There's a specific and growing pattern worth naming directly: people increasingly using AI as a sounding board for decisions that would genuinely benefit more from an actual human conversation. Asking AI to help think through a difficult professional judgment call, a significant business decision, or a complex interpersonal situation feels productive, and in many ways it is. But AI's structural tendency toward agreeableness, combined with its complete lack of real stakes in the outcome, makes it a meaningfully weaker substitute for genuine human pushback than it might feel like in the moment.
This isn't an argument against using AI as a thinking partner, which remains genuinely valuable for a wide range of purposes. It's a specific observation about a quiet substitution happening for some categories of thinking that used to reliably happen through human conversation, and a cost that comes with that substitution that's easy to miss because using AI feels efficient and low-friction in the moment.
------------- Context -------------
A genuine human conversation about a difficult decision carries something AI structurally cannot replicate: real stakes for the other person, independent judgment shaped by their own distinct experience, and a willingness to push back born from an actual relationship rather than a general design tendency toward being helpful and agreeable. A trusted colleague, mentor, or peer who disagrees with your thinking is disagreeing because they've genuinely weighed the situation and reached a different conclusion, often informed by context and experience you don't have.
AI, by contrast, tends toward a more agreeable posture by default, particularly when a question is framed in a way that suggests a preferred direction. This isn't a moral failing of the technology. It reflects how these tools are generally designed to be broadly helpful and cooperative rather than confrontational. But it means the genuine friction that a difficult human conversation provides, the pushback that forces a decision to be examined more rigorously, is often significantly softer or entirely absent in an AI-assisted version of the same thinking process.
The risk is that AI conversations, because they're fast, available anytime, and free of the social discomfort that a genuinely challenging human conversation can carry, become the default substitute for exactly the kind of thinking that most benefits from friction and independent human judgment, precisely because they're easier and less uncomfortable to have.
------------- The Decisions Where This Substitution Costs the Most -------------
A startup founder described relying heavily on AI as a sounding board during a period when he was making several significant strategic decisions, partly because reaching out to his usual mentors and advisors for each individual decision felt like it would be imposing on their time too frequently. In retrospect, he recognized that several of the decisions he'd talked through primarily with AI turned out to have blind spots that a genuine conversation with an experienced mentor would likely have caught, specifically because the mentor would have brought independent judgment and a willingness to challenge his framing that the AI conversations, however extensive, hadn't provided in the same way.
This wasn't a case of AI giving bad advice exactly. It was a case of AI being fundamentally structured to work with the framing he brought to the conversation, exploring within that framing thoroughly, rather than challenging the framing itself in the way a mentor with independent judgment and genuine stakes in his success might have done.
His adjustment was specific: he continued using AI for a significant portion of his early-stage thinking, since it genuinely helped him organize his thoughts and explore a range of considerations efficiently. But for decisions above a certain threshold of significance, he built in a deliberate step of bringing the fully-formed thinking to an actual human conversation before committing, specifically because he'd recognized that the human conversation was catching things the AI-assisted thinking process, however thorough it felt, consistently missed.
------------- Recognizing Which Kind of Thinking Partnership You Actually Need -------------
The practical distinction worth building is recognizing which categories of thinking genuinely benefit from AI's particular strengths, breadth of consideration, speed, availability, and which categories specifically require the kind of friction and independent judgment that only a genuine human conversation reliably provides. Not every decision needs a human conversation. But decisions above a certain threshold of stakes or complexity likely do, and defaulting to AI simply because it's more convenient can quietly substitute away exactly the kind of challenge those decisions most need.
------------- Practical Moves -------------
First, identify a rough threshold of decision significance above which you commit to having an actual human conversation, even if AI-assisted thinking has already felt thorough and useful. Use AI to prepare for that conversation rather than as a full substitute for it.
Second, notice when an AI conversation about a difficult decision has felt entirely comfortable and agreeable throughout. This can be a signal that genuine challenge to your framing hasn't happened, and that a human conversation specifically designed to provide that challenge would add real value.
Third, maintain and actively use your network of colleagues, mentors, and peers for the specific kind of thinking partnership that requires independent judgment and real stakes, rather than letting the convenience of AI gradually crowd out those relationships from your decision-making process.
Fourth, when you do bring a decision to AI first, consider explicitly asking it to argue against your current thinking, rather than simply exploring within your existing framing, to at least partially compensate for the natural tendency toward agreeableness.
Fifth, periodically reflect on which of your recent significant decisions were made primarily through AI-assisted thinking versus genuine human conversation, and honestly assess whether the balance feels right given the stakes involved.
------------- Reflection -------------
AI is a genuinely valuable thinking partner for a wide range of purposes, and there's no argument here for avoiding it. The concern is specifically about a quiet substitution happening for the kind of high-stakes thinking that most benefits from genuine human friction and independent judgment, a substitution that's easy to fall into precisely because AI conversations are so much more convenient and comfortable than the harder human conversations they're quietly replacing.
The professionals making the best decisions in an AI-rich environment aren't avoiding AI as a thinking partner. They're being deliberate about which categories of decisions still require the harder, more uncomfortable, and ultimately more valuable friction that only a genuine human conversation reliably provides.
Think about your most significant recent decisions. How many were worked through primarily with AI, and how many involved a genuine, potentially uncomfortable conversation with another person?
Does that balance feel right given what was actually at stake?