Can we talk about AI without talking about AI? Here we go. While traversing Germany, I booked a hotel for the night which fitted my needs: outside a city and, most importantly, self-check-in, as I was planning to stop at midnight, have a rest, and continue. Of course, there was a surprise. At 12:30 a.m., when I was trying to unlock the room with the keyless system, the lock wouldn’t budge. I checked with another guest that my approach was right; the lock was just not reacting. In such a situation, the fallback is to call a support number, wait on the line for 20 minutes, and get the issue resolved. The catch? The support line’s working hours are from 9 a.m. to 6 p.m. And if you have a problem, that’s your choice!
We can disentangle this situation into two - a technical and a business problem. On the technical side, even a junior knows that systems crash, hit a bug, or “misbehave.” Therefore, you must have a workflow for dealing with such situations. If you are running a sensitive, customer-facing system, then you have to have someone on-call for an overnight shift, which is definitely an extra expense. Secondly, you definitely want to capture such events, analyze the cause, potentially interview angry customers, and resolve the issue, which in turn will reduce on-call expenses (here I assume that there are two different rates for active and passive on-call duties). For example, Amazon has a Correction of Error (COE) process, which captures what happened, the impact on customers and/or the business, the root cause, the actions taken, and the lessons learned. Now, if you shut the blinds on your support at 6 p.m. while customers are actively checking in outside your support hours, sure thing — you are saving a lot of cost. And that brings us to the business problem.
Have you tried to book a hotel at 1 a.m.? If I ran a hospitality business, I would expect two types of customers at such hours: a hedonist type, very likely, and someone dumped by another hotel, very unlikely. As a business owner, you might think that you want to strive for a balance between your profit and customer satisfaction, which I now believe leads to mediocrity. You see, you can stomach some negative reviews and pay Booking.com or Google to be less affected by them. However, I really like the idea described in “Badass: Making Users Awesome” by Kathy Sierra, which says that your business should make the customer awesome or enable them to do epic things. A happy customer will share why they are happy or what makes them awesome, which will lead to a nice growth of your business. And letting a customer sleep under a bridge is actually the opposite.
The biggest marketplace in the world is actually driven by a similar idea — customer obsession. The customer is the center of the business, and everything is made to make it feel like an epic journey: you don’t like the product, we are happy to take it back; was it damaged or too late? No questions asked, we are happy to solve that. You might already be spoiled by their service, but believe me, some marketplaces make it an Everest to resolve any of these issues.
Now, Sam (OpenAI/ChatGPT), Dario (Anthropic/Claude), etc., are selling a fantasy that you can ignore the exact problems described above because an army of agents will take care of it, leaving you to just dream about how to make money and consume more agents. My bet is that if we suck at building businesses involving rule-based systems, we are going to multiply those failures with GenAI. An interesting nugget about the difference is this: in a rule-based system, programmers describe all potential outcomes. Any unwanted behavior is a bug or an exception, making the system’s behavior deterministic. With smart programmers and the right methodology, those errors can potentially be eliminated. But GenAI is a probabilistic model, so by its very definition, it will cause issues in a small number of cases.
I’m not a GenAI denier or an AI doomer, but I want to sell a different view of the near future—let’s say 2-10 years ahead. For the builders who can utilize GenAI tools, incorporate operational excellence practices into the process, and build an epic experience for their customers, this is actually the golden age. Sure, you might not be the subject of a flashy news story, like the ones about interns getting $100M+ offers, but businesses are ready to compensate you dearly for running their business flawlessly.