The Hardest Part Is Starting (And Why AI Actually Helps With That)

You know the conversation needs to happen. One of your team members is struggling, performance slipping, deadlines missed, quality issues starting to show up, and you've known for weeks. But every time you sit down to figure out how to approach it, you find something else that needs doing first: an email, a report, a site walk. Anything but that conversation.

That's not laziness. Your brain is protecting you from discomfort. The conversation involves conflict, uncertainty about how it lands, and the possibility of damaging a relationship you care about, which is a lot of emotional load before you've even opened your mouth. So the brain offers you literally anything else to do instead.

Psychologists call the underlying skill emotional regulation, your ability to manage your own state well enough to do difficult work. When it holds, you walk into the hard conversation. When it fails, you get procrastination, avoidance, and work that doesn't happen until it becomes urgent. And it fails for everyone. Attention and follow-through fluctuate across a day, across a week, across a career. Sleep deprivation makes it worse, stress compounds it, and the more emotionally loaded the task, the harder it is to start. After a rough night, even the most disciplined person on your team is capable of spending a whole morning avoiding one email.

What another person changes

When you're stuck and you talk the problem through with a colleague, you usually come unstuck, and it's worth noticing that this isn't because they had better answers. Most of the time you already knew what needed doing. Explaining the problem out loud to another person settles your state enough that you can actually engage with it. Psychologists call this co-regulation: two people stabilizing each other so both can work.

My suspicion is that this is a large part of why in-person work matters more than people admit. Some of what happens in an office is idea exchange, but a lot of it is nervous systems keeping each other in the zone where hard work stays uncomfortable-but-doable instead of tipping into shutdown.

A model can give you a version of this. Not as well as a trusted colleague, but well enough to get you moving, and moving is most of the battle.

Why tutoring works, and what that has to do with models

There's research from the 1980s by Benjamin Bloom showing that one-on-one tutoring produces about two standard deviations better learning outcomes than classroom instruction. The interesting part is why. A tutor doesn't know more than the textbook. What a tutor does is keep you in what's called the zone of proximal development, the range where the work is hard enough to be engaging and not so hard that you quit, and they hold you there by constantly adjusting: a smaller problem here, a guiding question there, just enough structure that you stay in motion.

A model can simulate a useful part of that. When you're avoiding something, you can prompt it to help you break inertia:

"Take on the persona of a strategic advisor and an experienced executive coach. I need to have a difficult conversation with a team member whose performance is slipping. I've been avoiding it. Help me figure out how to start. Not the full plan, just the first small step that gets me moving."

What comes back is probably not exactly what you'll do. It's a springboard, something to react to and push against, and reacting is a much lighter lift than composing from nothing.

A protocol that works

1. Name what you're avoiding and why, in one sentence. "I need to address this performance issue, but I'm worried about damaging the relationship and I don't know how to structure the conversation."
2. Give the model a role suited to the work: an executive coach for the hard conversation, a systems-minded project manager for the gnarly plan, a pragmatic senior engineer for the technical decision you keep deferring.
3. Ask for the smallest possible start. Not a plan, not a solution. "What are three ways I could start this in the next ten minutes?"
4. Pick one and do it immediately, before you can think better of it. A draft email. An outline. Three bullet points.
5. Use the momentum. That first action changes your state, and what felt impossible fifteen minutes ago is now just work.

I'll give a concrete version. You tell the model: "I need to talk with a valued team member about declining performance. I care about them, and I need change within 30 days. I've been putting this off. Help me figure out the first ten minutes of the conversation." It reflects the tension back, offers a few openings, drafts a short agenda you could send ahead of the meeting. You fix the parts that don't sound like you, send the agenda, and put the meeting on the calendar. The conversation you'd been carrying for weeks took ten minutes to start.

The same principle, built into tools

This is also why we think so much about friction in the tools we build at xSkel. A project manager with 300 active work orders faces a wall every morning. Where do you even begin checking for problems? The honest emotional response is avoidance, and it has nothing to do with how good the PM is.

So our monitoring tool asks the boring questions automatically every morning and hands back only what needs attention. The job changes from "check 300 things" to "review these eight," and suddenly it's doable. That's co-regulation at the process level. The tool isn't doing the PM's work; it's lowering the cost of starting it.

Guardrails

A model is not a therapist. If the load is clinical, persistent anxiety, depression, acute distress, that's a job for a professional and not for AI. You also stay accountable: the model helps you start, and you own the outcome. And for decisions with real consequences, insist on sources and uncertainty flags, which is exactly why our technical tools cite datasheets and abstain when they can't be sure.

Try it this week

Next time you catch yourself avoiding something that matters, write one sentence: "I'm avoiding ___ because ___." Ask a model for three ten-minute ways to start, pick one, and do it right then. Then notice what happened to the dread.