Back when I was studying computer science, there was a lot of focus on deterministic algorithms – algorithms that, when given a particular input, will always produce the same output, with the underlying machine always passing through the same sequence of states.
In this bold new world of Agentic AI, this is certainly not the case, as autonomous AI agents adapt and learn to consistently improve outputs as they grow and develop. This really provides food for thought in terms of non-deterministic business process automation and what may be possible.
SuperDataScience Podcast from Jon Krohn: Agentic AI – with Shingai Manjengwa
As part of my role at jaam, I work closely with our CTO, Graham Penman, to direct our team on how we can deliver Agentic AI to our customers, as well as leverage its benefits internally, and I like to keep my skills and knowledge current with hands-on technical work, and continuous learning.
As an avid podcast listener, I was delighted to find this Agentic AI with Shingai Manjengwa podcast late last year. Not only is Shingai a great speaker, but they really know their stuff. If you have a little time, it’s definitely worth a listen. For me, things start to get really interesting around nine minutes, thirty seconds.
My top three takeaways
It would be incredibly boring if I agreed with everything in this podcast – and I don’t – but there are some key topics which I find particularly fascinating:
- The premise that Agents are best when they have a key skillset to focus on and that the multiplier is enabling agents to seamlessly work together. This is critical to how Agentic AI can, and will, add value into business process automation. The approach to achieving this is one of the core topics that we are working hard to address with our delivery teams.
- The importance of auditing key inputs and outputs from an agent. For example, what were the prompts and what were the responses that were used to carry out actions. In the podcast, blockchain is discussed as being used for some of this and while this sounds very cool for some use cases, it won’t be needed for most. However, it’s essential for business processes that this level of auditing is in place with Agentic AI.
- The mantra that we need to be continuously testing and validating the agents we use. I one hundred per cent agree with this, not only because it’s the responsible thing to do, but also because underlying LLMs and surrounding tech is evolving so fast that we need to make sure to stay aware of how existing Agents are changing.
It’s a super exciting time in the AI world and AI agents have the potential to make a huge impact on business process automation. Keep watching this space.
PS – I mentioned I was a committed podcast listener, so thought I’d share one of my favourites. It’s The Infinite Monkey Cage and it has absolutely nothing to do with Agentic AI. I just enjoy it 😊