When I first used a generative AI chatbot over the holiday break at the end of 2022, I was astonished at its capabilities. The world’s information at my fingertips with near instantaneous responses personalized to my requests? It was like I was rediscovering the internet!
Upon returning to work that following January, I thought every knowledge worker in the world would be impacted by this technology. So, I reached out to our CEO to ask what our plan was and to offer my help in navigating the opportunities and risks these tools posed. Mind you, prior to this, my experience with artificial intelligence was informed mostly by sci-fi movies and the generally frustrating interactions I’d been having with my voice-activated smartphone assistant.
I was genuinely curious about how we would adapt our business in the post-generative-AI era and how it would change not only how we serve our clients but also how we work generally. Happily, our CEO was entirely on board with my offer and encouraged me to put together a team and get to work developing our own platform at KPMG in Canada, which we call “Kleo.” My interest in writing this post is to help you start thinking about how you can do the same.
What’s in a name?
Kleo is powered by our alliance with Microsoft and built on their Azure OpenAI cloud. It’s a suite of chatbots that harness the capabilities of large language models (LLM) to help improve our firm’s productivity, enhance our employee experience and inspire innovation. Launched in December 2023 to our more than 10,000 partners and employees across Canada, Kleo features a generalized chatbot (built on GPT-4), along with specific functional chatbots for HR, IT and Risk Management that answer our employees’ most common questions—things like “How do I setup my printer?” “How many vacation days do I have left?” “Am I allowed to deliver this service for an audit client?” And so on.
We are adding features that support new use cases every few months—including most recently a bot to help accelerate the development, and ensure the consistency, of our proposals.
Fundamentally, though, when we built Kleo, we knew we needed to do it in a way that upheld our commitments to our clients to keep their information safe and secure. Trust and integrity are what KPMG is all about; when we go down a path of technology innovation, we do not compromise on these principles.
Generative AI, of course, was (and is) still new. To navigate this space appropriately meant we needed to assemble a team of specialists across our organization that could balance the desire for cutting-edge innovation with pragmatism. This cross-functional team comprised the following domains and skillsets:
- Technology development – These are the individuals that understand innovative technologies and how to build and implement them
- Legal and privacy – Legal and regulatory uncertainties around generative AI remain, not least as they relate to privacy, so we needed to be sure the business was clear on how to mitigate them
- Risk management – Similarly, we could not take any chances that what we built was out of compliance with our terms of client service
- Business representatives – Engagement with key members of each of our core business functions (Audit, Tax and Advisory) was necessary to ensure we were adding value to our client service teams
- Security – The highest standards of information security that we use to protect client information had to be met or exceeded
- Change management, communication, and training – While Gen Z may find the use of generative AI tools intuitive, successfully rolling out something like this to a diverse employee base of more than 10,000 people meant developing a plan that would be meaningful to as many of them as possible
- Analytics and reporting – The team that will help us measure Kleo’s value
- Project management – With so many moving parts, successful completion, delivery, and ongoing development is dependent on dedicated project managers.
We are fortunate to have all these capabilities in-house, and each will be essential to the success of your own generative AI program, as well.
Surface, scratched
For us, the story has just begun. But I can tell you that everyone we talk to about Kleo is keenly interested in creating their own version. After all, the belief that AI is going to fundamentally change the way we work has taken hold, even though a lot of questions remain. “How and where do we start?" “What are the risks and who needs to sign off on them?" "How much will it cost?" "What does ‘great’ look like?"
I know what you’re thinking: These are hard questions, especially for larger organizations. I can confidently say that we understand. We’ve learned so much through our own experience and, with blog posts like this one, are in the process of packaging up our insights to help our clients tackle the same challenges and accelerate their own development.
As of right now, Kleo is fulfilling its mandate and more than 60 per cent of our people are using it regularly. When we started down this road, the goal was to provide a safe and secure way for our teams to leverage generative AI. Where we are heading is toward a future where the way we work will be completely reimagined. I couldn’t be more pleased and will have a lot more to say about this and related issues in coming posts, so please stay tuned. In the meantime, don’t hesitate to reach out with your thoughts and questions.
