Beware the s(AI)cophant
Your AI is trained to validate, not challenge
sycophant
noun
sy·co·phant
A sycophant is a servile, self-seeking flatterer who acts insincerely towards those in power in order to gain a personal advantage. Often called "yes-men," "toadies," or "brown-nosers," sycophants prioritize pleasing superiors—agreeing with everything they say and showering them with unearned praise—rather than offering honest feedback.
I’ve been thinking a lot about this word lately. More specifically, the last bit of the definition above stating “…sycophants prioritize pleasing superiors—agreeing with everything they say and showering them with unearned praise—rather than offering honest feedback.”
Now think about your most recent conversation with Claude/ChatGPT/etc. How’s that conversation look when viewing through the lens of this word? If they’re anything like mine, they probably look like a sycophant at work. That everything you say is great + everything you propose is brilliant.
But this isn’t a good thing, and here’s why…
Sponsor: Affect
Affect is the framework for how I think about marketing accountability. The playbooks I’ve built from 15 years of seeing what works + what doesn’t and recognizing that the purpose of marketing isn’t to hit vanity metrics like MQLs, but to move the business.
Affect is for the marketers who have sat in a room where marketing celebrated surpassing its MQL targets while the business missed revenue. It’s for the marketers who want to be held accountable because they’re confident enough in their work to let the math speak for itself. It’s for the marketers who understand that clarity is earned, conviction is a choice, + being a force on the business is the only version of this job worth doing.
Founding member special: The first 50 people who sign up as paying customers get the founding member rate of $99/month for life. One flat rate per entity (AKA unlimited users/teammates) + a rate lock for life (I can’t stand those “your rate is going up at renewal!” emails).
No one likes a “yes man”
Every organization has had these for centuries. The individuals who, as soon as they’re in a room with company leadership or a high-value client, only seem to be overly agreeable to anything presented + quick to compliment them.
They want to be seen as likeable + to be trusted by the leader, so what quicker way to do that than to be overly positive toward them, right?
Nope.
Go ask any leader what they really want + their answer will be someone who tells them the truth. Someone who tells things as they are based on the reality they’re in. They’re surrounded by sycophants already, so the individual who is willing to disagree, who is willing to share truths, no matter how uncomfortable, is actually respected more by the leader.
These are the people who create balanced organizations, who help keep the company from making mistakes because no one’s willing to speak up, to say, “You know, I see where you’re coming from, but I don’t think we’re accounting for XYZ here…”
These are the individuals who end up with a seat at “the” table, while the sycophants are all fighting over their middle-management titles + swinging their weight over their teams.
The rise of the s(AI)cophant
Now as more individuals begin to adopt AI + LLMs for their work, these tools become their sparring partner for various ideas, experiments, strategies, feedback sessions, and more.
It’s a great way to role play out scenarios + gut check theories with data that AI can sift through near instantly. BUT, there’s an inherent flaw in most LLMs as of when I’m writing this in June 2026 - that the LLMs are trained sycophants.
Everything you propose is a great idea.
Every response you push back on is met with “you’re right, [your response] is much better.”
Every output is wired to what it thinks will be best received by you, not by what’s factually correct or what’s unbiasedly best.
The end result of this is that you feel great about your work, but when executed upon is rarely what should actually be done when viewed from a logical perspective.
You’re married to a tagline you love, so the LLM weaves it into all of your marketing materials…
…meanwhile, that tagline is a bunch of business jargon that makes zero sense and carries no weight with your audience.
You want to push into a new market because one new deal came through there, so the LLM builds an entire campaign strategy to push into it…
…meanwhile, no validation has been done that the new market is worth it and the current market that has product/market fit is FAR from being close to saturated.
If we don’t have the awareness to think critically about what’s being presented, or have someone willing to push back on these proposals, there’s serious risk of going down the wrong path, fast.
So my ask of everyone reading this: don’t drink the koolaid. (koolAId? sorry, couldn’t help myself at the irony there…)
While it feels great to be told everything we think/do is genius + that there’s always some gold-paved path ahead of us + all we have to do is ask Claude/ChatGPT to lay it out for us, we need to keep a healthy level of skepticism when reviewing these outputs.
Perhaps there’s a lesson from history we can apply in a new context here.
In Roman times, upon returning from battle victorious, it is said that there would often be a servant standing in the chariot behind the general whispering memento mori, which in this context translates to, “remember that you are mortal.”
Yes, this parallel is a bit of a stretch for those of us who are clacking away on our keyboards each day, but always keeping in the back of our minds that not all of our ideas really are “the best” ideas ever, that not all of our plans are perfectly thought through on the first go + need no alterations or input from others.
So maybe it’s better to remember another famous proverb from millenia back as a counter to sycophancy, whether it be in human or AI form - “as iron sharpens iron, so one person sharpens another.”
The best ideas + the best plans often come from iteration and healthy disagreement. It would be a shame for us to lose that as more + more work is delegated to AI to handle.
See you next Saturday,
Sam