In Australia, 2023 was a year of excitement in generative AI, but in 2024 the work is shifting towards discovering practical opportunities to leverage this technology in our daily business lives. While experiments have begun in earnest, we have seen the importance of understanding a company’s readiness for different levels or tiers of generative AI integration.
By considering what depth of integration into our workflows and data systems we are aiming for, we can better understand what’s possible in the short term and what needs to be done to achieve long term value.
We can view generative AI at three levels of integration today.
Level 1: The basic tools and services publicly available today, such as ChatGPT, Gemini, Claude, and Copilot for Microsoft 365. These generative AI tools have no explicit integration into our private business systems with minimal customisations or extensions to meet our specific needs.
Level 2: These are the generative AI solutions that offer dedicated capabilities for specific tasks, such as Copilot for Github or Atlassian Intelligence. These move away from general utility towards offering help to specific roles within the business. At this stage, these have been typically focused on engineering work given the performance of generative AI in supporting and reviewing code development. However, this is an area where we are seeing significant customer demand as workflows are being enhanced with Copilot Studio.
Level 3: These are generative AI tools that have been deeply integrated with corporate knowledge bases to provide new foundational AI capabilities across a variety of organisational needs. This could be a bank building a knowledge base to support its contact centre teams or some other dedicated, private solution for an enterprise to turn sections of its extensive corporate history into a dynamic intelligence resource.
We can see how the three levels speak to a depth of AI integration, with any given organisation potentially operating with tools across multiple levels to meet the needs of different areas of the business. The deeper you go, the greater is the potential for truly powerful enhancements to business outcomes. But the work required also escalates quickly on the road to those improvements, as does the work needed to maintain dynamic insights over time.
Most of us have experienced Level 1 systems at this point. We see the potential for productivity benefits, but as these tools are often minimally integrated into our workflows and data, they are quickly forgotten after the initial novelty wears off. Or for users who try to embrace Level 1 tools more deeply, the limited integration can become a point of friction and frustration.
The integration of Level 2 in financial services, particularly through tools like CoPilot Studio, marks a significant advancement in streamlining complex workflows. This level of automation sees a digital assistant that not only summarises positions but also enhances banking services. Predictive responses and the ability to sift through extensive documents for inconsistencies are just the beginning. Our FastStart workshops play a pivotal role in this process, gathering role-based personas to tackle specific tasks and providing invaluable insights. These workshops, coupled with client consultations, and engagements reveal immediate benefits through structured implementations.
Arguably, Level 3 systems hold the greatest potential for boosting efficiency for diverse and specific needs throughout an enterprise. However, the journey to Level 3 generative AI is contingent upon extensive data cleaning to ensure the quality and readiness of data for advanced applications. This strategic focus on data quality can provide organizations with a significant competitive advantage, leading to transformative outcomes in the dynamic landscape of financial services.
Should you buy a Level 1 Copilot license for all your employees? How are you sure they will use it? What value will it drive? Or will it be hundreds of thousands of dollars per month doing nothing in particular? But what about buying Copilot for Github for your QA team? Or dedicated AI research tools for your legal team? Specificity matters, and we’ve seen these use cases come to life when they are thoughtfully targeted.
Like every digital transformation discussion, integrating any level of generative AI into business operations is not a one and done program. It’s a new cycle of adaptation and enhancement, with a practical value that relies on constant review and iteration with users and customers at the core of the process.
It’s essential to look at what level you are working at today, and how you are approaching the structures that are required to safely put generative AI in place to deliver the benefits.
