Markus Zirn

I've built neural networks back in the early 90s....back then we modeled 25 neurons :-) It does really help to understand an LLM "in the small". What took 30 (!!!) years was simply the computing power to scale this up to hundreds of millions of neurons. Now, the results are incredible, but feel almost incomprehensible. At the core, an LLM is a neural network, really a system of interconnected neurons which learns/gets trained by reinforcing the weights how signals propagate in the network. This is our current best understanding how a human brain works. At scale, it seems to now work pretty incredibly well (better than a human brain?). It's fundamentally different from how we used computers (deterministic programming) to date.

GenAI is really helping get rid of tasks that are "annoyances", like adding call summaries to the CRM after a customer conversation or the same after hiring interviews...Everyone knows from movies how much cops hate doing the report after solving the actual crime....GenAI is ideal to take this of our plates! Who doesn't love that?

In my experience, most Enterprise Software typically is still too complex and difficult. What differentiates Enterprise Software from Consumer Software is that it can handle all the exceptions, customizations and convoluted processes. The result: most Enterprise Software never gets fully adopted. It’s ultimately used in much simpler ways than originally designed for. When it comes to Enterprise Software product strategy, it’s all about avoiding complexity and simplifying relentlessly. Much progress has been made in the new world of SaaS. Much more effort is ahead of us.

In my 25 years in Enterprise software, I’ve learned two ways: a) Being hands-on, meaning you lead by example. Enterprise software, especially for a Platform company like Workato, is fairly complex. So you constantly have to teach and guide your team. If you can’t do it yourself, how can the rest of the company? b) Maniacal focus on customers. It’s all about customers. For an Enterprise Software Platform, everyone in the company needs to be curious about exactly what your product does for the customer. It’s all about curiosity about the customers’ business and customer outcomes. That’s the main source of inspiration for the entire company!

Time spent building your network is never time lost. Invest in cultivating relationships you already have but also spend time seeking out new ones. Some of my favorite low-lift ways to do this is are: 1. Staying active on LinkedIn. 2. Anytime I visit a new city, I’ll proactively seek out local connections and set up a coffee meeting.

AI projects require experimentation. As an IT leader, you will have to add some of the mindset of your data science colleagues to your repertoire to get the most out of GenAI. Before, executing a technical solution implementation was predictable and could be fully scoped. These were deterministic processes with clear rules. There typically was little surprise during the implementation. AI solutions are different. They will require quite a bit of experimentation. They require you to evaluate results and then tweak and adjust the solution - the data set as well as the type of LLM used. Fine-tuning is required. These are solutions that are probabilistic in nature (rather than deterministic). Your leadership style needs to adapt.