Most projects start with a brief that is half a page of lowercase notes. Turning it into a plan with tasks, owners, and dependencies can take a product manager days. It does not have to.
1. Pull out the outcomes
Read the brief once and write down what must be true at the end. “Checkout is live before Nov 18.” “New pricing screens ship.” These become your epics.
2. Break each outcome into tasks
For each epic, list the work in the order it has to happen. Keep tasks small enough to estimate in days, not weeks.
3. Write the dependencies down
For every task, ask “what does this wait on?” Briefs hide these in phrases like “mobile ships after that”. Make each one an explicit link.
4. Check for loops
If A waits on B and B waits on A, the plan cannot be scheduled. Loops creep in on large plans; check for them before anything is committed.
5. Flag, do not guess
The most useful part of a plan review is the list of things nobody knows yet. “Not sure who does the receipts email??” should become a task with no owner, clearly marked, not a task quietly assigned to whoever was last in the room.
6. Let the schedule come from the graph
Once tasks, estimates, and dependencies are in place, dates follow. Do not hand-pick dates; compute them, so the critical path is visible from day one.
Where software can help
A language model is good at steps 1 to 3: reading messy text and proposing structure. Code is better at steps 4 and 6: checking loops and computing dates. A person should own step 5 and the final approval.
Common mistakes
- Planning tasks before outcomes. Without clear outcomes, tasks multiply and nobody can tell when the project is done.
- Hiding unknowns. Assigning an unclear task to someone just to fill the column creates a surprise later.
- Dating by hand. Hand-picked dates ignore dependencies. Compute them.
- Skipping review. A generated plan is a draft. The team that does the work should check it before anything is created.
A quick template
For each epic, write: the outcome, the tasks in order, an estimate per task, what each waits on, and an owner or an explicit “no owner yet”. That is enough to compute a schedule and find the critical path on day one.
That is how Keel's brief to plan works: paste the brief, review a plan with the gaps flagged, and create nothing until you approve. See brief to plan.
