Skip to main content
Resources

      AI agents, or reasoning machines, will become integral to every organisation’s technology ecosystem.

      AI has evolved from basic model-based frameworks that could complete small tasks using rules and logic, like playing chess or solving an algebraic equation. They were fast but had limited scope or flexibility. Early AI operated within predefined rules and couldn’t learn from new data or environmental changes.

      Thanks to machine learning, AI agents can now process complex problems to amplify human potential in new and innovative ways. We now see multi-agent systems that can communicate, collaborate and even adapt in real time. They mirror the dynamics of human teams. 

      AI agents: the dawn of reasoning machines

      Our report explores the evolution of AI from single task automation to complex reasoning and outlines how our framework can prepare your organisation.
       

      Download report

      Agentic AI – the evolution from execution to reasoning

      Complex reasoning machines are comprised of AI agents that can understand context to reason and make autonomous decisions. They use sophisticated machine learning approaches like deep learning, natural language understanding and advanced language algorithms. And they can navigate complex scenarios with the cognitive maturity that resembles the human thought process.

      Their potential spans domains from health and finance to logistics, retail and beyond.

      • Transforming work

        We’ve seen AI agents fundamentally improve organisational performance.

      • Project tracking

        These agents monitor project status on aspects including status, budgets and timelines. This enables better resource management and outcomes.

      • Managing budgets

        Agents can review outstanding purchase orders and manage financial plans. This helps organisations make informed financial decisions and maintain their financial health.

      • IT helpdesk

        AI agents can answer technical issues efficiently. This reduces downtime and improves productivity.

      • Employee self-service

        AI agents can give information about company policy and processes, such as annual leave requests. This simplifies human resources and improves the staff experience.

      • Software testing

        Agents can create test data and test applications based on acceptance criteria to find errors. This improves quality, reliability and the user experience.

      KPMG’s Generative AI Maturity Pathway

      Our structured roadmap is for all organisations, whether you’re exploring AI for the first time or looking to mature your practices. The Generative AI Maturity Pathway has five progressive stages from exploration to optimisation. We designed it to help organisations release the potential of AI with the confidence that the highest levels of accuracy, security and efficiency are maintained.

      • Explore

        We use the chat interfaces of public large language models (LLMs) to test value and identify appropriate use cases.

      • Experiment

        This stage is a playground for using services like Open AI, Bedrock and Gemini. Experimentation helps refine your AI strategy and develop prototypes.

      • Foundation

        Here, we develop foundational production applications using LLM services. It could include virtual assistants or integrating AI into key applications. The focus at this stage is on accuracy and security.

      • Scale

        This stage creates tailored user interfaces and embeds AI workflows into core business processes. We focus on monitoring use and performance to drive innovation and efficiency.

      • Optimise

        This is the final stage in the pathway, delivering AI outputs with exceptionally high accuracy and efficiency. There will be robust and scalable AI infrastructure that supports a very wide range of use cases that deliver significant business value.

      How we can help

      KPMG Australia has helped some of our largest and most complex clients save the ‘trial-and-error’ and leap from Level One of the Generative AI Maturity Pathway to Level Five in under 12 weeks. Our proven pathway helps organisations establish production-grade, ready-to-scale AI capabilities and agent libraries to unlock the potential of AI in a structured, secure and safe way.

      Contact Us

      AI agents: the dawn of reasoning machines

      Our report explores the evolution of AI from single task automation to complex reasoning and outlines how our framework can prepare your organisation.
       

      Download report

      Get in touch

      Related AI insights and services

      We put people, trust and governance at the core of AI – helping organisations accelerate value with confidence, using our Trusted AI framework.
      You can with AI

      Learn how we navigated risk to design, build and deploy KPMG’s Trusted AI agent – KymChat – within months of ChatGPT’s launch.

      Generative AI - Changing the game with KymChat

      FAQs: Agentic AI

      Agentic AI uses several AI services that work together to make decisions or perform complex tasks. They can reason, plan and learn from new information and external systems.

      There are many types of AI. Generative AI is currently used widely to create new information based on learned patterns. Agentic AI can make autonomous decisions to achieve specific goals.

      AI agents will fundamentally change how businesses operate. The main benefits are increased efficiency, enhanced decision-making, cost savings and personalisation.

      We suggest organisations consider the following steps to ensure they get the most out of AI agents:

      • Invest in tools and training
      • Implement robust data security
      • Understand the regulatory landscape
      • Experiment with AI models
      • Create a culture of innovation