I'm a reasonably experienced backend developer paired with a mobile developer for knowledge sharing. He's a genuinely good person, but during implementation discussions he often types my explanation or question into Claude immediately and then returns with "Claude says…" On calls, he sometimes pauses the conversation to ask Claude and reads the response aloud.
I use AI tools regularly, so the issue isn't that he uses them. The problem is that I'm trying to understand his reasoning and collaborate with him, not receive a second-hand answer from an AI assistant. It sometimes feels like he trusts the generated response more than the technical discussion happening between us.
How can I bring this up directly without making things unnecessarily awkward? Has anyone else dealt with a coworker who uses AI as a relay for nearly every conversation?
3 Answers
Be direct, but frame it around the kind of collaboration you want rather than criticizing his AI use. Something like: “I’m happy to use Claude as a second opinion, but in these conversations I’d really like to hear how you would approach it, even if you’re uncertain. If I wanted Claude’s answer, I could ask it myself.” That makes the expectation clear without attacking him personally.
There may be a confidence issue underneath this. He might not trust himself to answer quickly, so Claude gives him something that feels authoritative. Try asking, “Does what I’m saying not make sense, or are you looking for a second opinion?” That invites an honest explanation and gives him permission to say he doesn’t know.
It’s also reasonable to set a boundary during calls: “Let’s talk through our assumptions first, then we can use Claude to check anything we’re unsure about.” That preserves the human discussion instead of interrupting it every time.
AI is useful for checking assumptions, researching unfamiliar details, or preparing a response. It becomes a problem when someone delegates all judgment to it and simply forwards the output. Developers still need to understand the context, challenge the suggestion, and take responsibility for the decision. Otherwise they’re just acting as a human connector between the team and an AI subscription.
A good compromise is to have him explain the recommendation in his own words and identify its risks or assumptions. That keeps the tool involved while making sure he actually understands the result.
You can also ask, “Okay, but what do you think?” whenever he starts reading the answer. Repeating that calmly may be enough to break the habit.