The thing AI startup bros don't understand about enterprise-level complexities
Or, why even if you can spin up an app with four engineers and a bit of cloud magic, you will have problems delivering AI at scale inside a Fortune 500 company
Hello to the 3,002 subscribers who read Consulting Intel!
I came across this tweet the other day by some random AI startup bro mocking Accenture and its so-called “AI experts”.
Now, let me be clear: I have zero interest in defending Accenture. I do not work for them (though - like a lot of you 😅 - I did at one point in my career), and I am sure we could spend all day discussing the flaws of an organization that size.
But what I do want to address is the lazy, arrogant and frankly idiotic take behind that tweet.
I have never heard before of this person, but let us first define in general terms who these people are, those that I call the startup bros. They usually either come from money or, at some point, managed to build a mildly successful product, maybe they got a few thousand users, and maybe some mid-sized tech company acquired them for a few million.
Suddenly, they became “rich” and think they are divine gifts to the world of technology.
Obviously, what they fail to realize is that, more often than not, they simply “got lucky”: right time, right place, right VC backing.
Fooled by Randomness when. somehow, this stroke of fortune mutates into an unshakable belief that they have cracked the code of enterprise technology, and that their ability to build a glorified proof of concept makes them uniquely suited to run complex AI transformations.
It would be cute if it wasn’t mad.
These guys think because they can move fast in a greenfield environment, spinning up an app with four engineers, a bit of cloud magic and a few Domino’s pizzas, they would have no problem delivering AI at scale inside a Fortune 500 company.
This is where their ignorance shines.
Enterprise technology work is not as much about building as it is about integrating.
Enterprise technology work is about making something function within a labyrinth of legacy systems, regulatory requirements, internal politics, operational chaos.
Enterprise technology work is about dealing with a technology landscape so tangled that even the company itself does not really understand it!
Enterprise technology work is about deploying AI in an organization that:
has thousands of applications (most of which should have been decommissioned a decade ago but somehow still exist, and are run by soon-to-be-retirees);
operates in some heavily regulated industry where one misstep results in billions in fines;
has entire departments dedicated to actively slowing things down to reduce operational risk;
is listed on the stock exchange, meaning every major tech decision comes under investor scrutiny.
It is the difference between fixing up your bullshit moped in your garage VS being handed a 40-year-old Boeing 747 mid-flight and being told to upgrade the engine without landing the plane.
Enterprise AI needs more than just “engineers”
Another thing AI startup bros get wrong is they assume the only “real” AI experts are PhD-level researchers pushing the boundaries of the field.
Do these people exist inside consulting firms? Absolutely. But implementing AI at scale is not an academic exercise: it is a business transformation problem.
I can guarantee that if you assemble a team solely of machine learning engineers to make enterprise AI happen, your project would fail miserably, and that’s because to succeed you would also need:
risk and compliance experts to ensure the model does not get the company sued;
project managers who know how to navigate internal politics and get things approved;
testers who can evaluate AI-generated output at scale, ensuring it actually works;
business analysts who translate ideas from people that have no clue what AI even means into technology capabilities;
a few more things…
The AI revolution inside large organizations will not be led by 25-year-old Python engineers wearing Atari t-shirts who can fine-tune an LLM, but by teams who understand how to implement AI swimming with the sharks in the Big Company Shit Ocean.
Could AI startup bros survive in enterprise AI? No.
I have worked with these guys: if you dropped them into an enterprise AI program, most would not last six months. Not because they are not smart, but because they would not have the patience to deal with the realities of large-scale transformation.
They would get frustrated by the slow-moving processes.
They would rage-quit after the third compliance review.
They would throw their hands up in the air the first time a procurement team took six months to approve a software license.
And that is fine.
There is nothing wrong with being a builder in a startup environment: it is a completely different game and can be a much more lucrative one (that is if you get lucky, of course).
What is wrong and truly laughable, is pretending that consulting firms do not provide real value when they are often the ones ideating, designing, implementing, and/or running the operations behind almost every major brand in the world, including the brands our friendly AI startup bros use every day!
Before mocking the so-called “AI experts” inside these firms, maybe understand the difference between hacking together an AI (glorified) demo and deploying AI inside a $200 billion company.
One is a cute project. The other is a war.
Ciao, until next time! 👋
✍ The Management Consultant
PS: If you like this newsletter, I have one huge favor to ask.
Share it around with friends, family and colleagues.
This is the most effective way to support me (…and to keep me motivated to continue writing 😁).
Thank you 👇👇👇
🎯 INTERESTING SH*T
A couple of things I found on the internet that you may like…
A copy of The Sovereign Child: I still don’t know if I am fully convinced about the thesis but, admittedly, I haven’t finished the book yet.
One of the points Ilya makes, roughly: “AI in a few years will gain capabilities and self-awareness that we are not able to understand right now”…
🚨 SPONSORSHIP
More than 3,000 consultants globally read Consulting Intel and rate it a 8.2/10 newsletter!
The readers include management consultants from McKinsey, BCG, Bain, Deloitte, EY, EY-Parthenon, KPMG, PwC, Accenture, Oliver Wyman, PA Consulting, and various boutique consultancies worldwide.
Many corporate employees and independent consultants are regular readers of Consulting Intel.
If you think your products or services could resonate with this audience, get in touch!
👀 JOIN THE DISCORD SERVER
If you like this newsletter, you will love our Discord Server.
In there, you will find a tight-knit community of 170+ management consultants from all over the world discussing real-life challenges, giving each other support and recommending the good stuff to keep our knowledge top notch.
It’s absolutely FREE, and I’d love to see you in there!
You've definitely won yourself a new subscriber this week and I've already shared this with 10+ people. I don’t see anyone else writing on this important subject with this level of fluency: bravo
You are right, folks “selling” AI to enterprises are mostly selling “snake oil”
That said, the real value / danger is new businesses that are ground up “AI powered” that take out existing behemoths especially in tech.
There is no company I am more bearish on that SalesForce.
The era of disaggregation of applications is upon us