# Anthropic: Accelerating frontier AI in Financial Services. — Belinda Neal, Keon Shahab — session 2026-10-13T13:18:00.000Z → 2026-10-13T13:36:00.000Z

_56 transcript lines · 35 slides_

## Transcript

In the vault. And so this, for us, is the key lesson. The vault enables compounding because it provides the right evidence, the right internal and external information at the fingertips of our investors and agents, and it automatically captures the synthesis that they generate and puts it back in the corpus for the next turn. And that makes us go around the circle faster and better. AND SO WHEN WE THINK BACK TO OUR, THE QUESTION WE BEGAN WITH, WHAT IT TAKES TO MAKE AN AGENT WORK The right context and the right understanding of what's going on in your world. And so our question for you is, does your AI know what's going on in the world? And does it remember what your firm has learned about it? Because we are teaching ours too. And if these sounds like an sound like an exciting challenge, we'd love to talk. Thank you. Coming up next, we're delighted to welcome two speakers from Anthropic. Please put your hands together for Belinda Neill, managing director of financial services, and Kian Shahab, applied AI architect. Good morning. I'm Belinda, and this is Keyon, and we're from Entropic. We work with financial services firms. As Keyon gets set up here, we'll start with a little bit of background. My name is Belinda Neill, and I'm the Managing Director of Financial Services at Entropic. Previously to joining Anthropic this year, I worked at Goldman Sachs for nineteen years across engineering, global markets, and equity research. My role at at Anthropic sits at the confluence of our industry, our product, our research. And our go to market teams, really in accelerating the adoption of AI within financial services. Keon. Good morning, everyone. Thanks for being here. My name is Keon Shahab. I'm an applied AI architect at Anthropic. Prior to Anthropic, I spent five years at Databricks helping financial services companies modernize their data and AI architecture. So we're excited to speak to you this morning. And our talk is going to cover three ideas. Firstly, the frontier of intelligence. How is this room taking the frontier models and applying In ways that create alpha or edge within your firm. The second is the frontier of adoption. How are you scaling what you've built based on the foundation of a few to many? And then last, how are we applying it to financial services? We're going to bring an example in here about how we've been innovating in the wealth industry in Claude Financial Advisors, and that paradigm compounds potentially to other areas of financial services as well. So let's get started and talk about the frontier of intelligence. So every so often, a model Crosses a threshold. And when it does, a way of working changes. We first saw this encoding last year with Opus 4.5 and Claude Code. Many of you in the room are using these tools, and you've gone from auto complete encoding to managing fleets of agents where you're really driving and accelerating a multiple effect of how you're doing coding. We also saw this threshold cross this year in cyber with Claude Mythos preview, and many of you are also spending time on the cyber programs in your firms. Next, finance. We've started now in finance with early agents and bringing Claude into the products and services that you use day to day. One example here is Citadel told us at our briefing event in May that their investment professionals started to use Claude for Excel, and it gives them a step change function in how they can test, build, and trust their work. We're really excited for the potential of Claude within the tools and the use cases that this room is building. But here's the challenge. Intelligence is ahead of adoption. What a model can do on its own is doubling every four months. PwC published early this year that the capabilities of AI are going to three quarters of companies. Sorry, Sorry, three quarters of the economic gains of AI are going to a fifth of companies. This is what we call the diffusion gap. If you're not building close enough to the curve, there's opportunity for you to take the frontier and be able to apply it in your workflows. We work with firms Across financial services across our model family, most of which hopefully this room is familiar with. We have Haiku for speed and volume, Sonnet for everyday work, Opus as the daily driver for financial services, complex tasks, and agentic coding, And Fable, our frontier model that can effectively be used for the hardest work in your firms. One example here from a millennium on our most recent model blog was a PM used Fable to find a piece of code that had a And it was unexplained in five years. Fable was able to find it. We're starting to see our models get used in this industry in new and creative ways, and we believe the frontier of intelligence is just getting started. Every model release gets smarter, cheaper, and safer, so it's really important to take a minute to call out how the model frontier is continuing to develop here. Firstly, the intelligence is going up. Now a model can work for sixteen plus hours on its own. This is up from just minutes in 2023. In terms of cost per task, it's going down. The most recent Opus 5.5... Opus 5.5 model was 40% more cost efficient than Opus five. You can also see those impact metrics for Haiku, Sonnet, and Fable. And lastly, safety. This is something we care a lot about at Anthropic. In the most recent model of Fable 5.1, only one percent of prompt injection attacks were successful. This is relative to other frontier models in the region of nine to fifty three percent. So we're really investing in safety alongside capability anthropic. And the banks know this. The last most recent Evident AI 2026 report that came out last week showed us that banks are 26% ahead of their AI maturity scores. We're really investing in scaling in applying this technology to our industry. A few use cases to call out from this report, which are publicly citable use cases that have been reported, is firstly on areas like fraud. One bank found that with auto loans, they were able to reduce $120,000,000 of potential fraud in 12 Months. In another area, in post earnings research, a draft of a report was able to be generated in fifteen minutes relative to forty five minutes. And you can see other use cases here and more broadly across the industry as we start to talk about the value of AI and the impact we're generating on our business. So where are we? In 2023, we built chatbots. Everyone was asking a question and going backwards and forwards with the model. In 2024, we brought pilots to firms, and we started to get more knowledge bases and Data into these AI systems. From 2025 to now, we're now building agents. And in the most case in this room, we're building them hand by hand, one by one. And our opportunity here to scale is not the model. Our opportunity here to scale is building the foundation for agents that can go from one to many. So the next part of the talk, we wanted to spend time focusing on how you can build the agent foundation that will allow you to scale the frontier of adoption. So over to you, Keyon. Amazing. Thank you, Belinda. So Belinda, as you I've just said there's no better time than now to start experimenting and building with agents. For this portion of the talk, we want to talk a little bit about how you can go from that experimentation phase towards productionizing and scaling agents. But first, some observations that we've seen across the industry. A year ago, we could think about this as the year of exploration where engineers were working on a handful of use cases. They were in POC and pilot mode and they took their learnings and transition that into this year where We've seen this be the Age of Production. So we've seen these handful of Agents go to production. A lot of them have been built by hand. What I think will happen next year is we'll go into the Age of Scale, where we'll have many Agents built on one solidified foundation. And as Belinda just mentioned, the models are now the easiest part of the equation here. We know that the models can handle high quality tasks, and we're confident about that. But what we want to talk About here is building this five layer framework for scaling and productionizing your agents. Starting from the bottom, we first want to look at what the agent can do. So we can give it context for our business use cases via connectors so that it can reach our systems and data. Then we want to give it context for our actual business rules and processes via skills. Once that's done, we want to think about building the trust layer for the agents. We can think about that in three general steps who the agent is As with an identity and permissions layer, how we can sandbox and determine the approval layer of the agent so that someone is signing off before it acts, and lastly, how we know the agent will still work once we adjust any variables over time. So let's look at each of these five. Perfect. So if you started building agents already, you may think of giving those agents access to your underlying systems and data manually, wiring them by hand to a handful of different Source systems. And that could work for maybe five agents and five systems. Then you'll have 25 systems to maintain and manage. But as you think about scaling up your agents into new specialized various tasks, you'll want to have those agents work off of one secure connectivity layer, which is MCP. That'll make it a lot easier for you to build at scale with one repeatable framework. Next, Let's say now we have the Agent have context to our underlying business data and systems. Now we want to give it more context into our actual business rules and processes. And that's done through skills. And this is where the SMEs that really do the work and understand the process and the work come into the picture. And from here on, we're going to look at a agent use case for a trade break investigation agent. And we're going to follow this across for the next Steps. In this case, a trade break agent is meant to look at two records of a single trade that don't match. And it's here you can see steps one through three a little bit of a simplified version. But this is the business context that an SME provides. And we think about skills as living, breathing, versioned code, essentially, even though they're just plain text. They're meant to be dynamic. They're meant to be updated frequently. All right. So now we have our Context layer for the Agent. We have the Agent having access to our source systems and data. We have the Agent that understands how to go from a generalized model to a specialized Agent with skills. Now we want to talk about building the trust layer with that agent. And first, let's start with the identity and permissions layer. So here you can see the identity of the agent. We're going to assign it. It's going to be the trade break investigator. We're going to give it a role with permissions called ops analyst. And you can see here once Agent needs to go and reach our source systems like our order data and our market data. It'll have permission to read but if it tries to go reach for our CRM data, it'll be denied and we will do that on purpose so that won't overreach for sources of data that it shouldn't have access to And lastly, we don't necessarily want this agent to perform on behalf of us. We want it to just propose new amendments to our orders data, not actually make those amendments. And that's where we'll look at on the next slide how we can keep the human in the loop here. All right. So this is kind of the layer of who signs off before the Agent acts. Right? Many of you have heard the term sandbox. We can use that terminology here as well for putting an approval layer around the Agent. So that you can see here the four step process. The agent will actually go and propose a booking amendment. It'll actually try to proceed through the process with the context that it has and the source systems and data that it has access to. But then it'll look For a human and SME someone in the loop in the process to approve. Right? Once they approve, then that amendment will be made and the change will be applied. This is the layer where you can also put guardrails, deterministic guardrails called hooks around your Agent so that it's never overreaching beyond what you would like it to do within its confines. All right. So now, we have a full Agent. Let's say it's in production. It's working great. And Let's say that we want to make an adjustment to this Agent. Let's say that a new model is shipped or we want to make an adjustment to the business context layer with the skills. How do we actually know if these adjustments are improving the Agent or if they're causing a regression. So that's our evals layer. This is where your SMEs, your business users who are in the loop, who understand the process, help determine a set of test cases to test the agent against. With a benchmark or a baseline so that once you apply any change to any variable, you'll be able to determine whether that variable change actually improved the behavior of the Agent or if it caused it to regress, as you'll see here in some of the examples. All right. So hopefully you've seen by building once with a layer and with a foundational layer that you trust you can work at speed to deploy many agents. At scale. For us, financial services and regulated industries doesn't mean slow. It means that we're seeing a lot of innovation from the financial sector. And we're seeing it go hand in hand with security and safety as well. And those three layers that we just walked through who the agent is acting as, who signs off, and how you still know that agent works are the key components of building a safe and secure agent framework. And once you have this Framework that you've built you can use it to apply to any new agent that you want to build for any specialty use case so you can see at the top you'll have your foundation and then at the bottom you'll see where you can actually bring in more specialty along the way as you continue to scale. That specialty is worked in combination with the business and engineering to make sure that the proper business rules and context are being integrated into the agent workflow. And you might be wondering, we haven't mentioned anything like Cloud Code yet. Well, everything that we've talked about so far has, let's say, four different doors that you can access it from. So a business person can access it from Cloud Cowork, an engineer can access it from Cloud Code, or you can take this to the next step and give the agent more work to do in your own UI environment via things like managed agents or agent SDK. Now so far we've talked about building a secure trustable agent Foundation framework. Now let's talk a little bit about the operating model. Some observations that we've seen, especially in financial services, is that the business is now becoming closer to engineering. And the way that we're observing this is that the business will directly work with the agent that engineering has built. They'll have direct capabilities to correct that agent. Or amend it proposed adjustments and that's where engineering is in the loop as well. Once you see the list of proposed adjustments you can solidify that as a scale change after you have reviewed and tested all that within your eval framework. And then anything new that you ship goes right back to the business. And they'll go through the same loop again to continuously improve the overall agent and the process. All right. So So far, we've talked about building the Agent framework in a way that you can build once for scale. And we've talked about how you can turn a generalized, highly intelligent model into a specialized Agent. Now I want to hand it off to Belinda to talk a little bit more about what we're seeing in the industry by building some of the foundations ourselves for the next step here, which is Claude for financial advisors. Thank you, Kean. So what you've just seen is us building a foundation once in a firm. Now I want to talk about how we could take that same approach to our industry. This foundation we've built implored for financial advisors, and whilst you might not all in the room sit in the wealth or the advisor space, I want you to take this as a pattern and an opportunity to start to think about how we can actually coordinate at an industry level. So in September, we launched Clor to Financial Advisors, the orchestrator across the stack of an adviser. This was built on the foundation that advisers only spend approximately 20% of their time with clients. We wanted to find a way where advisors could spend more time with clients and more time growing their franchise. So what did we do? We went through the agent foundation and found ways to scale. At layer one, we bought connectors in. The industry system, the deterministic numbers, and the scaled system that advisors trust, we brought into the ecosystem.

## Slides

### 00:00:27

### 3 Reactive to proactive
# Summary and evidence, ready for action.

[Screenshot of an AI-driven research summary tool titled "AI impact on currencies" for "KOREA & T

### 00:00:49

# The Vault enables compounding.

- **Gather**: The right external and internal information, for humans and agents.
- **Learn**: Synthesis is contributed back to the corpus.

[Circular diagram titled "Reasoning Engine" showing a data flow cycle with stages: Gather, Hypothesis, Investigation, Compound Memory, Synthesis, Learning, and Structured Data (Vault, Market History).]

### 00:01:21

# Does your AI know what's going on in the world?
## And does it remember what your firm learned about it?

We're teaching ours to.

### Vault: When Fifty Years of Structured Data Isn't Enough

Aaron Linsky
Head of Product Engineering

Sam Green
Team Lead, Applied AI - AIA Labs

### 00:02:19

# Who you are hearing from

- **Belinda Neal**
  - Managing Director, Financial Services, Anthropic
  - *Prior: 19+ years at Goldman Sachs across Engineering, Global Markets and Equity Research*

- **Keon Shahab**
  - Applied AI Architect, Anthropic
  - *Prior: 5 years at Databricks, Lead Solutions Architect for Financial Services*

[Circular headshot of Belinda Neal]
[Circular headshot of Keon Shahab]

### 00:02:50

# Who you are hearing from

## Belinda Neal
Managing Director, Financial Services, Anthropic
Prior: 19+ years at Goldman Sachs across Engineering, Global Markets and Equity Research

## Keon Shahab
Applied AI Architect, Anthropic
Prior: 5 years at Databricks, Lead Solutions Architect for Financial Services

[Profile picture of Belinda Neal]
[Profile picture of Keon Shahab]

### 00:03:20

# Taking the frontier to the industry

### 1
#### THE FRONTIER OF INTELLIGENCE
**Applying intelligence**
What the best model can do today. An alpha opportunity.

### 2
#### THE FRONTIER OF ADOPTION
**Scaling agents**
Building the agent foundation and the operating model. A scale opportunity.

### 3
#### APPLIED TO FINANCIAL SERVICES
**Compounding impact**
The same foundation, built once for all of wealth. An industry opportunity.

This talk: how to take on the frontier opportunity across these three

## Keynote
- Belinda Neal
  Managing Director, Financial Services
- Keon Shahab
  Applied AI Architect
  ANTHROPIC

[Three numbered rectangular cards, each with a title, description, and a small icon: a lightning bolt for "Applying intelligence", a stack of coins for "Scaling agents", and a chain link for "Compounding impact".]

### 00:03:52

# Frontier AI changes what's possible

Autonomous fleets of agents

> "Our investment professionals live in data and analytical models. Claude for Excel meets them there and **gives them a step

### 00:04:20

# Frontier AI changes what's possible

## Autonomous fleets of agents

> "Our investment professionals live in data and analytical models. Claude for Excel meets them there and **gives them a step-change

### 00:04:52

# The diffusion gap: intelligence is ahead of adoption

- **4 months** CAPABILITY
  for the length of task a model can finish to double
- **74%** ADOPTION

### 00:05:21

# The Claude model family
Working across financial services to bring each model to the right use cases.

### Haiku 5.5
**Fastest and lowest cost**
-

### 00:05:50

# Smarter, cheaper and safer with every release

### INTELLIGENCE IS GOING UP
- **16+ hrs** of work a model can finish on its own
- In 2023 minutes

### COST PER TASK IS GOING DOWN
- **40%** less to run: Opus 5.5 against Opus 5
  - Haiku 5.5: 75% less
  - Sonnet 5.5: up to 30% less
  - Fable 5.1: 25% less

### AND IT IS GETTING SAFER
- **1.0%** of prompt-injection attacks succeed on Fable 5.1
  - Fable 5, the release before: 6.5%
  - Other frontier models: 9% to 53%

Capability: METR task-length measurements. Cost: Anthropic's launch pages, each against the model it replaced.
Safety: Gray Swan indirect prompt injection benchmark, attack success within 15 attempts. Figures to confirm before the talk.

## Keynote
- Belinda Neal
  - Managing Director, Financial Services
- Keon Shahab
  - Applied AI Architect
- ANTHROP\C

### 00:06:20

# Smarter, cheaper and safer with every release

## INTELLIGENCE IS GOING UP
- **16+ hrs** of work a model can finish on its own
- In 2023: minutes

## COST PER TASK IS GOING DOWN
- **40%** less to run: Opus 5.5 against Opus 5
- Haiku 5.5: 75% less
- Sonnet 5.5: up to 30% less
- Fable 5.1: 25% less

## AND IT IS GETTING SAFER
- **1.0%** of prompt-injection attacks succeed on Fable 5.1
- Fable 5, the release before: 6.5%
- Other frontier models: 9% to 53%

Capability: METR task-length measurements. Cost: Anthropic's launch pages, each against the model it replaced.
Safety: Gray Swan indirect prompt injection benchmark, attack success within 15 attempts. Figures to confirm before the talk.

### Keynote
Belinda Neal
Managing Director, Financial Services

Keon Shahab
Applied AI Architect
ANTHROPIC

### 00:06:50

# Banks are moving faster

## EVIDENT AI: AVERAGE AI MATURITY SCORE, 50 LARGEST BANKS
+26% in twelve months

## WHAT THE LEADING BANKS PUBLICLY REPORT
- **FRAUD**: Auto loan screening: more than $120M of potentially fraudulent applications stopped in twelve months.
- **LENDING**: Mortgage decisions: 70% of a bank's own applications decided on the same day.
- **ADVICE**: Advisor content search: client content found about 95% faster; gross sales up 20%.
- **RESEARCH**: Post-earnings research: drafts in about 15 minutes, down from 45.
- **ENGINEERING**: AI-assisted coding: 35% more code, with shorter cycle times.

Source: Evident AI Index: Key Findings Report, October 2026. 50 of the largest banks in North America, Europe and Asia.
Use cases are the banks' own public disclosures, as profiled in the report.

### Keynote
**Belinda Neal**
Managing Director, Financial Services

**Keon Shahab**
Applied AI Architect
ANTHROP\C

[Bar chart showing AI maturity scores increasing from 30.3 in 2023 to 46.1 in 2026]

### 00:07:19

# Banks are moving faster

## EVIDENT AI: AVERAGE AI MATURITY SCORE, 50 LARGEST BANKS
**+26%** in twelve months

### WHAT THE

### 00:07:52

# Agents work. Now scale them.

## Claude assists
- Ask it things. Marvel a bit. Move on.
  - "What's our policy for booking international travel?"

### 00:08:19

# AI Engineer NEW YORK

Presented by Arize

## Keynote
- Belinda Neal, Managing Director, Financial Services
- Keon Shahab, Applied AI Architect, ANTHROPIC

### A year ago
**Exploration**
Pilots and proofs of concept
Built by **A few engineers**

### This year
**Production**
A handful of agents, each built by hand
Built by **Engineering teams**

### Future
**Scale**
Thousands of specialists on one foundation
Built by **Engineers and the business**

[Illustration of three outlined circles]
[Illustration of five filled circles]
[Illustration of a large grid of small filled circles]

### 00:08:52

### A year ago
## Exploration
Pilots and proofs of concept
BUILT BY
**A few engineers**

### This year
## Production
A handful of agents, each built

### 00:09:19

# The agent foundation: five layers, built once

## WHAT LETS YOU TRUST IT
- 5 Traces and evals: How do you know it still works?
- 4 Sandbox and approval: Who signs off before it acts?
- 3 Identity and permissions: Who is it acting as?

## WHAT IT CAN DO
- 2 Skills: Does it know your business and your processes?
- 1 Connectors: Can it reach your systems and data?
- The model: Can it do the work? Most teams start here.

### 00:09:49

# The agent foundation: five layers, built once

### WHAT LETS YOU TRUST IT
- **5 Traces and evals** - How do you know it still works?
- **4 Sandbox and approval** - Who signs off before it acts?
- **3 Identity and permissions** - Who is it acting as?

### WHAT IT CAN DO
- **2 Skills** - Does it know your business and your processes?
- **1 Connectors** - Can it reach your systems and data?
- **The model** - Can it do the work? Most teams start here.

Illustrative example

## Keynote
- Belinda Neal, Managing Director, Financial Services
- Keon Shahab, Applied AI Architect
- ANTHROPIC

### 00:10:19

### Layer 1
# Connectors

Can't it reach your systems and data?

Wire each system once.
Control it in one place.
Any agent can use it.

Illust

### 00:10:52

# Skills
Does it know your business and your processes?

`break-investigation/SKILL.md`
How this firm investigates a trade break

1.  **Match both sides of

### 00:11:20

## Layer 2

# Skills

Does it know your business and your processes?

`break-investigation/SKILL.md`
How this firm investigates a trade break

1.

### 00:11:52

### Layer 3
## Identity and permissions
Who is it acting as?

- No agent is a super-user. Each one is scoped to its job.

```
identity: trade-

### 00:12:20

### Layer 3
# Identity and permissions
Who is it acting as?

- **identity**: trade-break-investigator
- **role**: ops-analyst
-

### 00:12:57

### Layer 4

# Sandbox and approval

Who signs off before it acts?

1.  **Agent proposes** a booking amendment
2.  **Staged as a dry run** nothing real moves
3.  **A person approves** the operations lead
4.  **Change is applied** exactly as approved

Nothing real moves until a named person approves.

The skill asks. A hook enforces.
Code that runs before every action. The model cannot skip it.

## Keynote
Belinda Neal
Managing Director, Financial Services

Keon Shahab
Applied AI Architect
ANTHROP\C

[Icon of stacked horizontal lines next to "Layer 4"]
[Lock icon next to "The skill asks. A hook enforces."]

### 00:13:19

### Layer 4
# Sandbox and approval
Who signs off before it acts?

1.  **Agent proposes** a booking amendment
2.  **Staged as a dry run** nothing real moves
3.  **A person approves** the operations lead
4.  **Change is applied** exactly as approved

Nothing real moves until a named person approves.

The skill asks. A hook enforces.
Code that runs before every action.
The model cannot skip it.

[Icon of three horizontal lines with a shorter line on top, next to "Layer 4"]
[Lock icon next to "The skill asks. A hook enforces."]

### 00:13:48

### Layer 5
# Traces and evals
## How do you know it still works?

### TRADE-BREAK CASES PASSING, OUT OF 120

- **Baseline**:

### 00:14:20

### Layer 5

# Traces and evals

How do you know it still works?

TRADE-BREAK CASES PASSING, OUT OF 120

- **Baseline

### 00:14:47

# A strong foundation lets you move fast

The controls a risk committee asks for are the same ones that let a team ship the next agent without starting over.

## THE FIRST THREE QUESTIONS A RISK COMMITTEE ASKS

- **Layer 3**
  Who is it acting as?
  **Identity and permissions**
- **Layer 4**
  Who signs off before it acts?
  **Sandbox and approval**
- **Layer 5**
  How do you know it still works?
  **Traces and evals**

Regulated does not mean slow. Every new agent inherits all three.

Illustrative example

[Three boxes, each with an icon (key, checkmark, calendar with graph), a "Layer" number, a question, and a bolded concept, illustrating a framework for risk assessment.]

### 00:15:24

# A new agent is mostly already built

## CHOSEN FROM THE FOUNDATION
- **connectors**: orders, market-data
- **role**: ops-analyst

### 00:15:47

# One agent layer, four doors

## A PERSON
### Claude, incl. Cowork
**Hands off a task**
The advisor, the analyst, the underwriter.

### 00:16:25

# Engineers and the business, in one loop

- **1 Works with the agent** (The Business)
  An operations analyst runs trade breaks through it every day.
- **

### 00:16:51

# Engineers and the business, in one loop

1.  **Works with the agent**
    An operations analyst runs trade breaks through it every day.
2.  **Corrects it**
    Says what it got wrong, in plain language, where the work happens.
3.  **Becomes a skill change**
    The agent drafts the skill edit and a new test case for review.
4.  **Reviewed and tested**
    Evals rerun. Operations approves the method, engineering the release.

The business owns the method. Engineering owns the release. Every agent improves.

illustrative example

### Keynote
Belinda Neal
Managing

### 00:17:24

# One foundation, many specialists

- Trade breaks (Operations)
- KYC review (Compliance)
- Credit memo (Credit)
- Onboarding (Client service)
- Surveillance (

### 00:17:50

# Compounding impact
The same foundation, built once for an industry: Claude for Financial Advisors.

### Keynote
**Belinda Neal**
Managing Director, Financial Services

**Keon Shahab**
Applied AI Architect
ANTHROP\C