No founder Iāve spoken to had a clean starting point with AI. When GPT arrived, entrepreneurs didnāt get a book to read or a set of rules to follow. They had a tool that suddenly existed, and a lot of uncertainty about what it was actually useful for. Most people dealt with that by experimenting in their own spare time, while trying to get through work that still needed doing.
Thatās been consistent across the 55+ conversations Iāve had for Millennial Masters over the past year. Founders tried AI on real tasks before they trusted it. They learned by using it, getting annoyed with it, and coming back to it when something else felt slow or frustrating.
Three years into the GPT era, AI still isnāt smooth. It can be stubborn, inconsistent, and surprisingly hard to control. Iāve spent hours fighting with this very draft while it ignored instructions I was being explicit about. Even so, AI has already worked its way into businesses through the kind of repetitive work that quietly eats time.
Hereās where founders have already started using AI to take pressure out of their work. šš»
The work that disappeared first
AI tends to stick first in work people keep putting off.
Carly Meyers describes the decision happening while the task is already open. āOne of the most powerful questionsā¦ is when they say, āI wonder if AI can do this?āā¦ This thing that Iām doing right now is boring. Itās monotonousā¦ I wonder if AI could help me with this.ā
Ryan Carruthers applies the same thinking to sales conversations that would otherwise sit untouched. āI reactivate dead leads via AIā¦ It doesnāt matter if it all goes to pot because theyāre dead anywayā¦ then I can start to get rid of some of the repetitive tasks.ā
James Augustin uses AI to break down sales call transcripts and flag what worked and what fell flat, without anyone having to trawl through the recordings. āI think the first thing is humans should be doing things that are innately humanā¦ they should be going out to sales meetingsā¦ recording podcastsā¦ talking about strategy.ā
With AI, people stop doing the boring work once they see it doesnāt need them anymore.
Building gets cheap, distribution doesnāt
As the cost of building drops, ideas move sooner. Testing something no longer eats up weeks of time or a large chunk of budget.
Simon Jenner sees this clearly in early product development. āWeāve got vibe coding, which allows you to use the AI prompt to do itā¦ It gives you the advantages of code, which is scalable.ā When building is cheap, it becomes easier to find out quickly whether something deserves more attention.
Gus van Rijckevorsel described how this plays out in game development. āSomething that we could have done in four months, we can do it in two weeks.ā That removes long (and expensive) stretches where nothing can happen. Teams can test features quickly and spot duds before weeks get burned on them.
As development gets cheaper and faster, the work moves elsewhere. Reaching the right people starts to matter more than perfecting the thing itself. Distribution, audience fit, and getting in front of a specific group become the harder problems once shipping stops being the bottleneck.
Whoās still needed
You donāt need as many people once AI is doing a meaningful share of the work.
Roles that existed to deal with volume, admin, or routine output start to disappear when the first pass is handled automatically.
Nick Holzherr has been blunt about what heās seeing. āTheyāre not hiring junior marketing people anymore. They are hiring more senior people that can use AI. So they just need less people.ā
Nick Telson-Sillett described the same pattern in customer support. AI deals with most incoming queries before anyone gets involved, which means people only step in when a conversation needs judgment, reassurance, or experience. āAI handles about 70% of our support tickets,ā he said.
The practical result is that teams stay smaller, and the people in them sit much closer to decisions that actually matter. AI absorbs the volume and repetition, while humans focus on the moments where context, trust, or money are on the line.
What doesnāt get delegated
As AI produces more output, deciding what to act on starts taking longer.
This shows up clearly in recruitment. Noel Andrews described how āresumes and CVs are now 100% useless because of AIā¦ Itās like AI scoring AI. Itās just rubbish.ā When applications, cover letters, and even interview answers are generated with the same tools, polish stops being useful. What matters instead is how someone thinks, reacts, and reasons when they donāt have time to outsource the answer.
The same pressure appears when deciding where automation belongs inside a business. Oliver Yonchev talked about how cheap access to thinking tools has changed expectations. āThe cost of deep thinking work and planning and strategy and business consulting is a Ā£20 a month subscription on ChatGPT.ā
Yota Trom described using AI as a way to refine ideas without handing responsibility over. āMy philosophy around that is that AI is my editor, not my creator.ā In practice, this keeps judgment with the person doing the work, even when the tool can suggest endless variations.
AI produces material quickly, but someone still has to decide what matters, what gets ignored, and what carries risk if it goes wrong. The work shifts toward judgment, because that part canāt be automated without consequences.
Owning the outcome
As AI takes on more execution, the value in a business shifts away from doing the work and towards owning the outcome.
This comes up most clearly when people talk about software. Thibault Louis-Lucas described a future where building the product stops being the hard part. āImagine ifā¦ you do not need developers at all to create products. So where is the value then if itās not in the product itself?ā When anyone can prompt a working tool into existence, attention moves to distribution, reach, and trust.
Peter Watson was blunt about what happens to agency businesses built entirely on execution. āI donāt think weāre far away from three button clicks, no advertising agency required.ā When the doing gets automated, businesses that sell time or output struggle to justify their place. What survives is direction, taste, and knowing what to do with the tools once they exist.
Nick Telson-Sillett pointed out that high-value sales still hinge on human trust. āIf you are selling anything over $50,000 and above, the person taking the bet on buying a software will want to speak to someone.ā AI can handle volume and qualification, but the moment risk enters the conversation, people still want a person on the other end.
Jordan Stachini raised concerns about how skills develop when junior roles disappear. Creative work has traditionally relied on people learning by doing, often badly, before getting good. When AI replaces that early stage, the path to senior judgment becomes harder to see.
Some people are already building around this shift. Carly Meyers described running her business by assigning AI agents distinct roles, treating them like an internal advisory layer. She consults them, delegates work to them, and uses their output to inform decisions. In her view, companies built by one or two people will become normal once execution stops being the bottleneck.
As execution keeps getting cheaper, what carries weight is deciding what to build, who itās for, and how it reaches them.
2026: Still figuring it out
There still isnāt a rulebook for using AI in a business, and after the last year of conversations itās hard to imagine one that would actually help.
What people are really dealing with is how their role shifts once a lot of the busywork is gone. The time doesnāt magically free up. It gets pulled into decisions they canāt dodge anymore, or choices they used to push down the line.
AI will still dominate the conversation in 2026, mostly because most people are still trying to make it useful without letting it take over their day.
Iāll keep digging into that through Millennial Masters by staying close to how founders and operators actually live with these tools once the novelty wears off and the trade-offs become harder to ignore.
More AI insights from Millennial Masters:
