Agents@Work: Lachlan (Inside KPMG’s AI Workforce Approach)

Quick Overview

Lachlan Hardisty, Director of AI Lab at KPMG, explains that businesses struggle with AI accuracy not because of underlying models, but due to the interpretation and the lack of clear definition for what accuracy means for an agent, which necessitates a focus on traceable, explainable, and human-aligned outcomes rather than just model performance metrics.

Key Points: Businesses struggle with AI accuracy because they lack a clear definition of what accuracy means for an agent, often failing to define the intended outcome. KPMG's AI Lab focuses on the cutting edge of AI research and development, particularly in creating agents that can be deployed quickly and cost-effectively. The key differentiator for successful AI adoption is ensuring agents are traceable, explainable, and align with human expectations, rather than just relying on existing stock models. KPMG is actively involved in piloting and prototyping AI agents, including those for healthcare scheduling, and aims to be a leader in the agentic AI space. A major concern for businesses is the trust deficit in AI, where people fear job displacement, leading to hesitation in adoption. KPMG's approach involves clearly defining the desired business outcome and acceptable accuracy levels before implementation, contrasting with simply deploying existing models. The future trend involves AI agents moving beyond simple human-to-human communication simulations towards complex, multi-agent collaboration and automation of business processes.

Context: This video features an interview segment from 'Agents@Work' where Jacky Koh, Founder & Co-CEO of Relevance AI, speaks with Lachlan Hardisty, Director of the AI Lab at KPMG in Melbourne. The discussion centers on the practical challenges organizations face when implementing AI agents, specifically addressing the common hurdle of defining and achieving 'accuracy' in real-world business applications beyond simple model performance.

Detailed Analysis

Lachlan Hardisty, Director of the AI Lab at KPMG, argues that the primary reason businesses struggle with the accuracy of AI agents is not the underlying models themselves, but the interpretation of what accuracy truly means in a business context. He emphasizes that organizations often fail to define the necessary output or acceptable level of outcome for an agent, leading to disappointment. Hardisty notes that KPMG's AI Lab focuses on cutting-edge research and development to enable rapid prototyping and deployment of useful AI agents, helping large organizations shape and roll out their AI strategies. A significant concern he observes is the low trust in AI in Australia (and globally), driven by fears of job displacement, which slows adoption. He stresses that successful AI implementation requires clear articulation of the desired business outcome and acceptable accuracy thresholds, rather than just plugging in pre-existing models. The goal is to move towards systems where agents are traceable, explainable, and integrated holistically into digital and workforce strategies, ultimately leading to more efficient processes like contract negotiation and procurement. He concludes that the future lies in complex, multi-agent systems collaborating effectively, rather than simple, isolated agents.