Sobotni wieczór: dłoń z pilotem, miska popcornu i telewizor z niebieską wizualizacją technologii

Marcin Kujanek

Popcorn and a remote. Three tech channels of the week worth flipping to

Saturday, after eight in the evening. Too late for coffee, so there is a bowl of popcorn on the table and a remote in my hand. This week in tech was so packed that it is hard to watch it all. So I am flipping the channels for you. Three channels, three stories and one thought that stayed with me after the last click.

Get comfortable. Here we go.

A graphics card with blue lighting next to a deck of playing cards, an empty executive chair behind

Channel 1: The News. Nvidia's CEO leaves the stage

On 12 November Jensen Huang resigned as chief executive of Nvidia. He co-founded the company in 1993 and turned a maker of graphics cards for gamers into one of the most important companies of the AI era. Behind almost every model you read about in the news, somewhere, Nvidia chips are doing the work.

The reason for his departure is unusual for an industry that normally says goodbye to CEOs with talk of "a new chapter". Huang admitted to a relationship with a person whom US authorities suspected of passing information to Chinese intelligence. The board met late into the night. CFO Colette Kress is now interim CEO, and for the first time in its history Nvidia has a board chair, Harvey Jones. On Thursday the shares dropped sharply at the open, but recovered most of the losses by the end of the day.

For more than thirty years, Jensen Huang was the one dealing the cards in this industry. Quite literally, graphics cards. This week, the board dealt the cards.

And a note for screenwriters, who have been told for two years that AI will take their jobs. Relax. The company that powers practically every AI model in the world got a script this week that none of those models would ever have come up with. AI writes better and better, but "real life" still has the stronger writers' room ;)

The lesson is more serious than it looks. Even the most powerful technology rests on people and their decisions. No chip will change that.

A business owner looks at a tablet showing a verified company profile in ChatGPT

Channel 2: The Shopping Channel. Your company just got a listing in ChatGPT

Flipping on, because this one is about your company.

On 10 November OpenAI launched business profiles in ChatGPT in beta. A company can claim its profile through the Business Hub, verify it and get a blue check. It can also update its own information, such as prices or product availability.

The most interesting part is something else. The Business Hub shows how often ChatGPT mentions your company, how many people click through from a conversation to your website and which topics people ask about. Until now this was a black box. We knew customers ask AI assistants about suppliers, but we did not know what answers they got.

There is a catch. Business agents, which answer customers in ChatGPT on behalf of a company, are available at launch only to selected US businesses. The profile itself is still a good start. More than a decade ago every company set up a business listing on Google. A similar story is starting now, except that on the other side there is an assistant talking to your future customer.

I tried it on us straight away. On 12 November I submitted 360 Software House in the Business Hub. ChatGPT recognised the company and the website, and we landed on the waiting list. I will let you know what happens next.

A small test for tonight, between two handfuls of popcorn. Ask ChatGPT: "What do you know about [your company name]?". The answer may pleasantly surprise you. Or give you something to think about.

A company owner and a production manager check one shared system on a single screen, printouts set aside

Channel 3: The Documentary. Five apps versus one system

To close the evening, something that genuinely pulled me in. A study that sounds like a technical curiosity, but in fact tells the story of almost every company I know.

On 11 November Anthropic's research team published the results of an experiment called "Worse Together". The researchers looked at what happens when several people share one resource: a budget, a calendar of appointments or a group food order. They compared two setups.

  1. Everyone has their own AI assistant. The assistants can message each other and agree on what to do.
  2. One shared assistant serves everyone. It sees the whole picture and makes decisions for the group.

The result? The shared assistant won in all four tested situations, by roughly 10 to 45 percentage points. In a group order with a shared budget, one assistant reached 88% of the best possible outcome. A team of separate assistants that could talk to each other reached 44%.

The most interesting part is why. Each assistant saw only its own user's information. It filled in the rest with guesses. Sometimes it placed the order before reading all the messages from the others. An important request from one person did reach the group, but at the moment of decision it was no longer in front of whoever clicked "order".

Sound familiar?

To be fair, this is a study with simulated users and it is about AI agents, not spreadsheets. But I know the mechanism from companies very well. Only instead of AI assistants, there are apps.

  • Sales has its CRM.
  • Production has its Excel sheet.
  • The warehouse has its own program.
  • Accounting has its invoicing system.
  • And everything is still held together by an inbox and a phone call to Barbara, who "knows where everything is".

Each of these apps does its own job well. The problem starts between them. The information exists somewhere in the company, but not where the decision is made. Someone copies data from one place to another. Someone decides based on yesterday's status. Someone guesses, because there is no time to look. Exactly like the assistants in the study that ordered before they finished reading their messages.

That is why I believe one well-designed system beats five apps. Not because five apps are bad. Because one system sees the whole picture. It works on the same data, at the same moment, for every department. The order, production, the warehouse and the invoice tell the same story, not five different versions of it.

"One system" does not mean throwing everything away and building from scratch. We often start with the one process that hurts the most. We connect data where most information gets lost today. Only then do we add the next pieces. I wrote more about connecting processes into one whole in the article Business operating system: how to connect company processes?

Before you switch off the TV

One question for you tonight. Count how many apps your team has to open to answer a customer's simple question: "What stage is my order at?".

If the answer is more than one, let's talk. Book a free consultation. No presentation full of promises, just a conversation about where information gets lost in your company today and which process is worth starting with.

And now I am putting the remote down and going back to my popcorn. Have a great evening! :)

How do you rate this article?

Your rating:

SCHEDULE A FREE CONSULTATION