A scope of work fails at its edges, not its centre. Everyone agrees on the main deliverable; the argument is about whether revisions, integrations, content, training or post-launch fixes were part of it. A document that lists inclusions without naming exclusions leaves those questions to be settled later, under worse conditions.
Every structure below has the same four parts: objective (what changes for the client), inclusions, exclusions, and change control (how scope changes get priced and approved). Add dependencies — what you need from the client, and by when — because that is the clause that decides whether a slipped date is your problem or theirs.
The six structures
1. Fixed scope
Use when: the deliverable and the end point are both definable now.
Objective: [what changes for the client]
Included:
· [Deliverable 1] — [one line defining its extent]
· [Deliverable 2] — [one line defining its extent]
· [N] rounds of revision on [specific artefacts]
Excluded:
· [Adjacent item 1] — available as a separate quote
· [Adjacent item 2]
Dependencies: [item] from [team] by [date]
Change control: Changes are quoted in writing before work starts. Approved changes adjust both fee and timeline.
2. Milestone-based
Use when: the work is long enough that the client needs evidence before releasing the next payment.
| Milestone | Deliverable | Acceptance | Date | Fee |
|---|---|---|---|---|
| 1 · [name] | [deliverable] | [how it is accepted] | [date] | [fee] |
| 2 · [name] | [deliverable] | [how it is accepted] | [date] | [fee] |
| 3 · [name] | [deliverable] | [how it is accepted] | [date] | [fee] |
Use when: the client wants to see output before committing the full budget. The acceptance column matters more than the deliverable column.
3. Retainer
Use when: the value is continuity rather than a single artefact.
Monthly capacity: [N hours / N deliverables]
Recurring deliverables: [item], [item]
Response time: [timeframe] for [request types]
Rollover: [policy on unused capacity]
Excluded: [item], [item]
Use when: the work is steady and the client wants priority access. State what happens to unused hours, because that is the clause people remember.
4. Pilot or trial
Use when: the client wants proof before a larger commitment.
Question this pilot answers: [the specific decision]
Scope: [delimited work]
Duration: [fixed window]
Output: [concrete artefact]
Fee: [amount], credited against [larger engagement] if you proceed within [period]
5. Time and materials
Use when: the scope genuinely cannot be known in advance — investigation, debugging, discovery.
Rates: [role] at [rate]; [role] at [rate]
Estimate range: [low]–[high]
Notification threshold: we will stop and confirm before exceeding [amount]
Excluded: [item]
Use when: scoping the work would cost more than doing it. The notification threshold is what keeps this arrangement from becoming an open cheque.
6. Audit or assessment
Use when: the deliverable is a finding rather than a build.
What is examined: [specific scope]
Method: [how]
Output: [report with findings and priorities]
Not included: [remediation / implementation]
Use when: the client needs to know where they stand before deciding what to fix. Say explicitly that fixing things is a separate piece of work.
The clauses worth keeping in all six
Three clauses prevent most disputes regardless of structure. A revision definition — what counts as one round, and what constitutes a change of direction instead. A dependency clause — client inputs and their deadlines, with the effect of a missed date. An acceptance window — how long the client has to review before a milestone is considered accepted.
Where AI helps and where it does not
Generating the skeleton is worth it. The four-part structure, the standard clause phrasing, the table layout — a tool produces these in a minute and they transfer between projects. Generating the specifics is not. Inclusions, exclusions, dependencies and rates are things you know and the model does not; drafting them produces confident text that does not match your engagement, and the error surfaces later as a dispute.
What "done" means
A scope document that does not define completion leaves the ending open to negotiation. State what done looks like for each deliverable — the file formats supplied, the environment it is installed in, the number of review rounds included, and what happens to items still open at the end date. Also state what is explicitly not part of completion: training, documentation beyond a stated scope, post-launch monitoring, and content or assets supplied by the client. The argument is rarely about the work; it is about where the work stopped.
Template the structure, write the boundaries
A tool gets the document started. The exclusions are the part that protects you — write those yourself.
Try Rytr →Frequently Asked Questions
What is the difference between a scope of work and a contract?
The contract is the legal agreement; the scope of work describes what will actually be produced. It is usually attached to and referenced by the contract.
Why do I need an exclusions section?
Because silence gets read as inclusion. Naming what is outside the scope prevents the most common dispute — the client assuming adjacent work was covered.
How many revisions should I include?
A specific number, tied to specific artefacts, with a statement that further changes are quoted. "Unlimited revisions" invites indefinite work.
What is change control?
The process for handling scope changes: quote in writing, approve, then adjust fee and timeline. It prevents drift from becoming a disagreement.
Should a pilot be free?
Usually not. A paid pilot with the fee credited against the larger engagement filters for serious interest and creates commitment on both sides.
Can AI write our scope of work?
It can produce the skeleton and standard clauses. Inclusions, exclusions, dependencies and rates must come from you.