Skip to content

Lesson 6 of 6

Get the Plan Agreed and Keep the Scope Under Control

Present the plan so that it is agreed in the room, answer the three reactions you should expect, and handle later requests with a trade and a change log.

6 minForward Deployed Engineering

After this lesson you will be able to present a build plan so that the customer agrees to it, respond to the questions it will raise, and keep the scope from growing without anyone deciding that it should.

Present the plan for a decision

A plan that has been emailed has not been agreed. Hold a short meeting, thirty minutes is enough, with the four kinds of people from lesson 2: the sponsor, somebody who does the work, the owner of the systems and the person who can say no. At Harrow that is the operations director, a person from the order desk, the IT contractor and the finance director.

Send the page the day before, so that the meeting can be spent on questions. Then go through it in this order:

  • The problem and the evidence, in the customer's words and numbers.
  • The options you considered, including the one that was asked for.
  • What you recommend, and what the customer will see first and when.
  • What is out of scope, read aloud, item by item.
  • What you need from each person in the room.
  • The decision you are asking for today.

Read the out-of-scope list aloud even though it feels awkward. An item that everyone heard and nobody objected to is agreed. An item that sat unread at the bottom of a page will be disputed later.

End by asking for the decision in so many words: "Are we agreed that this is the first piece of work, and can I have the emails and the export by Friday?". Write down who said yes.

Three reactions to expect

Three reactions come up in almost every one of these meetings. Prepare for each.

"Why is this not what we asked for?" At Harrow the question will be why the plan has no AI agent in it. Answer from the evidence and the customer's goal, as in lesson 4: most of the late orders come from three senders whose emails can be read reliably without a model, that gives a result in weeks, and step 4 of the plan measures whether a model is worth adding for the rest.

"Can it also do this?" Somebody will ask for phone orders, or replies to customers. Do not answer yes or no in the room. Say what it would take to find out, and note it in the change log described below.

"Can it be faster?" Say what the dates depend on. If the examples and the import format arrive on the first day, the early end of each range is realistic. If you are asked to drop the week in which the order desk compares drafts, explain what that week buys: it is how everyone learns whether the drafts can be trusted before real orders depend on them.

Trade, do not refuse

After the plan is agreed, requests keep coming, and most of them are reasonable. Agreeing to everything makes every date wrong. Refusing everything makes you hard to work with. The useful answer is a trade: something can come in if something else moves, and the sponsor chooses.

Suppose that in week 2 the order desk asks whether the tool could also read orders from a fourth customer. A good reply is: "Yes, that is possible. It would take about three days, so the measuring in step 4 would start three days later. Would you like that, or shall I put it on the list for after step 4?".

That reply does three things. It takes the request seriously. It states the cost in days and in what moves. And it hands the choice to the customer, who owns the priorities. Customers very rarely resent being given a clear trade.

Some requests are different in kind. If somebody asks for the person who confirms each draft to be removed, that is not a trade against a date. It breaks the finance director's rule, so the answer is that the finance director would need to agree first.

Keep a change log

Keep one list of every request to change the plan, whatever was decided. Put it at the bottom of plan.md or in a file beside it. Each entry has a date, what was asked, who asked, what was decided, and what it did to the plan.

## Change log

2026-03-17  Asked: read orders from a fourth customer (order desk).
            Decided: yes, by the operations director.
            Effect: step 4 starts three days later.

2026-03-19  Asked: send customers an automatic reply (sales).
            Decided: not in this piece of work. Review after step 4.
            Effect: none.

2026-03-24  Asked: import drafts without a person confirming (order desk).
            Decided: no. The finance director requires the check.
            Effect: none.

The log has two uses. When somebody asks in week five why step 4 is late, the answer is on the page with a name beside it. And when the work is finished, the log is a ready list of what the customer wants next, in their own words.

When to rescope

Sometimes the plan itself should change. That is not a failure, provided it happens openly. Three things justify it.

  • The evidence changes. The month of orders might show that the three regular senders are only a third of the email, and the first build would then remove much less of the problem than the sample suggested.
  • An early check fails. If the IT contractor finds that a file import does not run the credit check, the build as planned cannot go live.
  • The customer's priority changes. If Harrow loses a large customer over wrong deliveries, corrections may suddenly matter more than lateness.

When one of these happens, rewrite the affected parts of the page, say what changed and why, and get it agreed again by the same people. What you must not do is quietly build something different from the page. The plan is worth having only while it describes what is really happening.

How to tell that scoping is done

You can stop scoping and start building when all of these are true:

  • The problem is written down with evidence, and the sponsor agrees it is the problem.
  • There is a success measure with a baseline, and the sponsor has agreed the target.
  • The out-of-scope list has been read aloud to the people it affects.
  • The first milestone is no more than two weeks away, and somebody on the customer's side knows they will be asked to look at it.
  • Everything you need from the customer has a name and a date beside it.
  • Every risk has an early check.

Scoping the Harrow work took a few conversations, one export and one short script. The result is a page the customer has agreed to, a first build that is much smaller than the one requested, and a date on which the customer will see something. That is a good trade for a few days of work, and it is the part of an FDE's job that most often decides whether the rest succeeds.

Key takeaways

  • Send the plan a day early and use the meeting to get a decision, not to read the page aloud.
  • Expect to be asked why it is not what was requested, whether it can do more, and whether it can be faster; prepare each answer.
  • Answer a new request with a trade: what it would move, and who decides.
  • Record every request and its outcome in a change log, so that scope changes only by decision.
  • Rescope openly when the evidence changes, when an early check fails or when the customer's priority changes.

Try it

Do a whole pass on a new ask, in sixty minutes: "A chain of opticians says: we want a chatbot so customers can book appointments." Invent the answers the four kinds of people would give, including one rule from a person who can say no and one number you would measure. Then write the one-page plan with all nine headings, and a change log with two requests you expect and how you would answer each. A good result proposes a first piece of work that can be shown within two weeks and lists at least three things as out of scope, each with a reason.

Official documentation

Ready to use this in a job?

Rolebound applies to matching roles for you.

Start now

Written with AI and checked automatically. Nobody has run these steps by hand, and tools change, so a command may behave differently in the version you install. Spotted a mistake? Tell us.

    Cookies on Rolebound

    We use necessary cookies to sign you in and keep the site working — those cannot be turned off. We would also like to set analytics and advertising cookies so we can see which pages help people get interviews and which adverts brought them here. Nothing optional is set unless you say yes, and you can change your mind at any time from “Cookie settings” in the footer. Read our Privacy Policy.