AI might be the biggest buzzword in tech - but without process context, it’s just guessing. AI is like a GPS, it only works when it knows the road ahead. In this episode of the Trust the Process Podcast, we explore how Process Intelligence gives AI the context it needs to operate in the real world - making it not just smart, but business-savvy.
The following is a transcript of the podcast, edited and organized for readability.
I’m 22, wearing the most “consultant” outfit I can imagine — blazer, blouse, and a pair of high heels I definitely can’t walk in. I’ve got my first-ever consulting interview with Bain & Company in Munich. For me, it was the dream I’ve spent weeks prepping business case riddles, behavioral questions — you name it. I’m nervous but ready.
I live just outside the city, so I take my old Toyota — a car with over 200,000 kilometers and an old-school navigation system — the kind that would tell you where to go, but not really what you’re heading into. And what I’m heading into… is traffic.
Now, I did leave early, but apparently not early enough. It usually takes me about 30 minutes to get into town and now that 30 minutes turns into an hour. And I’m stressed and late. I’m starting to sweat through my wonderful outfit, I’m getting more angry with myself by the minute, and I am watching the minutes slip by. Everyone who has already been late for an interview — I think you can feel with me right now.
I do eventually find parking, but I’m already twenty minutes late. So I sprint through the streets of Munich. I reach the Bain reception, panting. They did somehow still let me in, but the interview was a blur of stress and panic. I was definitely not Bain & Company material that day.
I don’t know why, but somehow I still got the internship. But that moment stuck with me.
Because today, things are different. I still have interviews — but now I have Google Maps. A living, breathing system that knows where I’m going and what’s in my way. It doesn’t just give directions — it gives context. And that’s exactly the shift happening in business today.
AI used to be like my old-school navigation system — static. Powerful, sure — but blind to real-world complexity. Now, thanks to Process Intelligence, AI can operate more like a modern GPS. It’s dynamic, predictive and aware of everything happening around it.
In this episode, we’ll explore how Process Intelligence fuels AI — giving it the context it needs to work in the real world. Because without it, AI is just guessing. But with it? It knows exactly where it’s going — and how to get there.
Before we can understand why AI needs Process Intelligence, we need to get clear on how AI actually works. AI doesn’t see the world the way we do. It doesn’t “think” in a human sense. It operates on data — structured, unstructured, historical, and real-time. Some AI models recognize patterns. Some generate new content. Some make decisions. But fundamentally, they all learn from the relationships buried in data.
Steve Brown: "So AI comes in many different flavors, and what's really defined the last 10 years is the emergence of a new category of AI, which is generative AI. Before that, AI was known as discriminative AI, which doesn't mean it discriminates against people, it means it's good at sorting data into groups. So that was great for image recognition and audio recognition, those sorts of applications or spam filters."
So one learns to classify. The other learns to create. Discriminative AI sorts and classifies — it sees a dog and knows it’s a dog. Generative AI, on the other hand, can create something entirely new — it can paint a picture of a dog that never existed before. But here’s the twist: even generative AI doesn’t understand what it creates. It doesn’t know a dog is furry or loyal — it just knows what the word “dog” typically appears next to.
Steve Brown: "Fundamentally, if you look at the heart of a generative AI, it's using a thing called a transformer… and a transformer is just guessing what should come next. It's a really good guesser and it's a very sophisticated guesser. I think we've built AI that understand language. They're now starting to understand video feeds input so they can see and hear and understand the world on that level. We now need to teach 'em what's called world models, so how the physics of the world works. And then a next logical step is how companies, organizations work, how groups of people work together."
Think back to when you were a kid learning the difference between a cat and a dog. You weren’t given a rulebook. You just saw enough of each — pointy ears, wagging tails — and slowly, your brain built up the pattern. Language worked the same way. You didn’t just memorize words; you absorbed meaning from context. Now imagine we have to teach that to a machine.
Steve Brown: “So it is through the training process that these AIs start to find structures in the data and that is expressed in tokens. So it's captured in what's called vector space. So a token is a set of lots of numbers. It could be up to 13,000 numbers that represent that concept. So I'll give you an example stored in imagine of 13,000 dimensions, high dimensional space. You have the word king and the word queen. The direction between those in this 13,000 dimensional space, which just makes my brain hurt, the direction between them is the same direction as between man and woman because it's showing this is a relationship between them with gender also the direction between king and man and queen and woman is the same direction. It shows royalty. So you start to encode the relationship between concepts using this space, and that's how these generative AI work.”
You explain that a ‘king’ is to a ‘queen’ as a ‘man’ is to a… you guessed it… woman? That’s how AI tools like ChatGPT learn — by spotting patterns and relationships in language.
But here’s the thing: AI doesn’t know anything. It’s just running the world’s most advanced guessing game, predicting what comes next based on stats.
Think of it like asking for directions and being told, “Just head northeast.” That’s called gradient descent — a technique to get closer to the answer. But it doesn’t know the road’s closed or only goes one-way. AI needs more than guesses. It needs context if it’s to be effective. And that’s where Process Intelligence comes in. Just like Google Maps adds street names and traffic data, Process Intelligence gives AI the insider knowledge on how your business actually runs.
Here’s where things get fun. Because Process Intelligence is like giving your AI a map, traffic updates, and a co-driver who knows the shortcuts. But at first, the connection between AI and Process Intelligence isn’t obvious. It’s not a straight line — it’s spaghetti.
Process Intelligence uses process mining to build a digital twin of your business. It spots inefficiencies, adds context, and codifies that data for AI. Without it, AI is like a GPS with no roads.
To see what this looks like in the real world, we spoke to Sachin Chauhan, AI evangelist at Microsoft, who recently presented his work to the United Nations.
Picture this: a major European grocery chain wants to personalize the shopping experience — online and in-store. But customers are showing up in different systems under different IDs. Fragmented data. Messy journeys. No clarity.
Sachin Chauhan: “They were able to connect this anonymous customer to the known customer, which means they have to really understand who is this customer, and they were able to have to impact, number one, gain new customer based on the real-time personalization of offers or the coupons, which is contextual to that person and the geography. Secondly, once you know the customer, it is easy for you to also cross-sell if it is already an existing customer, you already have a lot more information about this person in your customer 360 degree platform. Once you're able to contextualize your offers, recommendations as per the customer needs, which is based not on your intuition but actually based on the data you already have, I think it gives rise to more sales and improved customer satisfaction.”
They used Process Intelligence to map out the customer journey and teach the AI what a “customer” actually looks like — across 100+ systems. With Process Intelligence input, they were able to forecast demand and cut inventory costs by $3 million, just by helping the AI understand when and where products were needed.
Sachin Chauhan: “This is one of the biggest retailer in Europe, and they were struggling with the customer experience and customer personalization. They were not able to map out the clear customer journey and they were not able to understand why they were having a huge amount of customer churn and the satisfaction score of their end user base is also going down. And that indirectly impacted their revenue. So this customer has decided to take up the process intelligence as one of the platform from Celonis to actually leverage it to connect with the different systems.”
AI agents need contextual input that can only come from the company and its processes.
Sachin Chauhan: “You need to first capture that information and then map it with the data you are already storing. Which means you have to first understand and connect the dots to create a golden record so that you understand who is this customer, and then you combine the information you are already holding with this customer. If this is a known customer and you find out in your golden records, in your customer data hub that this is the customer I already know, I can really push more contextual and personal recommendation or coupons, which are more making sense for him to buy, him or her to buy.”
Even more than that, Process Intelligence doesn’t just teach agents about customer behavior — it also adds critical information around things like stock levels, delays in delivery channels, and staffing shortages.
Sachin Chauhan: “Also another example which is with one of the biggest retailers, their challenge was the leakage in their processes. So process intelligence system from Celonis was able to help this retailer cut down on the inventory so they don't have to over provision. They were able to forecast their demand properly and they were able to cut down the additional cost, which in revenue is around $3 million for them, saving $3 million for them by just having a proper inventory fit for their demand so that they're not over maintaining or under provisioning.”
Now, imagine this AI agent: it recognizes that Sachin shops every Friday. It knows if his favorite brand of hummus is in stock, if the delivery route is reliable, and if he’s a high-priority customer.
And it doesn’t just report this. It acts on it. In real time.
Sachin Chauhan: “They have executed it for more than eight to nine months, and eventually that has given rise to the satisfaction score has gone up by two times, and their customer experience has also increased to a very satisfactory level. They also have been able to save cost on the manual effort while they have it in place, so they were able to save at least two million euros in cost by automating the manual efforts.”
This means AI can now make sense of a company’s unique processes with the context around it. It doesn’t just guess. It connects raw data from ERP, CRM, and supply chain systems — then translates it into a business-readable format. So instead of seeing table names and IDs, AI sees “invoice overdue” or “shipment delayed.” It finally understands how your business works. It’s not magic — it’s structured process knowledge.
AI can’t fix what it doesn’t understand. Without insight into how your business runs — its logic, flow, and rules — AI is flying blind. And that’s risky.
Process Intelligence gives AI the map. It adds structure, timing, and context — transforming your AI from a smart assistant into a true operator. One that understands not just the data, but the business behind it.
As generative AI evolves and companies build intelligent agents, one thing becomes clear: the future of AI isn’t just about thinking. It’s about knowing, and doing.
