Project Management · Intermediate
Draft a work breakdown structure from a charter
The charter is approved. It fits on one page: an objective, five deliverables, a budget and a go-live date. The sponsor wants a plan by Friday. Between that page and a schedule anyone can trust is a work breakdown structure, and the blank first draft is the hardest part to start.
What AI is for here, and what it isn't
A work breakdown structure (WBS) breaks the project's total scope into smaller deliverables until each piece is small enough to estimate, assign and track. It's the foundation of the schedule, the budget and the responsibility assignments, which is why mistakes in it spread everywhere.
AI is good at the first draft: proposing a sensible hierarchy for a familiar kind of project, listing the deliverables people tend to forget, and noticing dependencies. It removes the blank page.
It doesn't know your organization's way of working, your team's skills, the vendor's real lead times, or the work that's already been done. And it should never produce the estimates. Estimates come from the people who'll do the work; their involvement is also what makes them commit to the plan. A plan the AI wrote and the team never shaped is a plan nobody owns.
Start from deliverables, not activities
The most common WBS mistake is listing activities ("hold workshops," "do testing") instead of the deliverables they produce ("approved requirements," "tested release"). A deliverable-based WBS has two properties worth protecting:
- The 100% rule. The WBS contains all of the project's scope and nothing outside it. Each level's children add up to exactly the whole of their parent. The rule is set out in PMI's Practice Standard for Work Breakdown Structures (second edition), and it's what makes the WBS a check on scope as well as a planning tool. It also implies a second rule: no two elements should overlap, or the same work gets estimated twice.
- Nouns, not verbs. Each element is a thing that will exist. Activities live one level down, in the schedule.
Take the deliverables from your charter and write them as the top level. Add the ones every project has but charters forget: project management itself, and transition (training, handover, support).
Generate the first-pass breakdown
You are helping a project manager draft a first-pass, deliverable-based
work breakdown structure. The team will review and change it.
CHARTER / SCOPE: [paste objective, deliverables, in/out of scope,
constraints, key dates]
PROJECT TYPE: [e.g., CRM migration, office relocation]
TEAM SHAPE: [roles available, internal and vendor]
SIMILAR PAST PLANS: [paste or summarize, if any]
PRODUCE:
1. A WBS to three levels, numbered (1, 1.1, 1.1.1), using noun-based
deliverable names. Include project management and transition.
2. For each level-3 element, a one-line description of what "done"
looks like.
3. A list of deliverables commonly forgotten on this type of project,
marked as "suggested, confirm".
4. Dependencies between elements (A must finish before B starts), and
external dependencies (vendors, approvals, other projects).
5. QUESTIONS FOR ESTIMATORS: what each work-package owner will need to
know before they can estimate.
RULES: Respect the 100% rule. Exclude anything listed out of scope.
Do not estimate effort, duration or cost. Do not assign people.
The questions for estimators turn the draft into a working session agenda: instead of asking someone "how long will the data migration take?", you can ask "how long will it take to migrate 40,000 accounts with these three de-duplication rules, given test data by 15 January?"
Check the draft against the charter
Before showing anyone, check the draft yourself:
- Every charter deliverable appears somewhere in the WBS.
- Nothing out of scope has crept in. Models tend to add plausible extras ("mobile app," "API documentation"). Remove them or raise them as scope questions.
- Each level-3 element is small enough for one person or team to own, and to estimate in days or weeks rather than months.
- The "commonly forgotten" list: decide which items belong. Data migration, environments, security review, training, communications and post-go-live support are the usual omissions, and each one is a schedule surprise waiting to happen.
Build it out with the people doing the work
Run a working session with the work-package owners. Show the draft as a draft: it's easier for people to correct a structure than to invent one, and the session will move fast.
For each work package, the owner:
- Confirms or changes the breakdown below their element.
- Answers, or takes away, the estimator questions.
- Gives an estimate as a range (best case, likely, worst case), with the assumptions behind it.
- Names the dependencies they need from others.
Write the assumptions down with the estimates. "Two weeks, assuming the vendor delivers the test environment by 10 January" is a plan. "Two weeks" is a hope.
Worked example: the office relocation
Fictional charter. Move a 90-person office to a new floor across town by 30 April. Deliverables: fitted-out space, working IT and telephony, moved staff and equipment, decommissioned old space. Out of scope: furniture procurement (facilities' separate contract), changes to network design.
The AI's first pass (level 2 shown) proposes:
- Project management
- New space readiness (fit-out acceptance, access control, signage)
- IT and telephony (network cabling, Wi-Fi, printers, meeting room AV, phones)
- Move execution (move plan, packing, crate logistics, desk allocation)
- Old space exit (equipment removal, data destruction, handover to landlord)
- Transition (staff communications, first-day support, snag list)
Commonly forgotten (confirm): ISP circuit lead time at the new site; security badge enrollment; mail redirection; certified data destruction for old devices; insurance for equipment in transit.
Two things happen in review. The PM removes "New furniture selection," which the model had added under 2: it's out of scope. And the facilities lead points out the item that changes the whole schedule: the ISP circuit lead time. The AI listed it as "confirm," and a call to the provider confirms 60 business days. The circuit order goes in that week, and the WBS gains a dependency that drives everything in IT and telephony.
The AI didn't know the lead time. It knew to ask about it. That's the right division of labor.
Baseline it
When the owners have estimated their packages and the dependencies make sense, the WBS becomes the basis for the schedule and budget. Get the sponsor's approval of the baseline, and from then on treat changes to it as changes to scope.
WBS ID: [1.2.3]
Name: [deliverable, as a noun]
Parent: [1.2 name]
Description: [what's included]
Done when: [acceptance criteria]
Owner: [one name]
Estimate: best [n] / likely [n] / worst [n] [days / weeks]
Assumptions: [what the estimate depends on]
Depends on: [WBS IDs, external dependencies]
Needed by: [WBS IDs that depend on this]
Source: [AI draft / team session, date]
What stays with you: the estimates, the sequencing and the resourcing, worked out with the people who'll do the work.
Further reading: Project Management Institute, Practice Standard for Work Breakdown Structures, second edition (2006), especially section 2.4 on the 100% rule, and PMI's overview of the work breakdown structure.