learn-ai-project-management-with-phoebe / PM session 2 of 10
Learn AI + Project Management with Phoebe · PM track · Session 2 of 10

Mandate to charter: two paragraphs of email into a programme you can defend

Northwind week 1. Your entire brief is an email from the COO that says "get all our data into one place" and "ideally before Christmas". Tonight you turn that into a charter with ordered priorities, written non-goals, and an honest list of everything the sponsor still has to tell you. The trick is not making the model write more. It is making it refuse to invent.

🟡 PM track PMs · TPMs · delivery leads Charter draft 45 min
0-3 · Where we are 3-15 · What a mandate leaves out 15-40 · Build-along: charter + non-goals 40-45 · Q&A
Part 0

Where we are: Northwind, week 1

Last session you built the workspace: one Project for the Northwind data platform, a six-document context pack, and a house-rules file. Tonight the pack is nearly empty, because week 1 is honest about how programmes actually start. You have one thing in it - a two-paragraph email from Elena V., the COO and your sponsor. Nine source systems, 812,000 rows a night at steady state, a legacy warehouse someone wants switched off, and none of that is written down anywhere yet. It is in people's heads, and your job this week is to get it onto a page that Elena will put her name against.

Live - presented in session Self-study - read after class ★ Try it now prompt Official sources covered
★ What you walk out with today A Northwind charter draft with every invented fact stripped out and marked TBC, a GAPS list that turns into ten questions for your sponsor, five written non-goals that will stop scope arriving sideways in week 6, and the move that makes all of it work: asking the model for the questions instead of the answers.
Part 1 · covers PMBOK 7 stewardship + planning

What a mandate leaves out 6 min live

Here is the whole brief. Read it the way you would read it at 08:14 on a Monday, with the rest of your week already booked. It is not a bad email. Elena is describing an outcome she wants, which is exactly her job. Turning an outcome into a scoped, sequenced, owned programme is exactly yours.

📩 The mandate - Elena V. (COO), Monday 08:14Subject: data project, can we finally get moving Hi - following up on the board session last week. We need to get all our data into one place so we can finally see sales properly, stores and online together. Right now every report I get tells me something different and I have stopped trusting the numbers. Ideally before the Christmas peak. Budget is tight this year so let us be sensible about it. Talk to Sofia, she has views on the technical side. Thanks, Elena
THE BRIEF YOU GOT THE CHARTER FIELDS The mandate email "data in one place" "see sales properly" "before Christmas peak" "budget is tight" "talk to Sofia" What you must ask the other six fields, before anyone commits Context the retailer, the mess Objective + measure "see sales" is not one In scope which of the 9 sources Out of scope the field that saves you Priorities P1 to P4 ordered, not all urgent Constraints budget? people? date? Assumptions what we bet is true Decision rights who signs, who is told The two dark fields are given. The six pale ones are questions you must ask.
🔍 Click to zoom - the mandate gives you two charter fields and implies a third; the rest are questions
LiveFive things the email does not say4 min

Every one of these is a decision that will get made whether or not you make it deliberately. Left blank, they get made by whoever shouts on the day.

  • Success measure. "See sales properly" is a feeling, not a measure. Does success mean one agreed daily sales number across stores and online? Available by 07:00? Reconciling to finance within a stated tolerance? Until that sentence exists, nobody can tell you whether the programme worked.
  • Scope boundary. "All our data" is nine source systems. Are all nine in the first phase, or three that carry the sales question and six that follow? "All" is the most expensive word in the email.
  • Priority order. Speed, cost and completeness are all implied and they conflict. Christmas peak pulls one way, "budget is tight" pulls the other, "all our data" pulls a third. Something has to give and Elena has not said which.
  • Constraint reality. "Budget is tight" is not a number. And nothing in the email mentions that you are one data engineer short, which is the constraint that will actually decide the date.
  • Decision rights. "Talk to Sofia" could mean Sofia K. decides the architecture, or advises on it, or has to be consulted before you do. Those are three different programmes.
Real world

The word "all". A delivery lead took a near-identical mandate at a grocery chain and started scoping all fourteen source systems, because the email said "all". Six weeks of discovery later the sponsor saw the plan and said, honestly baffled, "I only ever cared about tills and online." Two of the fourteen systems answered the actual question. The other twelve were a phase-two conversation that nobody had ever had, because nobody had asked what "all" meant out loud.

Self-studyWhy sponsors write it like this2 min read

It is tempting to read the email as vague or careless. It is neither. Elena is doing the thing a sponsor is supposed to do: naming an outcome and the pain behind it, then handing the how to someone whose job is the how. PMBOK 7 calls this stewardship, and the split is deliberate - the sponsor owns the why and the money, the delivery lead owns the shape of the work.

  • She has told you the real pain. "Every report tells me something different and I have stopped trusting the numbers." That is worth more than a spec. Trust in one number is the actual success measure hiding in the email.
  • She has given you an ally and a lean. "Talk to Sofia" is a routing instruction, use it in week 1 rather than week 5. "Budget is tight" may not survive a business case, but it tells you which way she leans when speed and cost collide.

Your job is not to complain that the brief is thin. It is to convert it into something specific enough that Elena can look at it and say "no, not that" - which is the only reliable way to find out what she actually meant.

Part 2 · covers PMBOK 7 planning + the PMO prequel

The charter fields that actually earn their place 5 min live

A charter template can run to thirty fields and most of them are ceremony. Nine do the work. These nine are the ones that later stop an argument, and the reason they matter here is that they are also the exact structure you hand the model - a draft with no field list comes back as prose, and prose hides the gaps.

FieldWhat it settles later
ContextWhy now. Stops phase two re-litigating the whole premise
Objective + success measureThe number or condition that means "done and it worked"
In scopeWhich of the 9 Northwind sources, in which phase
Out of scope (non-goals)The request in week 6 that you can decline in one sentence
Priorities P1 to P4, orderedWhat gives way when two good things collide
ConstraintsMoney, people, hard dates, systems you cannot touch
AssumptionsWhat you are betting on. Each one is a risk in waiting
Decision rights + named sponsorWho signs, who is consulted, who is merely told. One human: Elena V., not "the business"
LivePriorities are only real when they are ordered3 min

Ask any sponsor to rate priorities and you get four P1s, because everything genuinely does matter to them. A priority list where everything is P1 contains no information. It cannot resolve a single trade-off, which is the only job it has.

Priorities are only real when they are ordered P1 - ONE TRUSTED SALES NUMBER Stores and online. This is what never gives way. P2 - CHRISTMAS PEAK DATE The only date that has to be hit. P3 - BREADTH: REMAINING SOURCES The rest of the nine systems, later phases. P4 - DECOMMISSION LEGACY WAREHOUSE Saves cost. Changes nothing for Elena. First to give way. When the vendor extract slips, P3 gives way and P1 does not - the order does the deciding.
🔍 Click to zoom - the order is what resolves the trade-off, not a debate on the day

For Northwind, the ordered version might read:

  • P1 - one trusted sales number across stores and online. This is the pain in the email.
  • P2 - the Christmas peak date for that number, and nothing else needs to hit it.
  • P3 - breadth, the remaining source systems, then P4 - decommissioning the legacy warehouse, which saves cost but changes nothing for Elena.

Now the list does work. When the vendor extract slips, P3 gives way and P1 does not. When someone asks for a self-service model, it is not on the list at all. And notice what the order reveals: if Elena reorders P2 above P1, she is telling you she would rather have a date than a trustworthy number, which is a conversation you very much want to have in week 1 rather than week 14.

Real world

The reorder that changed the plan. A PM walked a sponsor through four priorities she had drafted as speed, coverage, cost, retirement. The sponsor moved cost from third to first without hesitating, then explained that a board commitment on run-rate landed in January. Nothing in the brief had mentioned it. That single drag-and-drop killed two workstreams, saved four months, and only happened because the list was ordered enough to be argued with.

LiveNon-goals: the field that saves the programme3 min

Scope rarely arrives through the front door. Nobody sends a change request titled "please double the programme". It arrives sideways, in a corridor, as a reasonable-sounding sentence: "while you are in there, could you also...". The written non-goal is the only cheap way to answer that.

  • An unwritten non-goal is not a non-goal. If it lives in your head, the answer in the corridor is a negotiation. If it is in a charter Elena signed, the answer is a link.
  • Non-goals are not refusals. "Not in this phase, revisit at M4" is a non-goal. It respects the request and dates the conversation, which is why they rarely cause offence.
  • Write the ones people will actually ask for. A non-goal nobody would have requested is filler. The useful ones feel slightly uncomfortable to write down, which is how you know they were live.
The test for a real non-goal Read it out and ask: would anyone plausibly have asked for this in the next six months? If no, delete it. If yes, and you feel a flicker of "hmm, will Elena be annoyed" - keep it, that is the one that saves you.
Real world

The corridor request. Week 6 of a warehouse build, a regional director caught the delivery lead by the lifts and asked for live till data on his phone, "just for my stores, it should be easy now the pipes exist". It was not easy: nightly batch and real-time streaming are different architectures, different cost, different on-call. Because "real-time POS is not in this phase" was line two of a signed charter, the answer took nine seconds instead of three weeks of investigation the team would never have got back.

Part 3 · the highest-value move at this stage

Ask the model for the questions, not the answers 4 min live

Here is the counter-intuitive bit. At charter stage, the most valuable output is not a charter. It is the list of ten questions you must ask Elena before you commit to anything, generated by something that has read the email carefully and is not embarrassed to point out what is missing. Then you tell it explicitly not to answer them itself.

LiveWhy the questions beat the draft2 min

A model asked to write a charter from a thin email will produce a complete-looking charter, because completeness is what charters look like. It will quietly fill the gaps with the average of every data platform charter it has ever seen. That average is not wrong exactly - it is plausible, generic, and belongs to no one. Worse, it is hard to see: a gap filled with confident prose reads exactly like a fact.

Ask for the questions instead and the same reading ability works for you. It is genuinely good at noticing that "before Christmas peak" is not a date, that "all our data" has no boundary, and that nobody has been named as the person who signs off the model. That is rail one doing its job in reverse: a weak draft means a missing document, so the fastest route to a strong draft is finding out which documents are missing.

Real world

The question that moved a milestone. A TPM ran her mandate email through this move and got back a question she had not thought to ask: "who signs off that a data quality number is acceptable, and what is the threshold?" She took it to the sponsor, who did not know either, which forced a twenty-minute conversation with the finance lead. Out came the 99% figure that became the M3 gate. Without the question, "good enough quality" would have been discovered at the gate, in front of the steerco.

LiveTaking ten questions to a sponsor without looking unprepared2 min

There is a real fear here and it is worth naming: arriving at a COO's desk with ten questions can look like you have done nothing. Done well, it looks like the opposite. Four rules make the difference.

Bring the draft, lead with the draft. Open with "here is the charter as I understand it" and let her read a page that is 70% filled in. The questions are the visible holes in something substantial, not a blank form.

Propose an answer to every question. Never "what is the success measure?" Instead: "I have written the success measure as one agreed daily sales number, stores and online, available by 07:00. Is that what you meant?" It is far easier to correct a wrong answer than to produce a right one on the spot.

Say what each answer unlocks. "I need the scope boundary because it changes the plan by about two months." A question with a consequence attached is a decision request. A question without one is homework you are handing back.

Route the ones that are not hers. Of ten questions, maybe four are Elena's. The rest belong to Sofia K. on architecture, Priya N. on the vendor, Dan R. on reporting. Show her that you have already routed them - that is the move that reads as prepared.

Ten is the ceiling, not the target Take the four that change the plan. Park the rest in the GAPS list and answer them yourself during discovery. A sponsor who is asked four sharp questions answers them. A sponsor handed ten defers all of them to a follow-up that never gets booked.
Demo 1 of 2

Mandate email to charter draft, plus a GAPS list ★ 12 min · everyone builds

Open the Northwind workspace from b1, paste Elena's email in if it is not there already, and run this. The output you want is a charter that looks slightly unfinished, because it is. Every TBC in it is a question with a name attached.

Confirm the email is the only source in the workspace for this task, then run the prompt below. Read the charter first, then the GAPS list, then go back through the charter looking for anything asserted that the email never said.

Hunt for smuggled facts: a date, a budget figure, a team size, a technology choice, a phase-two plan. Each one is a place the model quietly averaged over your programme. Rewrite every one as TBC with the role who owns the answer.

Pick the four questions you take to Elena V. this week, and route the rest: architecture to Sofia K., the POS extract and the vendor Veridian to Priya N., reporting expectations to Dan R.

★ Try it now - charter draft with the gaps left openI am the delivery lead on Northwind. The only brief I have is the mandate email in this workspace. Draft a charter using exactly these fields, in this order: context · objective and success measure · in scope · out of scope · priorities P1 to P4 · constraints · assumptions · decision rights · named sponsor HOUSE RULES FOR THIS DRAFT - Use only what the email actually says. Do not infer a date, a budget, a headcount, a technology, or a scope boundary. - Mark anything the email does not settle as "TBC - ask [role]", naming the role who can answer: sponsor, data architect, vendor manager, BI lead. - Priorities must be an ordered list, P1 to P4, each with one line on what it means to give way. Do not mark two items P1. - Where you propose wording for a TBC field, label it "PROPOSED" so I can see it is yours and not Elena's. THEN, as a separate section, output GAPS: one line per TBC, in this shape - [field] · the question I should ask · who to ask · what it changes if the answer goes the other way Do not produce a timeline, a schedule, or a milestone date. I do not have one yet. Do not answer the GAPS questions yourself.
The version that fails - fluent, complete, and quietly fictionalWrite me a project charter for a data platform project. Make it professional and comprehensive, with clear objectives, scope, timeline and success criteria.
Why the second one fails It asks for "comprehensive", so every gap gets filled. You will get a twelve-week timeline nobody agreed, a success criterion of "improved data-driven decision making", and scope covering all nine systems because "comprehensive" implies it. Nothing in it is flagged as invented, so at first glance it reads better than the good one. That is precisely the danger: the honest draft looks worse and is worth ten times more, because you can see where the holes are.
Demo 2 of 2

The five non-goals, written down and agreed ★ 10 min · build your own

Now the field that pays for the whole session. Five sentences, agreed by Elena V. in week 1, that will each save you a fortnight somewhere between now and M6. Here is the Northwind set - yours will differ, but the shape will not.

Five sentences that stop scope arriving sideways No real-time POS Store ops, revisit after M5 No self-serve ML AI-headline asks; M6 stays one case Not replacing ERP Finance may ask; not replacing it No new BI tool Dan R.'s team; hides what failed No old archives 2 yrs history, not ten - phase 2 Five non-goals Elena signed in week one - unwritten, each becomes a corridor negotiation later.
🔍 Click to zoom - a written non-goal turns a corridor ambush into a nine-second answer
Non-goalWho will ask, and what it costs to say yes
No real-time POS dataStore operations, once the pipes exist. Streaming is a different architecture, cost and on-call rota. Revisit after M5
No self-service ML for every teamAnyone who has read an AI headline. M6 is one use case, deliberately. A platform for everyone is a programme of its own
We are not replacing the ERPFinance, when they see the modelled warehouse. Reading from the ERP is not replacing it. Never in this programme
No new BI toolDan R.'s team, reasonably. A tool migration inside a data migration doubles the risk and hides which one failed
No migration of historical archivesWhoever remembers 2019. Two years of history, not ten. Archive migration is a phase-two business case

Draft the five with the prompt below, using the charter and the mandate email as the only sources. Sharpen each into one sentence a sponsor can say yes or no to in ten seconds - if it needs a paragraph of caveats, it is not agreed, it is deferred.

Add the "instead" for each: later phase, another team, or never. A non-goal with no destination becomes a grudge.

Take all five to Elena V. in the same conversation as your four questions and ask her to strike any she disagrees with. Then put the survivors at the top of the charter, not in an appendix, and paste them into the workspace context pack so every later draft knows what you already cut.

★ Try it now - draft the non-goalsFrom the charter draft and the mandate email in this workspace, propose five things this programme is explicitly NOT doing in this phase. For each one give me: - the non-goal, in one plain sentence a COO can agree or reject in ten seconds - who will realistically ask for it, and roughly when in the delivery - what saying yes would cost us: scope, architecture, risk, or people - where it goes instead: a later phase (say which), another team, or never Rules - Draw them from what the mandate email leaves open, not from generic data-platform scope lists. - Do not soften them. "We may consider this later" is not a non-goal. - Do not invent a phase-two plan or a date to defer things to. If the destination is not settled, write "TBC - ask sponsor". - Flag any that contradict something the email does ask for, so I catch it before Elena does.
Real world

The line in the appendix nobody read. One programme did write its non-goals - and buried them on page nine of a charter deck. In month four a director asked for the archive migration, the delivery lead said "that is out of scope", and the director said, genuinely, "since when?" Technically since week 1. Practically since never, because it had been agreed by silence rather than out loud. They lost the argument and six weeks. Non-goals belong on page one, read aloud, with a nod from the sponsor. The document is not the agreement - the conversation is, and the document is what you keep afterwards.

Homework

Try it yourself - this week ◐ 30-45 min total

Source material

Official sources covered

Taught from the delivery canon plus the by-hand prequel course, with vendor documentation for the workspace mechanics. PMI certification and the full normative text of its standards stay with PMI. This session covers:

PMBOK Guide 7th ed. - planning performance domain + stewardshipParts 1-2 · charter fields, scope boundaries, sponsor and delivery-lead split
learn-tech-project-pmo-with-phoebe - the Northwind charter, by handParts 2-3 · the same artifact built without AI, for comparison
Anthropic docs - Projects, workspace instructions, contextDemos · where the charter and GAPS list live; depth in b1 and b9
Check yourself

Three questions before you go 🎯 ◐ 90 seconds

1 · Your charter draft says "delivery by mid-November". Elena's email said only "ideally before the Christmas peak". What do you do?

Rail two, and the most common smuggled fact of all. Nobody was asked and nothing was capacity-checked, so mid-November is invention with a professional font. AI drafts structure, never commitments.

2 · Your sponsor marks three of the four priorities as P1. Why does that break the charter?

Priorities exist to settle collisions. Everything-is-P1 carries no information and quietly hands the decision to whoever escalates loudest on the day. Push for the order, even if it takes two attempts.

3 · Week 6, a regional director asks for live till data on his phone. What makes that a nine-second conversation instead of a three-week investigation?

Scope arrives sideways, in corridors, as reasonable requests. The only cheap defence is a non-goal that was written down and agreed out loud in week 1 - on page one, not in an appendix.

PM session 2 cheat sheet · pin this

A mandate is an outcomeThe sponsor owns why and the money. You own the shape of the work. Convert, do not complain.
Five things it never sayssuccess measure, scope boundary, priority order, constraint reality, decision rights.
Nine fieldscontext, objective + measure, in scope, out of scope, P1-P4, constraints, assumptions, decision rights, named sponsor.
Ordered or nothingEverything cannot be P1. An unordered priority list resolves no trade-off.
Non-goals save youFive sentences, agreed out loud, on page one. Scope arrives sideways, not by change request.
Ask for the questionsThe best charter-stage output is the list of what you must ask, with the model refusing to answer it.
TBC - ask [role]Every gap carries a name: sponsor, data architect, vendor manager, BI lead. Never a blank.
Four, not tenTake four questions to the sponsor, each with a proposed answer and what it changes. Route the rest.