IntellectAI Deploys Multi-Agent Systems to Automate Bank GRC Workflows

Risk Rewired: From Guardrails to Growth Engines With IntellectAI — Ai4

Ai4

Embedding automated governance directly into multi-agent AI pipelines allows financial institutions to scale thousands of autonomous agents while cutting compliance turnaround times by 90%.

AI-assisted summary of Risk Rewired: From Guardrails to Growth Engines With IntellectAI — original source. Check the original for full context.

Source published:

Detailed notes

## Background Published on YouTube by the Ai4 channel on August 27, 2025, this presentation features Rajesh Muthuramalingam, Senior Vice President at IntellectAI. The session outlines how financial institutions can transform governance, risk, and compliance (GRC) from slow-moving innovation bottlenecks into strategic growth enablers by embedding automated risk controls directly into the enterprise AI lifecycle.

## TL;DR Rajesh Muthuramalingam demonstrates why traditional, reactive GRC frameworks fail in the era of agentic AI, where tier-one institutions are deploying hundreds of thousands of autonomous agents into production. Introducing IntellectAI's Purple Fabric platform, he advocates for a 'governance by design' architecture featuring native guardrails, deterministic data lineage, and model optimization. The presentation spotlights concrete production case studies, including a major UK wealth manager reducing complaint resolution turnarounds from 50 days to 20 minutes and a $1.6 trillion sovereign fund boosting ESG document analysis capacity by 1,000x. Muthuramalingam concludes by illustrating how embedded governance allows organizations to safely deliver production-grade agentic MVPs within an 8-week timeframe.

## Atmosphere & style The presentation carries the polished, authoritative rhythm of a seasoned fintech leader delivering a pragmatic keynote on the final morning of an enterprise tech conference. Muthuramalingam balances technical depth with approachable humor, using self-aware quips about hardcore governance buffs and analogies from The Wolf of Wall Street to hook the audience before pivoting into rigorous architectural diagrams and enterprise benchmarks. The tone is brisk, optimistic, and relentlessly focused on operational execution, making complex compliance engineering feel like an agile commercial driver rather than an administrative chore.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=EIA-kj37X74&t=4s) IntellectAI has focused on operationalizing multi-agent enterprise AI into live production over the past 18 months via its Purple Fabric platform. - [00:04](https://www.youtube.com/watch?v=EIA-kj37X74&t=4s) Intellect Design Arena supports major global financial institutions, powering 6 of the top 10 North American banks and 10 of the top 12 in Europe. - [01:41](https://www.youtube.com/watch?v=EIA-kj37X74&t=101s) Believing enterprise AI momentum can proceed without governance is reckless 'fictional thinking' that leads to severe operational fallout. - [01:41](https://www.youtube.com/watch?v=EIA-kj37X74&t=101s) Enterprise risk must transition from reactive, seatbelt-style damage mitigation to dynamic, predictive controls built like autonomous driving systems. - [03:58](https://www.youtube.com/watch?v=EIA-kj37X74&t=238s) Traditional GRC teams face severe strain and burnout from fragmented spreadsheets, manual audits, and escalating regulations such as SEC climate rules and UK Consumer Duty. - [03:58](https://www.youtube.com/watch?v=EIA-kj37X74&t=238s) The massive scale of enterprise AI adoption, such as a Tier-1 UK bank targeting 300,000 production agents, makes manual compliance reviews obsolete. - [06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) Compliance must be 'governance by design,' embedding continuous risk and regulatory oversight directly into business workflows rather than applying bolt-on controls. - [06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) Purple Fabric operates on four key architectural pillars: Data to Knowledge, Multi-Agent Orchestration, LLM Optimization, and Native Enterprise Governance. - [09:24](https://www.youtube.com/watch?v=EIA-kj37X74&t=564s) A multi-agent complaints workflow for a £170B+ UK wealth manager cut case handling times by 90% (from 50 days to 20 minutes) and reduced costs by over 50%. - [09:24](https://www.youtube.com/watch?v=EIA-kj37X74&t=564s) IntellectAI's ESG Edge platform enabled a $1.6T sovereign wealth fund to analyze 10M+ documents across 9,000 entities, scaling research capacity 1,000x at >90% accuracy. - [12:53](https://www.youtube.com/watch?v=EIA-kj37X74&t=773s) Enterprise multi-agent systems demand mandatory safeguards including explainability, full auditability, deterministic data lineage, prompt injection defense, and bias detection. - [12:53](https://www.youtube.com/watch?v=EIA-kj37X74&t=773s) Enterprises can successfully move from initial use-case identification to a fully governance-compliant production MVP in under 8 weeks.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=EIA-kj37X74&t=4s) Introduction and IntellectAI's Purple Fabric Platform

Welcoming the audience on the third day of the conference, Rajesh Muthuramalingam jokes that anyone attending a governance track this late in the event must be "absolutely hardcore governance buffs." Over the past 18 months, his team at IntellectAI has focused on taking agentic AI from theoretical experiments into enterprise production. These implementations are powered by Purple Fabric, IntellectAI's proprietary enterprise platform designed to operationalize multi-agent systems with built-in oversight.

IntellectAI is part of Intellect Design Arena, an established financial technology organization with a 30-year legacy operating across more than 57 countries with nearly 350 enterprise clients. While the company has delivered fintech products for over a decade and a half, its client base includes six of the top ten financial institutions in North America and ten of the top twelve in Europe. Through Purple Fabric and generative AI capabilities, IntellectAI is now scaling these technologies beyond traditional financial services into broader enterprise sectors.

- **Corporate Footprint:** 30-year track record, active in 57+ countries, serving ~350 enterprise customers. - **Market Penetration:** Powering 6 in 10 tier-one North American financial institutions and 10 in 12 European leaders. - **Core Focus:** Productionizing multi-agent AI systems over the last 18 months via the Purple Fabric platform.

---

[01:41](https://www.youtube.com/watch?v=EIA-kj37X74&t=101s) The Wolf of Wall Street Analogy and Outdated Risk Thinking

To illustrate what happens when innovation outpaces oversight, Muthuramalingam polls the room on who has watched *The Wolf of Wall Street*. He recalls the film's opening half: Leonardo DiCaprio aggressively pushing penny stocks amidst wild parties, uninhibited market frenzy, and extreme momentum, completely devoid of any mention or discussion of risk management. While entertaining on screen, the film serves as a heightened dramatization of real-world recklessness. When organizations engage in "fictional thinking"—believing that momentum can exist without governance, guardrails, and oversight—the fallout is inevitable and severe.

Muthuramalingam argues that enterprise perspectives on risk remain dangerously outdated. Traditional governance treats risk like seat belts or anti-lock brakes in conventional cars: static, reactive controls designed for damage mitigation after something goes wrong. In contrast, modern risk management must mirror the engineering of autonomous, self-driving vehicles, where risk controls are active, dynamic, and predictive. Enterprises can no longer rely on hindsight fixes while the speed of AI deployment accelerates; compliance systems must match the velocity of business innovation in real time.

---

[03:58](https://www.youtube.com/watch?v=EIA-kj37X74&t=238s) The Broken GRC Landscape and the Scale of Modern AI

Governance, Risk, and Compliance (GRC) functions across the enterprise are facing systemic strain. Traditional setups are weighed down by siloed data, legacy processes, manual audits, and compounding regulatory mandates. Risk and compliance professionals are burned out trying to keep pace with evolving mandates—such as the SEC's climate-related disclosure rules in the United States or Consumer Duty regulations in the United Kingdom. Corporate boards are demanding predictive foresight rather than static spreadsheet reports compiled after the fact.

```mermaid flowchart TD accTitle: Modern GRC Transformation Challenges subgraph Legacy["Traditional GRC Failures"] Silos["Siloed Systems & Legacy Audits"] Burnout["Stretched, Burned-Out Teams"] Reactive["Spreadsheet-Driven Hindsight"] end

subgraph Shift["Enterprise AI Demands"] Scale["Scale: 300,000+ Production Agents"] Regs["Dynamic Global Regulations"] BuiltIn["Embedded Intelligent Compliance"] end

Legacy -->|Must Transform Into| Shift ```

This structural strain is reaching a breaking point as the sheer volume of AI agents entering enterprise workflows escalates. Muthuramalingam points to a major UK high street bank that has publicly committed to building and deploying 300,000 AI agents into production within the next three to five years. Orchestrating and monitoring 300,000 autonomous agents fundamentally challenges conventional operational resilience. Bolt-on controls and manual review queues cannot scale to meet that volume without bringing business agility to a complete halt.

- **Regulatory Pressure Points:** SEC climate disclosure mandates in the US and FCA Consumer Duty requirements in the UK. - **Scale Benchmark:** Tier-1 UK high street bank targeting 300,000 production AI agents within 3 to 5 years. - **GRC Deficits:** Over-reliance on manual compliance spreadsheets, reactive audit trails, and fragmented toolsets.

---

[06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) The New Playbook: Governance by Design and Purple Fabric Architecture

Because the legacy playbook is obsolete, organizations require a new paradigm: intelligent compliance that is "governance by design." Risk controls cannot exist as an afterthought or an awkward overlay; they must be woven natively into end-user business workflows so compliance occurs continuously and invisibly. Just as autonomous vehicles require a completely redesigned powertrain rather than a mechanical seat belt, enterprise AI demands agentic infrastructure to autonomously police and enforce governance standards.

```mermaid flowchart LR accTitle: Purple Fabric Platform Core Architecture subgraph PF["Purple Fabric Enterprise Platform"] direction TB D2K["Data to Knowledge<br/>(Ingestion & Pre-processing)"] OE["Orchestration Engine<br/>('Digital Experts' Multi-Agent Hub)"] OPT["LLM Optimization Hub<br/>(Cost, Speed & Accuracy)"] GOV["Enterprise Governance<br/>(Built-in Guardrails & Lineage)"] end

D2K --> OE OE --> OPT GOV -.->|Native Lifecycle Governance| D2K GOV -.->|Native Lifecycle Governance| OE GOV -.->|Native Lifecycle Governance| OPT ```

Purple Fabric operates as a zero-code enterprise business impact platform built around four foundational pillars: 1. **Data to Knowledge:** Comprehensive ingestion and pre-processing that converts raw, unstructured enterprise assets into structured knowledge graphs and embeddings for AI consumption. 2. **Orchestration Engine ("Digital Experts"):** A multi-agent framework capable of chaining specialized autonomous agents together to execute complex business workflows. 3. **LLM Optimization Hub:** Dynamic model routing and optimization that balances latency, compute cost, and response accuracy across diverse underlying foundation models. 4. **Enterprise Governance:** A non-negotiable, built-in layer that delivers real-time regulatory intelligence, continuous assurance, and automated auditability across every agent transaction.

---

[09:24](https://www.youtube.com/watch?v=EIA-kj37X74&t=564s) Real-World Case Studies: Complaints Investigation and ESG Intelligence

Muthuramalingam shares two live production deployments demonstrating Purple Fabric's measurable business impact. The first involves the UK's largest advice-led wealth management firm, managing over £170 billion in assets under management (AUM). Following the implementation of UK Consumer Duty regulations, the firm experienced a massive surge in customer complaints, leading to missed service level agreements (SLAs), compliance exposure, and severe staff burnout. IntellectAI deployed a multi-agent solution that automates the complaints journey end-to-end—from initial intake and evidence gathering to generating formal customer resolution letters. Operating in live production since October, the system reduced case handling times by 90% (slashing processing times from 50 days down to 20 minutes per case) while driving down operational costs by over 50%.

```mermaid flowchart TD accTitle: Case Studies Outcomes Summary subgraph CS1["UK Wealth Manager (£170B+ AUM)"] C1["End-to-End Complaints Multi-Agent System"] O1["Case Duration: 50 Days -> 20 Minutes"] O2["90% Time Reduction & 50%+ Cost Savings"] end

subgraph CS2["European Sovereign Fund ($1.6T AUM)"] C2["ESG Edge: 10M+ Documents Across 9,000 Entities"] O3["1,000x Increase in Analysis Capacity"] O4[">90% Verified Extraction Accuracy"] end

CS1 --> Success["Production Outcomes"] CS2 --> Success ```

The second case study highlights "ESG Edge," built for one of Europe's largest sovereign wealth funds managing $1.6 trillion in assets. Rather than treating ESG metrics as superficial marketing disclosures, the fund wanted deep non-financial parameters embedded directly alongside financial metrics in its investment risk framework. IntellectAI built a multi-agent system that analyzes over 10 million unstructured documents across 9,000 portfolio companies. The platform extracts verified, highly granular ESG data points at greater than 90% accuracy, enabling investment analysts to conduct 1,000 times more company analysis without adding headcount.

- **Complaints Automation Metrics:** 90% decrease in handling turnaround, moving from a 50-day average to 20 minutes, with over 50% cost reduction. - **ESG Analysis at Scale:** 10+ million documents parsed across 9,000 global portfolio assets for a $1.6 trillion sovereign fund. - **Analytical Throughput:** 1,000x expansion in ESG research depth maintained at >90% data accuracy.

---

[12:53](https://www.youtube.com/watch?v=EIA-kj37X74&t=773s) Governance Non-Negotiables and Rapid 8-Week Deployment

Reflecting on these production rollouts, Muthuramalingam details the mandatory governance requirements for multi-agent enterprise systems. Solutions must guarantee absolute explainability, full auditability, and deterministic data lineage tracing back to source documents. Platforms must feature built-in bias detection, toxicity filters, prompt injection guardrails, and real-time model oversight. When firms transition from managing human processes to overseeing hundreds of thousands of AI agents, the internal risk function must evolve entirely.

Successfully completing these initial projects builds organizational confidence, creating an internal "hearts and minds journey" across compliance and technology teams. For example, after successfully deploying the complaints engine, the UK wealth management firm expanded its footprint by launching an enterprise GPT chat advisor to assist advisors in their daily client engagements. In closing, Muthuramalingam underscores IntellectAI's execution speed: moving from initial use case identification to a fully governance-ready, risk-compliant MVP deployed in live production in under 8 weeks.

- **Governance Pillars:** End-to-end data lineage, explainability, automated toxicity and bias mitigation, prompt injection shielding, and continuous model performance monitoring. - **Enterprise Expansion:** UK wealth manager progressed from complaints automation to an advisory-facing enterprise GPT assistant. - **Time to Value:** Complete transition from use-case scoping to production-grade, risk-certified MVP deployment within an 8-week delivery timeframe.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=EIA-kj37X74&t=4s) absolutely hardcore governance buffs — Rajesh Muthuramalingam

> [01:41](https://www.youtube.com/watch?v=EIA-kj37X74&t=101s) fictional thinking — Rajesh Muthuramalingam

> [06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) governance by design — Rajesh Muthuramalingam

> [06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) Digital Experts — Rajesh Muthuramalingam

> [06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) Data to Knowledge — Rajesh Muthuramalingam

> [09:24](https://www.youtube.com/watch?v=EIA-kj37X74&t=564s) ESG Edge — Rajesh Muthuramalingam

> [12:53](https://www.youtube.com/watch?v=EIA-kj37X74&t=773s) hearts and minds journey — Rajesh Muthuramalingam

> governance from a barrier into a strategic enabler of agility, trust, and competitive performance — Rajesh Muthuramalingam

## Mentioned Rajesh Muthuramalingam, IntellectAI, Purple Fabric, Intellect Design Arena, The Wolf of Wall Street, Leonardo DiCaprio, Securities and Exchange Commission (SEC), Financial Conduct Authority (FCA) Consumer Duty, ESG Edge, Ai4

## Video description Rajesh Muthuramalingam, Senior Vice President - AI Products, Delivery and UK GTM, IntellectAI

Abstracts

In financial services, governance, risk, and compliance (GRC) are non-negotiable — but when it comes to AI, these functions are often seen as slow-moving gate keepers slowing innovation rather than enabling it. In some cases, governance becomes a tool for resisting change, used to stall progress in the name of risk control. However, leading institutions are proving that strong governance can be a competitive differentiator, not a bottleneck. This session will explore how banks, insurers, and financial institutions are redefining AI governance to meet regulatory demands while accelerating responsible AI deployment. Instead of reactive oversight, enterprises are embedding ethical, regulatory, and risk controls directly into the end-to-end AI lifecycle — reducing compliance gaps while accelerating time-to-value.

Financial services and GRC leaders attending the session will see real world implementations for GRC use cases that will give insights on how to harness enterprise AI platforms and solutions to transform governance from a barrier into a strategic enabler of agility, trust, and competitive performance.

[Source](https://www.youtube.com/watch?v=EIA-kj37X74)

中文简报

IntellectAI部署多智能体系统，自动化银行GRC工作流 将自动化治理直接嵌入多智能体AI流水线，使金融机构能够在扩展数千个自主智能体的同时，将合规处理周期缩短90%。 Rajesh Muthuramalingam阐释了在智能体AI时代，当顶尖机构向生产环境中部署数十万个自主智能体时，传统的被动式GRC框架为何难以应对。结合IntellectAI的Purple Fabric平台，他倡导采用具备原生防护栏、确定性数据溯源以及模型优化特性的“设计即治理”（governance by design）架构。演讲重点介绍了具体的生产案例，包括一家英国大型财富管理机构将投诉处理周期从50天缩短至20分钟，以及一家规模达1.6万亿美元的主权基金将ESG文档分析能力提升了1000倍。Muthuramalingam最后展示了嵌入式治理如何让机构在8周时间内安全交付生产级智能体MVP。 IntellectAI 依托服务北美排名前十中 6 家银行以及欧洲排名前十二中 10 家机构的深厚积淀，在过去 18 个月中通过其 Purple Fabric 平台推动多智能体企业级 AI 在实际生产环境中落地。 (0:04) 企业传统的 GRC 治理模式将风险视为事后补救的被动防线，面对有英国大行计划在 3 到 5 年内上线 30 万个 AI 智能体的庞大规模，人工与电子表格审查已彻底失效，必须转向如自动驾驶般的动态预测型“设计即治理”机制。 (3:58) Purple Fabric 平台围绕数据转知识、多智能体编排、LLM 优化及原生企业治理四大支柱构建，将合规与防护栏无缝嵌入全生命周期业务流程中。 (6:06) 实际生产案例显示，该平台帮助管理超 1700 亿英镑资产的英国财富管理机构将投诉处理时长从 50 天压缩至 20 分钟，并为管理 1.6 万亿美元的欧洲主权基金以超 90% 的准确率分析 9000 家企业的 1000 万份文档，带来 1000 倍的分析产能提升。 (9:24) IntellectAI 通过具备可解释性、数据血缘和防提示词注入的原生合规框架，可在 8 周内完成从用例定义到上线生产级、符合监管要求的 MVP 部署。 (12:53)

Reading summary

IntellectAI's Rajesh Muthuramalingam argues that legacy governance, risk, and compliance frameworks cannot handle the velocity of enterprise artificial intelligence, requiring organizations to replace reactive controls with dynamic, natively integrated 'governance by design.' The Breakdown of Traditional GRC at Scale (3:58) Muthuramalingam contends that conventional Governance, Risk, and Compliance (GRC) functions operate on outdated models, likening traditional risk practices to reactive seat belts designed only for post-incident damage control. As enterprises adopt AI, legacy systems relying on manual audits, siloed data, and retrospective spreadsheets are straining under compounding regulatory mandates, such as the SEC's climate disclosures in the United States and the Financial Conduct Authority's Consumer Duty in the United Kingdom. This structural friction becomes unmanageable as the sheer volume of enterprise autonomous systems grows. Muthuramalingam highlights an unnamed tier-one UK bank planning to deploy 300,000 AI agents into production within three to five years. In his view, manual review queues and bolt-on governance checks cannot scale to monitor hundreds of thousands of concurrent agents without crippling enterprise agility. Embedding Governance Into Multi-Agent Architecture (6:06) To prevent innovation from outpacing oversight, Muthuramalingam advocates for intelligent compliance integrated directly into operational workflows. Paralleling the engineering shift required for autonomous vehicles, enterprise AI requires infrastructure where compliance rules execute automatically and continuously across every transaction. IntellectAI implements this approach through its zero-code Purple Fabric platform, which structures multi-agent deployment across four core pillars: a 'Data to Knowledge' layer that converts unstructured enterprise data into structured knowledge graphs; an 'Orchestration Engine' coordinating specialized agent workflows; an 'LLM Optimization Hub' balancing compute costs, latency, and response accuracy; and an underlying 'Enterprise Governance' layer providing continuous auditability, lineage tracking, and regulatory intelligence. Measured Outcomes in Live Production Deployments (9:24) Muthuramalingam presents two institutional case studies to demonstrate the practical application of agentic oversight. In the first, a UK wealth management firm overseeing more than £170 billion in assets faced missed service level agreements and compliance risks following the introduction of UK Consumer Duty rules. IntellectAI deployed a multi-agent system to automate complaint intake, evidence assembly, and resolution drafting. In live production since October, the implementation reportedly reduced average case handling times by 90%—from 50 days to 20 minutes—while lowering operational costs by over 50%. In the second deployment, Europe's largest sovereign wealth fund, managing $1.6 trillion in assets, implemented 'ESG Edge' to assess non-financial risks across its portfolio. The platform's multi-agent network parses more than 10 million unstructured documents across 9,000 corporate assets, extracting verified sustainability metrics at over 90% accuracy and expanding research capacity by 1,000 times without expanding analyst headcount. Guardrail Standards and Fast-Track Deployment Timelines (12:53) According to Muthuramalingam, enterprise-grade multi-agent governance requires specific technical baselines, including deterministic data lineage back to primary sources, full explainability, automated toxicity and bias mitigation, and defenses against prompt injection. Establishing these baseline guardrails in early use cases helps build organizational trust, as seen when the UK wealth manager subsequently expanded from complaints handling to deploying internal advisory-facing chat tools. Muthuramalingam asserts that standardizing these guardrails enables rapid implementation, claiming that IntellectAI routinely transitions clients from initial problem identification to a fully governance-certified, production-ready minimum viable product within an eight-week timeframe. While IntellectAI demonstrates dramatic operational metrics and rapid eight-week implementations across major financial institutions, successfully maintaining automated oversight as enterprises scale toward hundreds of thousands of active production agents remains an evolving industry challenge.

Detailed notes — zh

## Background 该内容由行业媒体频道 Ai4 于 2025 年 8 月 27 日在 YouTube 平台发布，由 IntellectAI 高级副总裁 Rajesh Muthuramalingam 担任主讲。视频简介指出，本次演讲旨在探讨银行和保险等金融机构如何打破传统 GRC（治理、风险与合规）阻碍创新的刻板印象，通过将合规内嵌于端到端 AI 生命周期中，将治理重构为推动业务敏捷与可信落地的核心增长引擎。 ## TL;DR 演讲围绕智能体 AI（Agentic AI）时代的治理革新展开，指出面对未来数十万智能体并发及严苛监管的挑战，企业必须从传统事后补救转向内生式、前瞻驱动的智能合规。IntellectAI 依托其零代码 Purple Fabric 平台（涵盖数据转化、编排引擎、LLM 优化及企业治理四大支柱），展示了两个落地生产案例：帮助管理规模 1,700 亿英镑的财富机构将投诉调查时间从 50 天压缩至 20 分钟并降低 50% 成本；赋能 1.6 万亿美元主权基金实现 1,000 倍的 ESG 文档分析吞吐量。最后，演讲强调了全链路可解释性与内置安全护栏的重要性，并提出 8 周交付合规就绪生产级 MVP 的敏捷落地承诺。 ## Atmosphere & style 整场演讲呈现出务实、自信且高度聚焦于工业级落地的专业氛围。讲者以幽默的行业破冰开场，随后借由《华尔街之狼》与自动驾驶等生动类比，将晦涩的合规治理话题讲得通俗而深刻。现场叙事节奏紧凑，通过震撼的生产级真实数据与技术架构展示，既化解了金融机构对 AI 风险的失控焦虑，又传递出清晰的落地掌控力与前瞻野心。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=EIA-kj37X74&t=4s) IntellectAI 依托母公司 30 年金融科技积淀与企业级平台 Purple Fabric，已覆盖北美 60% 与欧洲 83% 的大型顶级金融机构，并在过去 18 个月持续推动 Agentic AI 的生产落地。 - [01:41](https://www.youtube.com/watch?v=EIA-kj37X74&t=101s) 企业风控思维亟需从传统外挂式“安全带”和事后补救（Hindsight），转变为类似于自动驾驶汽车的智能感知与前瞻预防（Foresight）。 - [03:58](https://www.youtube.com/watch?v=EIA-kj37X74&t=238s) 面对监管持续收紧（如 SEC 气候披露、英国 Consumer Duty）以及英国顶级商业银行计划 3–5 年部署 30 万个 AI 智能体的超大规模挑战，传统 GRC 管理手册已无法维系。 - [06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) 新型 GRC 范式主张“设计即治理”（Governance by Design），将智能合规机制直接内嵌至实际业务工作流中，使治理从创新阻碍转化为前瞻驱动力。 - [07:56](https://www.youtube.com/watch?v=EIA-kj37X74&t=476s) 支撑企业级 AI 落地的平台需具备四大核心支柱：数据转知识（Data to Knowledge）、多智能体编排引擎、大模型优化中心（平衡速度、成本与精度）以及全生命周期的企业级治理底座。 - [09:24](https://www.youtube.com/watch?v=EIA-kj37X74&t=564s) 英国规模超 1,700 亿英镑的财富管理机构借助多智能体自动化调查与答复系统，将单起投诉调查处理时间从 50 天缩减至 20 分钟，运营成本降低超 50%。 - [11:22](https://www.youtube.com/watch?v=EIA-kj37X74&t=682s) 欧洲 1.6 万亿美元主权财富基金通过智能体系统穿透 1,000 万份非结构化文档提取 9,000 家企业的非财务 ESG 数据，在保持 90% 以上高精准度的同时将分析吞吐量提升了 1,000 倍。 - [12:53](https://www.youtube.com/watch?v=EIA-kj37X74&t=773s) 规模化智能体部署的核心治理原则包括：全链路数据源血统追踪（Data Lineage）、内置偏见/毒性/提示词注入防御矩阵，以及面向海量智能体的自适应模型监督体系。 - [12:53](https://www.youtube.com/watch?v=EIA-kj37X74&t=773s) 企业无需在创新速度与安全合规之间妥协，通过内生治理平台，可在 8 周内完成从业务用例识别到“合规就绪”（Governance-ready）的生产级 MVP 投产。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=EIA-kj37X74&t=4s) 破冰与背景：IntellectAI 与 Purple Fabric 演讲者 Rajesh Muthuramalingam 开场便向现场观众打趣道，能够坚持来到大会第三天治理专场（Governance Track）的参会者，绝对都是“不折不扣的硬核治理发烧友”。过去 18 个月里，他和团队一直致力于将智能体 AI（Agentic AI）推向生产落地，而所有这些业务用例都是构建在一个名为 Purple Fabric 的企业级平台上。 Rajesh 随后介绍了背后的母公司 Intellect Design Arena。作为一家深耕金融科技领域长达 30 年的机构，Intellect 的业务遍及全球 57 个以上的国家和地区，服务近 350 家企业级客户。在过去 15 年间，公司不仅持续交付金融科技产品，自身也经历了深度的 AI 转型之旅。目前，北美排名前 10 的大型金融服务机构中有 6 家是其客户，欧洲排名前 12 的机构中更有 10 家采用了其解决方案；随着 AI 技术的演进，他们的服务边界也正迅速拓展至金融服务之外的更广阔领域。 - **业务版图**：服务覆盖全球 57+ 国家，拥有近 350 家企业级客户。 - **市场渗透率**：覆盖北美 60%（6/10）及欧洲 83%（10/12）的大型顶级金融机构。 - **技术沉淀**：团队在过去 18 个月持续推动 Agentic AI 的工业级落地。 --- ### [01:41](https://www.youtube.com/watch?v=EIA-kj37X74&t=101s) 从《华尔街之狼》到自动驾驶：重新定义风险管理 Rajesh 引用了经典电影《华尔街之狼》（The Wolf of Wall Street）来展开对风险的反思。电影前半段充斥着小李子（Leonardo DiCaprio）肆无忌惮地兜售廉价仙股（Penny Stocks）、彻夜狂欢与盲目亢奋的市场势头，自始至终没有提及任何合规监管或风险管理。然而电影的高潮与结局揭示了残酷的现实——脱离治理与监管的“虚构式思维”（fictional thinking）必然导致灾难性的崩盘。电影虽然运用了戏剧化的夸张手法，但其投射的现实后果却极为真实。 当今企业对待风险的方式依然严重过时，如同还在用传统汽车的“安全带”思维来管理风险。然而，现代自动驾驶汽车管理风险的逻辑与传统汽车有着本质不同。如果企业依然将风险管理视为静态机制，习惯于“事后补救”（hindsight）而非“前瞻预防”（foresight），就根本无法跟上当今业务狂飙突进的步伐。随着 AI 的全面普及，治理与合规系统必须具备与业务发展同频共振的速度。 --- ### [03:58](https://www.youtube.com/watch?v=EIA-kj37X74&t=238s) GRC 的现实困境与 30 万智能体带来的韧性挑战 当前企业的治理、风险与合规（GRC）正面临系统性危机：系统孤立割裂、遗留流程繁冗、审计手段落后，各行业还承受着沉重的监管超负荷（Regulatory Overload）。更严峻的是，合规与风控团队长期处于超负荷运转、极度疲惫（burned out）的状态。GRC 体系要么已经支离破碎，要么正处于崩溃边缘。与此同时，董事会亟需的是具备前瞻性的洞察，而不是深陷在电子表格中被动生成事后报告。 监管压力仍在持续加码，例如去年的 SEC 气候披露要求，以及英国推行的《消费者权益保护准则》（Consumer Duty）。更具颠覆性的是新技术引入对业务连续性和运营韧性（Operational Resilience）带来的冲击。Rajesh 分享了一个典型案例：英国一家大型主流商业银行公开宣布，计划在未来 3 到 5 年内构建并在生产环境中部署 30 万个 AI 智能体（AI Agents）。面对 30 万个在生产中自主运行的智能体，企业必须彻底重构其治理与风控逻辑，传统的管理手册已经完全失效。 - **监管加码**：包括美国 SEC 气候信息披露新规、英国 FCA Consumer Duty 法规等。 - **极端规模挑战**：英国顶级商业银行计划 3–5 年内在生产端上线 300,000 个 AI Agents。 --- ### [06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) 新治理法则：智能合规与内生式工作流 面对生成式与智能体 AI 的浪潮，企业必须打破旧有的操作范式，建立全新的 GRC Playbook。过去将风控与治理流程作为“外挂插件”（bolt-on）覆盖在业务系统之上的做法已不再可行。智能解决方案必须由“智能合规”（Intelligent Compliance）来保驾护航，将治理原则融入系统底层的“设计即治理”（Governance by Design）。 合规机制必须无缝嵌入到实际的业务工作流之中，员工在日常作业中无需跳转到独立流程去处理合规事务。唯有将治理下沉为内生基础设施，企业才能从被动的风险应对转向“前瞻驱动的风控”（foresight-driven risk）。如同自动驾驶汽车依托全套智能感知与决策系统来超越传统安全带一样，GRC 也需要借助 Agentic AI 本身作为底层驱动引擎，将合规从阻碍业务创新的关卡转化为企业敏捷前行的核心推力。 ```mermaid flowchart LR accTitle: Intelligent Governance Transformation Flow accDescription: Moving from legacy bolt-on compliance to intelligent, embedded, foresight-driven governance powered by Agentic AI. subgraph Legacy["传统模式 (Legacy GRC)"] A["外挂式合规 (Bolt-on)"] --> B["事后被动响应 (Hindsight)"] B --> C["割裂的工作流 & 表格"] end subgraph Future["新型模式 (Modern Agentic GRC)"] D["内生式治理 (By Design)"] --> E["前瞻驱动风控 (Foresight)"] E --> F["无缝嵌入业务工作流"] end Legacy -.->|"范式重构与 Agentic AI 赋能"| Future ``` --- ### [07:56](https://www.youtube.com/watch?v=EIA-kj37X74&t=476s) Purple Fabric 平台的四大核心支柱 为了将内生治理与 Agentic AI 深度融合，IntellectAI 推出了 Purple Fabric 企业级业务影响力平台。该平台采用零代码（Zero-code）架构，具备内置保障机制、多智能体编排能力以及实时监管情报。Rajesh 指出，任何致力于支撑企业级 AI 生产落地的平台，都应当构建在以下四大核心技术支柱之上： ```mermaid flowchart TD accTitle: Purple Fabric Four Pillars accDescription: The four core pillars of the Purple Fabric enterprise AI platform: Data to Knowledge, Orchestration Engine, LLM Optimization Hub, and Enterprise Governance. subgraph Platform["Purple Fabric 平台四大核心支柱"] P1["1. 数据转知识 (Data to Knowledge)<br/>数据预处理与结构化转化"] P2["2. 编排引擎 (Orchestration Engine)<br/>单/多智能体与数字专家无缝协同"] P3["3. 大模型优化中心 (LLM Optimization Hub)<br/>平衡速度、成本与精准度"] P4["4. 企业级治理 (Enterprise Governance)<br/>内置合规、风控与实时监管情报"] end ``` 1. **数据转知识（Data to Knowledge）**：负责海量复杂数据的清洗、预处理与知识转化，将其转化为 AI 智能体可直接消费的高质量结构化知识。 2. **编排引擎（Orchestration Engine）**：被定义为“数字专家”（Digital Experts）中枢，能够将单一或多智能体解决方案无缝协同编排，直接面向业务目标输出结果。 3. **LLM 优化中心（LLM Optimization Hub）**：针对企业级落地的核心诉求，在响应速度（Speed）、推理成本（Cost）与输出精准度（Accuracy）之间实现全局最优平衡。 4. **企业级治理（Enterprise Governance）**：作为不可妥协的基础底座，将风控策略、合规检查与安全护栏完全内置于平台全生命周期中。 --- ### [09:24](https://www.youtube.com/watch?v=EIA-kj37X74&t=564s) 落地案例一：英国万亿级财富管理机构的投诉调查智能体 Rajesh 详述了首个已完全进入生产环境的真实案例。客户是英国最大的顾问主导型财富管理机构，资产管理规模（FUM）超过 1,700 亿英镑。在英国出台 Consumer Duty 监管新规后，该机构遭遇了海量客户投诉涌入的严峻挑战，人工团队不仅承受巨大压力，还面临 SLA 超时和违规受罚的风险。 IntellectAI 团队利用 Agentic AI 重构了端到端的投诉处理流程。系统部署了多智能体协同方案，实现从投诉案件接入、事实核查、合规调查到最终决议信件（Resolution Letter）自动生成的全流程闭环。该系统自去年 10 月正式上线生产环境运行至今： - **处理时效飞跃**：单起投诉调查处理时间从原本的 **50 天骤降至 20 分钟**，案例处理耗时削减了 **90%**。 - **成本效益**：直接运营成本降低超过 **50%**。 - **业务外延**：该项目帮助该机构完成了引入 Agentic AI 的首轮治理心智建设；在此坚实基础上，他们随后成功上线了一款面向顾问的企业级 GPT 对话智能体，释放了大量顾问时间以专注服务高净值客户。 --- ### [11:22](https://www.youtube.com/watch?v=EIA-kj37X74&t=682s) 落地案例二：1.6 万亿美元主权财富基金的 ESG 深度洞察 第二个案例来自一家管理规模达 1.6 万亿美元的欧洲顶级主权财富基金（全球最大主权基金之一）。该基金拒绝流于形式的表面化 ESG 评级，而是希望深入穿透其投资组合中的 9,000 家上市公司，获取高度颗粒化、深度的非财务数据，并将这些 ESG 数据指标深度集成到其核心投资风险评估框架中。 IntellectAI 为其搭建了一套多智能体分析系统，系统自动穿透分析了超过 1,000 万份非结构化文档，精准提取出可直接输入风控模型的非财务数据指标。 - **分析产能爆发**：在完全不增加分析师团队编制的前提下，分析师对投资组合公司的 ESG 数据分析吞吐量提升了 **1,000 倍**。 - **模型精准度**：在处理超千万份非结构化文档的高压环境下，整体信息抽取与分析精准度持续保持在 **90% 以上**。 - **现实影响力**：该基金得以依据高度确凿的颗粒化数据，向被投公司施加真实的 ESG 改进压力，推动实质性的全球可持续发展。 --- ### [12:53](https://www.youtube.com/watch?v=EIA-kj37X74&t=773s) 规模化落地核心原则与 8 周投产承诺 基于众多复杂场景的落地实践，Rajesh 总结了企业规模化部署多智能体时在风控与治理方面不可妥协的核心原则： - **全链路可解释性与可追溯性（Explainability & Traceability）**：必须具备百分之百的数据源血统（Data Source Lineage）追踪能力，确保金融级高强监管下的每一步推理与结论都有据可查、可被审计。 - **内置安全防御矩阵**：偏见检测（Bias Detection）、毒性过滤（Toxicity Detection）以及提示词注入防御（Prompt Injection Guardrails）必须作为平台出厂标配，而非后置外挂。 - **新型模型监督体系（Model Oversight）**：面对数十万个智能体同时在线的未来，传统风控职能将彻底重塑，必须依靠自适应、自动化的智能体监控机制来进行生命周期治理。 在演讲尾声，Rajesh 强调，治理与合规不应成为业务创新的阻碍，而应成为推动增长的发动机。他给出了 IntellectAI 的生产承诺：**从识别业务用例、启动研发到构建出完全具备合规就绪（Governance-ready）与风控就绪的生产级 MVP，全周期控制在 8 周以内**。通过将信任与风控内嵌于工作流，企业完全可以在严守合规底线的同时实现指数级的业务跃迁。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=EIA-kj37X74&t=4s) 过去 18 个月里，我和团队一直致力于将智能体 AI（Agentic AI）推向生产落地，而所有这些业务用例都是构建在一个名为 Purple Fabric 的企业级平台上。 — Rajesh Muthuramalingam > [01:41](https://www.youtube.com/watch?v=EIA-kj37X74&t=101s) 如果企业依然将风险管理视为静态机制，习惯于“事后补救”（hindsight）而非“前瞻预防”（foresight），就根本无法跟上当今业务狂飙突进的步伐。 — Rajesh Muthuramalingam > [03:58](https://www.youtube.com/watch?v=EIA-kj37X74&t=238s) 合规与风控团队长期处于超负荷运转、极度疲惫（burned out）的状态。GRC 体系要么已经支离破碎，要么正处于崩溃边缘。 — Rajesh Muthuramalingam > [03:58](https://www.youtube.com/watch?v=EIA-kj37X74&t=238s) 英国一家大型主流商业银行公开宣布，计划在未来 3 到 5 年内构建并在生产环境中部署 30 万个 AI 智能体（AI Agents）。 — Rajesh Muthuramalingam > [06:06](https://www.youtube.com/watch?v=EIA-kj37X74&t=366s) 过去将风控与治理流程作为“外挂插件”（bolt-on）覆盖在业务系统之上的做法已不再可行。智能解决方案必须由“智能合规”（Intelligent Compliance）来保驾护航，将治理原则融入系统底层的“设计即治理”（Governance by Design）。 — Rajesh Muthuramalingam > [09:24](https://www.youtube.com/watch?v=EIA-kj37X74&t=564s) 单起投诉调查处理时间从原本的 50 天骤降至 20 分钟，案例处理耗时削减了 90%。 — Rajesh Muthuramalingam > [11:22](https://www.youtube.com/watch?v=EIA-kj37X74&t=682s) 在完全不增加分析师团队编制的前提下，分析师对投资组合公司的 ESG 数据分析吞吐量提升了 1,000 倍。 — Rajesh Muthuramalingam > [11:22](https://www.youtube.com/watch?v=EIA-kj37X74&t=682s) 在处理超千万份非结构化文档的高压环境下，整体信息抽取与分析精准度持续保持在 90% 以上。 — Rajesh Muthuramalingam > [12:53](https://www.youtube.com/watch?v=EIA-kj37X74&t=773s) 从识别业务用例、启动研发到构建出完全具备合规就绪（Governance-ready）与风控就绪的生产级 MVP，全周期控制在 8 周以内。 — Rajesh Muthuramalingam ## Mentioned Rajesh Muthuramalingam, IntellectAI, Intellect Design Arena, Purple Fabric, 《华尔街之狼》（The Wolf of Wall Street）, Leonardo DiCaprio, SEC, Consumer Duty, FCA, GPT ## Video description Rajesh Muthuramalingam, Senior Vice President - AI Products, Delivery and UK GTM, IntellectAI Abstracts In financial services, governance, risk, and compliance (GRC) are non-negotiable — but when it comes to AI, these functions are often seen as slow-moving gate keepers slowing innovation rather than enabling it. In some cases, governance becomes a tool for resisting change, used to stall progress in the name of risk control. However, leading institutions are proving that strong governance can be a competitive differentiator, not a bottleneck. This session will explore how banks, insurers, and financial institutions are redefining AI governance to meet regulatory demands while accelerating responsible AI deployment. Instead of reactive oversight, enterprises are embedding ethical, regulatory, and risk controls directly into the end-to-end AI lifecycle — reducing compliance gaps while accelerating time-to-value. Financial services and GRC leaders attending the session will see real world implementations for GRC use cases that will give insights on how to harness enterprise AI platforms and solutions to transform governance from a barrier into a strategic enabler of agility, trust, and competitive performance. [Source](https://www.youtube.com/watch?v=EIA-kj37X74)

Source quotes

[0:04] Rajesh Muthuramalingam: 过去 18 个月里，我和团队一直致力于将智能体 AI（Agentic AI）推向生产落地，而所有这些业务用例都是构建在一个名为 Purple Fabric 的企业级平台上。 [1:41] Rajesh Muthuramalingam: 如果企业依然将风险管理视为静态机制，习惯于“事后补救”（hindsight）而非“前瞻预防”（foresight），就根本无法跟上当今业务狂飙突进的步伐。 [3:58] Rajesh Muthuramalingam: 合规与风控团队长期处于超负荷运转、极度疲惫（burned out）的状态。GRC 体系要么已经支离破碎，要么正处于崩溃边缘。 [3:58] Rajesh Muthuramalingam: 英国一家大型主流商业银行公开宣布，计划在未来 3 到 5 年内构建并在生产环境中部署 30 万个 AI 智能体（AI Agents）。 [6:06] Rajesh Muthuramalingam: 过去将风控与治理流程作为“外挂插件”（bolt-on）覆盖在业务系统之上的做法已不再可行。智能解决方案必须由“智能合规”（Intelligent Compliance）来保驾护航，将治理原则融入系统底层的“设计即治理”（Governance by Design）。 [9:24] Rajesh Muthuramalingam: 单起投诉调查处理时间从原本的 50 天骤降至 20 分钟，案例处理耗时削减了 90%。 [11:22] Rajesh Muthuramalingam: 在完全不增加分析师团队编制的前提下，分析师对投资组合公司的 ESG 数据分析吞吐量提升了 1,000 倍。 [11:22] Rajesh Muthuramalingam: 在处理超千万份非结构化文档的高压环境下，整体信息抽取与分析精准度持续保持在 90% 以上。 [12:53] Rajesh Muthuramalingam: 从识别业务用例、启动研发到构建出完全具备合规就绪（Governance-ready）与风控就绪的生产级 MVP，全周期控制在 8 周以内。

Source quotes

[0:04] Rajesh Muthuramalingam: absolutely hardcore governance buffs [1:41] Rajesh Muthuramalingam: fictional thinking [6:06] Rajesh Muthuramalingam: governance by design [6:06] Rajesh Muthuramalingam: Digital Experts [6:06] Rajesh Muthuramalingam: Data to Knowledge [9:24] Rajesh Muthuramalingam: ESG Edge [12:53] Rajesh Muthuramalingam: hearts and minds journey Rajesh Muthuramalingam: governance from a barrier into a strategic enabler of agility, trust, and competitive performance

Original transcript

[0:04] You know what? Uh, day three governance [0:07] track. The ones here are absolutely [0:10] hardcore governance buffs, right? So, [0:13] thanks for thanks for joining this [0:14] session. Uh, my name is Rajesh [0:16] Mutramingham. Um, and me and my team [0:20] have been productionizing Agentic AI [0:22] over the last 18 months. And we work um [0:27] and and we built that on a platform [0:29] called Purple Fabric which I will talk [0:32] through in a uh in a in a little while. [0:35] But let me quickly introduce who we are. [0:37] Um [0:40] we are um part of Intellect Design [0:42] Arena. We are a 30-year um organization [0:46] providing fintech products. We're in 57 [0:49] plus countries uh close to 350 [0:52] enterprise customers. [0:54] And we've been providing fintech [0:58] products over the last uh decade and a [1:00] half, but also we've been on an AI [1:03] journey ourselves. Um, and Purple Fabric [1:06] is part of uh that suite of products [1:10] which I'll talk to you in a bit. [1:12] And we we work with the largest um u [1:17] banks and financial services in the [1:18] world. Uh as you can see um coming up. [1:24] Yes. So six in 10 um large financial [1:29] services in North America, 10 in 12 in [1:32] Europe. So yeah, we work with uh large [1:35] financial services but with with AI we [1:37] are expanding beyond financial services [1:39] as well. [1:41] Right. So who's seen this movie? Quick [1:46] show of hands. [1:48] All right. Good. That's a good show of [1:50] hands. So for those who haven't seen [1:52] this movie, I highly recommend this. But [1:54] for those who did see this movie, you [1:56] know how this started, right? So this is [1:58] just this was uh Leo selling penny [2:01] stocks and the first first half of the [2:03] movie was absolute wild parties and [2:07] unfettered um [2:09] um unfettered selling. There was [2:12] absolute frenzy. There was a lot of [2:13] market momentum, but uh there was no [2:16] mention of oversight. there was no there [2:18] was no risk management. I don't think [2:20] that was even discussed in that movie. [2:23] Uh and we all know how the movie ended [2:25] though. Um we all know that but it's not [2:28] just a movie. The movie is a heightened [2:30] representation of what actually [2:31] happened. And there is real world [2:35] fallout [2:36] uh when you have fictional thinking. And [2:39] when I say fictional thinking, [2:41] what I mean by that is if you do not [2:44] think about governance guardrails, [2:46] oversight [2:48] uh from the very beginning, uh your [2:50] fictional thinking is well, it will have [2:53] it will have its fallouts and I think [2:55] that's kind of what we got out of that [2:56] movie, right? So [2:59] what do we think about risk? The way we [3:01] think about risk is still still quite [3:04] outdated, right? So we still think about [3:07] risk like how we think about seat belts [3:10] in traditional cars. But risk in a [3:13] modern car, a self self-driving modern [3:16] car is quite different. Um the way you [3:18] manage risk is is going to be [3:20] fundamentally different in a [3:22] self-driving car. But we're still [3:24] treating risk uh as a as a static uh [3:28] static risk risk management. We're [3:30] thinking of um we're not thinking about [3:34] foresight. we're thinking about um how [3:36] do we fix things in hindsight. [3:39] So the key thing is we need to move at [3:43] the speed of the business and we you [3:45] know with all the AI proliferation it's [3:47] getting harder and harder and I think [3:49] it's really important that governance [3:51] system compliance keeps up with that [3:53] speed of the speed at which the business [3:55] is moving. [3:58] Quickly summarizing none of this is [4:01] something you guys are not aware of but [4:03] just to quickly summarize where the [4:05] problems are. We all understand we have [4:07] siloed systems legacy processes audits [4:10] is part of that legacy processes we have [4:14] regulatory overload in certain [4:15] industries. [4:17] We have burned out teams. It's really [4:19] important to emphasize that right. So [4:20] the teams your compliance functions your [4:23] risk functions your governance functions [4:25] they're stretched. the teams are burned [4:26] out and it's really hard to keep up with [4:30] things and clearly GRC is either broken [4:34] or it will break soon and and at the [4:37] same time with with the proliferation of [4:39] AI board wants foresight. They don't [4:42] want us to be stuck in spreadsheets [4:44] providing reactive uh reactive reports, [4:48] right? And the problem is it keeps [4:51] getting worse, right? It's it's it's not [4:54] like you you somehow catch up and and [4:57] things will be okay. What will happen is [4:59] something else will come up. Last year, [5:01] for example, there was SEC climate [5:03] disclosures. Uh in if you're AC from [5:05] across the pond in the UK, there's uh [5:08] consumer duty regulations. So, something [5:09] will keep hitting you. And with AI, [5:12] there's going to be a lot of emphasis on [5:14] AI accountability. [5:16] And also with all those new technologies [5:21] coming in uh in into the landscape you [5:24] will have operational resilience [5:26] challenge. Well let me pause on that [5:28] operational residence actually. So we [5:30] we've been working with a large high [5:32] street bank in the UK and they've openly [5:35] stated their ambition that they want to [5:38] build 300,000 agents and deploy them in [5:42] production in the next 3 to 5 years. [5:44] Just think about that, right? 300,000 AI [5:47] agents in production. Think about how [5:51] you need to completely rethink your [5:53] governance and risk management and also [5:56] think about the the impact it has on [5:58] your operational resilience. So I think [6:01] that's where the challenge is [6:04] and the thing is the old playbook [6:06] doesn't work anymore. So you need to [6:08] fundamentally rethink your playbook. [6:11] So I think this this is where the new [6:13] playbook is very different to the old [6:14] playbook from when it comes to [6:16] governance, risk and compliance. So you [6:18] need intelligent compliance. So you need [6:20] compliance to be built in. It needs to [6:22] be part of your design. It needs to be [6:25] governance by design. So you can't have [6:27] a bolt-on. You can't have a bolt-on of [6:30] risk risk processes and and governance [6:33] processes as an overlay to your existing [6:35] processes when it comes to AI. So the [6:38] whole thing needs to be intelligent [6:39] compliance. Intelligent solutions need [6:41] intelligent compliance and also [6:43] everything needs to be embedded as part [6:44] of your workflows. So it should be [6:46] seamless. Nobody should should have to [6:49] go and look at a different workflow to [6:51] manage your compliance. It should be [6:52] part of your actual workflow for the [6:56] business. And I think this is where it [6:59] leads to foresightdriven risk rather [7:01] than uh reactive risk management. [7:06] And [7:08] what that means is you need a completely [7:11] new engine. Like just like how uh a [7:14] self-driving car the way it manages risk [7:16] is way beyond a seat belt or an ABS. Uh [7:20] it it has a fundamentally different [7:22] engine to manage risk. I think that's [7:25] kind of where we are heading and there's [7:27] a lot being invested in in GRC annually. [7:30] This includes cyber security and the [7:31] rest of it. There's a lot of investment [7:33] in that. But also the way to solve that [7:36] is also agentic AI right this is where [7:38] agentic AI will power governance risk [7:41] and compliance and it's becoming [7:44] infrastructure right it's it needs to be [7:45] embedded it's not something uh it's not [7:48] something that you do after after the [7:50] fact it needs to be embedded [7:56] so this is where we come in Purple [7:58] Fabric um we are an enterprise business [8:02] impact platform We're a zero code [8:04] platform and all the use cases we built [8:07] has been on this platform and [8:11] this is where the governance is built [8:13] in. We have built-in assurance. We have [8:15] agentic AI obviously where we can [8:17] orchestrate multi- aent solutions and [8:19] also we can provide real-time regulatory [8:21] intelligence as well. Now I'll just [8:24] quickly go through what our platform is. [8:28] Um it's four key texts. It's data to [8:32] knowledge which is the first one. It's [8:34] really key that we do all the data [8:35] processing pre-processing [8:37] and convert the data to knowledge for [8:40] the AI agents to consume. Then we have [8:42] the orchestration engine which does uh [8:45] we what we call digital experts where [8:48] single or multi- aent solutions come [8:50] together to provide business outcomes [8:52] but the ability to orchestrate those [8:54] multi- aent solutions seamlessly is what [8:56] sets our platform apart and also LLM [8:59] optimization hub where we have to [9:01] optimize for speed cost and accuracy. Um [9:05] and enterprise governance is [9:08] non-negotiable. It's built in. So that's [9:10] uh that's one of our four pillars. So [9:12] these are the four key things that make [9:13] up our platform. But I I believe any [9:16] platform should should make up of made [9:18] up of those four things. But we do do [9:20] those four things quite well. [9:24] What I will do now is I'm going to um [9:26] talk about two stories, two business [9:28] impact stories um that'll bring things [9:31] to life a little bit more. The first one [9:34] is a complaints investigator that we [9:37] built for a large large wealth [9:39] management firm in the UK. In fact, they [9:42] the largest adviceled wealth manager in [9:45] the UK with over 170 billion funds under [9:48] management. And what they had was with [9:50] the advent of consumer duty regulations, [9:52] they were inundated with customer [9:54] complaints and they weren't able to keep [9:57] up with it. uh and it was really putting [10:00] their teams under tremendous stress and [10:02] also they were missing their SLAs and uh [10:05] they weren't compliant with regulatory [10:07] requirements. So we had to rethink the [10:10] way we help our customers. So what we [10:13] did was we reimagined the whole end to [10:16] end complaints journey using um agentic [10:19] AI. So the whole thing is multi- aent [10:22] solution deployed [10:24] solving a complaints problem end to end [10:26] from the moment a complaint comes in to [10:29] a resolution letter being generated. Now [10:32] this is in production. This is not a [10:34] pilot. We've been in production since [10:36] October last year. It's still in [10:37] production providing that business [10:39] impact for our customers. And what's [10:42] already is in 90% reduction in case [10:44] handling time. So it used to take about [10:46] 50 weeks uh 50 days rather and now it [10:49] takes about 20 minutes for them to to uh [10:52] to investigate one complaint and close [10:54] it. Also that's um given them at least a [10:58] 50% reduction in um in uh in cost. Um [11:02] but there's a lot of other benefits on [11:04] because of that as well. This was the [11:06] first solution for them from an agent [11:09] agentic AI perspective that has also set [11:11] the foundation for them to actually go [11:13] on and do bigger things after that [11:15] because they've gone through that [11:17] governance journey to get agentic AI [11:19] into their organization. [11:22] So the other story I wanted to talk [11:23] about is ESG edge. So we worked with the [11:26] world's largest sovereign fund uh a [11:29] European sovereign fund 1.6 trillion [11:31] funds under management. What they wanted [11:34] was they wanted to do ESG properly, [11:36] right? Like they did not want to do ESG [11:38] for the sake of ESG. They wanted to have [11:40] granular ESG data against the 9,000 [11:43] portfolio companies so that they can [11:45] integrate those data points into their [11:47] risk framework because they wanted to [11:50] make non they want to have non-financial [11:52] parameters to make investment decisions [11:54] alongside financial parameters. So what [11:57] we did was we built a a multi- aent [12:00] solution that looks through millions of [12:02] documents, in fact 10 million plus [12:04] documents against their 9,000 portfolio [12:07] companies and provides very specific [12:09] data points, AI powered data points from [12:13] all of those various unstructured data [12:15] sources and actually gives them the data [12:18] points to be integrated in their risk [12:19] framework. What this has resulted in is [12:21] their their analysts can now do thousand [12:24] times more analysis on on on their [12:27] portfolio companies when it comes to [12:28] ESG. So they haven't had to increase [12:31] their team size, but they've been able [12:33] to actually truly make a difference when [12:34] it comes to ESG. So they're now putting [12:36] pressure on their portfolio companies to [12:38] get better when it comes to ESG. So [12:41] we're really happy we're we're doing [12:42] this with them because this has made a a [12:44] small difference uh in the world and all [12:47] that why we're providing 90% accuracy uh [12:50] in fact more than 90% accuracy. [12:53] So what we learned from this um is it's [12:57] sort of it's it's a journey for for [12:59] these organizations to uh to deploy [13:02] multi- aent solutions but we've gone [13:04] through the journey and we've distilled [13:06] the things that is absolutely essential [13:08] for organization to be successful right [13:10] when it comes to risk management and [13:12] governance I think the the [13:14] non-negotiables are explainability [13:16] auditability and full traceability so [13:18] you need absolute data source lineage [13:20] especially in industries that are highly [13:22] regulated in and also you have you need [13:25] bias detection, toxicity detection and [13:27] prompt injection etc. all as standard [13:30] built-in and also you need model [13:32] oversight like I said earlier you know [13:33] the organization who wants to build [13:35] 300,000 plus agents and deploy them into [13:37] production the way they need to think [13:39] about model oversight is going to be [13:41] fundamentally different right they're [13:43] going to change their entire risk [13:44] function uh in fact we won't recognize a [13:47] risk function anymore right so these are [13:49] the things that we got out of that um [13:54] that exercise so this is what we learned [13:57] are the challenges um how we help our [13:59] customer through those challenges and [14:01] like I said um we it also meant that [14:04] these organizations once they've gone [14:06] through that journey their own internal [14:08] hearts and minds journey their own [14:09] governance journeys their own um journey [14:12] to sort of make sure that they are [14:14] comfortable with introducing AI into [14:16] their ecosystem while managing risk they [14:19] then now able to go on and do bigger [14:21] things so that wealth manager I [14:23] mentioned earlier now has deployed [14:26] an a chat agent [14:28] uh an AI powered chat agent where it it [14:31] provides the advice is an enterprise GPT [14:35] uh for them to be able to do um free up [14:39] them free up their time and and be more [14:41] effective with their with their uh with [14:43] their clients. [14:46] So I think the shift is imminent. I [14:48] think it's not it's in fact here. So I [14:51] think the key message is we need to sort [14:53] of make sure that the growth is built. [14:55] you know, you need to build the [14:56] governance through uh into your [14:58] workflows and make sure that that powers [15:00] your growth and it needs to be trust [15:03] trust driven compliance and risk [15:05] management. [15:07] And thank you very much. In fact, before [15:09] I close, I want to just quickly [15:13] highlight [15:15] um how long it took to deploy those [15:18] multi-agent solutions. So both of those [15:22] solutions we deployed from the [15:24] identification of use case to [15:26] production. The MVP was production in [15:28] under 8 weeks. So that's our promise. [15:31] The from the time we identify a use case [15:34] from zero to production is 8 weeks. [15:36] Fully governance ready and risk ready. [15:39] Thank you very much.