataloging, meaning generating metadata (data about data) such as what a data set describes and who owns it, which LLMs can help with.

“a lot of data governance ends up being this cataloging activity of generating metadata, data about data, data that describes my data set, who owns it,”

AI in Financial Services · 0:17:18

Andrew Sellers says financial services firms are increasingly adopting federated data governance with domain-driven ownership, where the people closest to the data own it.

“are more and more adopting what we call a federated governance model, but with domain-driven ownership that basically the people closest to the data that are producing the data ultimately own it”

AI in Financial Services · 0:17:51

Andrew Sellers says he has not met anyone who lets an LLM handle data governance without human oversight.

“I've not met anyone yet that's crazy enough to just sort of let the LLM have at it without any kind of human oversight.”

AI in Financial Services · 0:18:14