There is a particular kind of thing that sits on people's to-do lists for months. Not because it's hard, exactly, but because it's big enough that starting feels daunting. The home renovation you want to get done before winter. The freelance business you've been thinking about. The career change you've been circling for two years. The team process overhaul your manager asked for in March. The family move you've agreed on in principle but haven't started mapping out.

These aren't tasks — they're projects. They have multiple phases, dependencies, decisions that need to be made before other decisions, and enough moving parts that the whole thing feels like it requires a coherent plan before you can do anything. And so nothing happens.

AI is genuinely useful here, and not in the way most people expect. The value isn't that ChatGPT or Claude knows what your project requires better than you do. It doesn't. The value is that talking through a project with an AI forces you to articulate what you actually mean — and that articulation is often where the planning actually happens.

Why AI Works as a Planning Partner

When you plan in your head, you can skip steps. You can hold vague intentions without testing whether they're real plans. "I'll figure out the budget" can sit in your mental model as a resolved item without you ever figuring out the budget. Internal planning is fast, but it's full of gaps that feel like solid ground.

When you write something out — even just for yourself — those gaps become visible. But writing a plan from scratch is itself a blank-page problem. You need a structure to fill in.

What AI does well is give you that structure quickly, based on a description of what you're trying to do. It turns "I want to launch a freelance consulting practice" into a list of phases, questions, and early decisions in about thirty seconds. You will disagree with parts of it. Some parts won't apply. But you're no longer staring at a blank page — you're reacting to something, and reacting is much easier than originating.

This is the core dynamic worth understanding: AI as a starting-point generator, not an oracle. You bring the context and judgment. It brings structure and a set of prompts to react to.

A Simple Starting Approach

The most effective way I've found to use AI for project planning is a two-step conversation. The first step is description — you describe the project as plainly as you can. The second is refinement — you push back on what the AI produces and make it more specific to your actual situation.

For the first step, a prompt like this works well in either ChatGPT or Claude:

"I'm trying to plan [describe the project]. I want to end up with [describe the goal]. The constraints I'm working with are [describe the key constraints — time, budget, resources, dependencies]. Can you give me a breakdown of the main phases or workstreams I should be thinking about, and the key decisions I need to make early on?"

Be specific about constraints — that's what separates a useful plan from a generic one. "I want to launch a consulting practice" produces a generic list. "I want to launch a consulting practice while keeping my full-time job, with a target of two clients within six months, working about ten hours per week" produces something much more usable.

The second step is pushing back. Look at what the AI produced and ask:

Tell the AI those things and ask it to revise. "You've included X, but that's not relevant because Y. You've missed Z, which matters because W. Can you redo the plan with those changes?" After one or two rounds of this, you typically have something you can actually work from.

Using AI to Surface What You Haven't Thought Of

One of the more useful things AI can do in project planning is identify the questions you haven't asked yet. Not because it has special insight into your situation, but because it has seen many descriptions of similar projects and can pattern-match to common failure points.

A useful prompt once you have an initial plan:

"Looking at this plan, what are the most common ways that projects like this go wrong? What should I think about or decide early that I might be inclined to defer? What assumptions am I probably making that I should test before going further?"

This tends to produce genuinely useful material — not always, and not every point will be relevant, but regularly enough that it's worth the two minutes. In my own work, I've used this kind of prompt when planning a significant piece of writing and had the AI flag a sequencing dependency I hadn't thought through. The plan needed to change before I started. That saved significant rework later.

One honest caveat: AI doesn't know your specific context. It doesn't know your team dynamics, your organization's politics, your own energy levels and patterns, or the specific history of why previous attempts at similar projects stalled. The planning questions it raises are useful starting points, not authoritative checklists. You still need to bring the judgment about what actually applies to your situation.

Breaking a Big Project Into a First Week

One of the most common reasons projects stay stuck is that the plan is all phases and no immediate actions. You have a twelve-week timeline but no idea what to do on Monday morning.

AI is good at bridging this gap. Once you have a rough plan, you can ask:

"Given this plan and these constraints, what are the three to five most important things I should do in the first week? I want specific actions, not categories — things I could actually put on a calendar or a to-do list."

The specificity request matters. Without it, you tend to get things like "research vendors" or "define success metrics" — which are categories, not actions. Push for actual steps: "List three vendors that fit the profile I described and note one differentiator for each" or "Write down what success looks like in concrete terms — one sentence per stakeholder." Those you can do. Categories you can put off indefinitely.

What This Works For (and What It Doesn't)

The planning conversation approach works well for projects with a reasonably defined endpoint and some structure — work projects, personal goals, life logistics, creative undertakings. It works less well for things that are genuinely open-ended explorations, where you don't yet know enough to describe what you're trying to do. For those, a different kind of AI conversation tends to be more useful — more like talking through options than building a plan.

It also works better when you go into the conversation with some specificity about constraints. The more vague the brief, the more generic the output. "I want to improve my health" generates a different (and much less useful) plan than "I want to build a consistent exercise habit over the next ninety days, working with a bad knee, without joining a gym, with about thirty minutes available in the mornings."

Claude and ChatGPT are both good for this kind of work — they can hold a planning conversation through several rounds of back-and-forth without losing track of the context. If you're on a paid plan, you'll generally get better follow-through on complex planning conversations than on the free tiers, though the free versions work for simpler projects.

The Planning Happens in the Conversation

The thing that tends to surprise people when they try this for the first time is that the most valuable part isn't the document the AI produces at the end. It's what you figured out during the conversation — the constraints you articulated, the assumptions you tested, the sequencing you worked through, the early decisions you surfaced.

AI as a planning partner works because thinking out loud to something that responds is different from thinking in your head. The responses don't have to be brilliant to be useful — they just have to be responsive enough that you keep clarifying and refining what you actually mean. That process is where the planning happens. The output is just the record of it.

The project that's been on your list for three months probably doesn't need more time to think about it. It needs thirty minutes with an AI and a willingness to write down what you actually mean when you say you want to do it.