SMEPerspectivesSystems7 min How a COO is thinking about AI
I have a new obsession when it comes to AI. It’s boring, it’s mundane and it drives those around me insane. But it does solve a problem every business experiences. When I hear people talking about AI, it’s often about what it can do . The actions it can take on a person's behalf. It can move information faster, respond more eloquently (with less personality) and perform research at a pace and to a standard that makes me grateful I’m not a researcher. Before we dig into what I’m obsessing about within the realm of AI, it’s helpful to separate what we mean by “ AI ”. AI has become synonymous with two things; doing something autonomously, faster and access to knowledge. If we “ put ” AI in something, we’re talking about an agent. An agent can access specific data and systems and perform tasks, based on a series of instructions it is given to follow. When we “ use ” AI, we mean we’re talking to a chat bot which has access to specific data and generally the internet. When spoken about, we often blur the lines between the two and use the catch-all term of “AI”. In either case I’ve noticed a common thread between them and it’s what I’ve been obsessing over. It all comes down to data - how it’s structured and how it’s interpreted. Instructions If you’ve used ChatGPT, you too will have been impressed by its ability to find the information you’re after almost instantly. It’s wonderful! But it’s not magic and it’s not an accident. There are a series of instructions it follows when it receives the prompt you enter. It doesn’t simply slip its hand into the bucket of the internet and retrieve at random the information you require. Have you ever asked a chat bot to scan for information from your own files? Not quite as wonderful! In my experience, it’s hit and miss. It can find information, but it doesn’t know how to weigh it. Should it prioritise older or newer files? Are files written by specific people more important than others? Which files are the source of truth? And then there’s the more standard questions it doesn’t know intrinsically. Where should it search? How should it present the information? From what perspective should it analyse the data? Before you know it, you have a prompt so detailed it was probably easier to just find & analyse the file the old school way. AI needs to know the structure of the data and how to interpret it. You can have unstructured data, but it will reduce the effectiveness of any instruction because it will lack consistent rules to follow. When giving someone directions, saying, “ take the first left, over the bridge and then it’s on your right-hand side ” gives them a better chance of finding what they’re looking for than “ it’s that way ”. Obsession Every business I’ve worked in, the biggest cost is loss of information & institutional knowledge. It’s either lost as people leave or is forgotten over time. At turntabl, we realised the lessons of 6 years ago had been forgotten & we had to teach them all over again. Not anymore. If I’m able to structure data and communicate how it should be interpreted, then the knowledge can become accessible by anyone at any time. They won’t be searching for files, they’ll be able to talk to a bot to search for how to make a decision . One of the least teachable skills I’ve observed is decision making, and upstream of that, judgment. These two skills are what you find in founders, and it’s the reason they struggle to remove themselves from the business. People just can’t seem to make the same decisions as founders. And if they do, they’re not often as good. That doesn’t make employees incapable, it just means they haven’t had the same experiences the founder has. Founders make decisions which on the surface look hard to make, but normally they’re pretty straight forward when you’ve had their experiences and all the historical context. What if you could log why every decision was made? Up until recently you’d have to do a retrospective - what worked well, what didn’t go well, why did we make the decision we made - and then we’d have to collectively remember the conversation. 3 years later when the same event happens again, the lessons of the past have faded ( speaking from experience! ). Now you can record a call, store the transcript and the decision is remembered forever. What if you had everyone in the business thinking like this? You can institutionalise everybody’s knowledge and their company-specific experience. No more key-person dependency. No more knowledge hoarding. No more making the same mistake twice. And that’s just one use case. Here are a few other thoughts and ways I’m using AI at the moment: For personal use I need to store all my data in an easy to access, centralised location. Google Drive is more accessible than one drive because of its API. The problem here is file types ( there are issues reading google docs / word docs across AI systems ), so I’m going to either need to save documents in text format, or double down on Gemini. Does Gemini have the functionality of chatGPT? Am I using the right tool for me? ( I’m using both at the moment, this is an ongoing challenge! ) I can straddle multiple systems, but it’s detrimental in the long-term because I will have fractured data all over the place. I need to choose. Or I can choose not to choose, but the payoff will be increased costs and reduced functionality. For business use How do I increase the chance of being understood? I’ve been working cross culturally over Teams for almost a decade. That still doesn’t mean things don’t get lost in translation, especially when I speak a little fast and the internet plays up. You can lose a lot of information in a few seconds. I’ve been pushing everyone to put the call transcripts through Copilot - ask Copilot what you’ve been asked to do, whether you’ve interpreted it correctly, what you’ve misunderstood and why. This is incredibly powerful. How much information is lost after every call? I spend most of my calls either ideating and fleshing out ideas or coaching others. Can I use the transcript data to challenge my ideas and to improve my coaching? I have all this data, but if I do nothing with it then it has no long term value. How can I capitalise on the richness of every call? I’m not just asking others to get feedback from Copilot - I’m doing it myself. I’m getting it to critically challenge my thinking and to identify common themes. I’m asking it if I was speaking clearly, and if I wasn’t then where are the gaps ( I had a call recently where I was adamant I gave a clear, specific instruction. Turns out I was completely wrong. ) It’s important to note that if I’m not specific with my instructions to my AI chatbot of choice, its natural instinct is to praise me for my genius and coaching abilities. Good if I’ve ever got low self-esteem, not ideal for progress. These are just a couple of ways I’m thinking about AI for personal use and in business, specifically in my role as COO of a company with over 100 employees. I’m currently going deep on the loss of information, which is why all of my thinking is centred around data and the interpretation of that data. Workflow automation has been common for many years, which means it’s already in the minds of most employees. Company-specific knowledge loss is a blind spot I'm noticing, which is why I’m banging the drum every day. I’ll be sharing the details of specifically how I’m using AI in a coming article. It might help you think about it in a different way and hopefully solve a recurring challenge you’re experiencing in your business right now. More to come, Stoirm PS If you hear a lot about AI but don’t know where to start in your business, drop me a message and I’d be happy to talk it through with you
