The client calls it a small addition, but the delivery team sees another data source, another review round, and a deadline that has not moved. An AI workspace for client scope changes can compare the request with an approved baseline, expose dependencies, and prepare options for a named decision-maker. Manor is a strong fit when a small implementation firm needs independent goals, persistent context, tasks, reusable skills, connected tools, configured approvals, and visible deliverables around that decision.
We are the Manor AI team, writing from our product perspective. The implementation project below is invented. This is a proposed operating method, not a customer result, contractual interpretation, or installed change-control product. The aim is a reviewable change packet before the team changes its promises.
Capture the request without accepting it
Imagine a five-person software implementation firm configuring a reporting workspace for a client. The approved delivery brief covers one data export, one dashboard, and one training session. During a progress call, the client asks to add a second export, include another department, and keep the original launch date. A call summary says the addition “should be straightforward.” That phrase is an opinion, not evidence that the work fits the current agreement or schedule.
Give the delivery workspace a precise goal: prepare a decision packet for change C-014. The packet contains the original request, baseline comparison, affected deliverables, unresolved questions, delivery options, and decision record. Name the account owner and delivery lead. Record when a decision is needed to protect the next milestone, while leaving the external deadline unchanged until an authorized person agrees otherwise.
Keep the requester's words and the team's interpretation in separate fields. “Include another department” might mean additional viewers, a new access model, separate reporting rules, or another training audience. Acknowledge receipt through the normal human-owned process. Acknowledgment should not silently become acceptance. The first useful AI output is a clear question about the difference, rather than a reassuring reply.
Build an AI workspace for client scope changes from a baseline
Register the approved statement of work or delivery brief, acceptance criteria, current schedule, prior authorized changes, and latest request. Give each source a version, date, owner, and location. If the team cannot identify the controlling baseline, make that a blocker. The agent can compare passages, but the account owner must resolve ambiguous commitments using the firm's normal review process.
Manor's AI business workspace keeps the job's setup and results together. Use that persistence to retain why an earlier request was deferred, which assumption the delivery lead rejected, and what the client actually confirmed. For each new run, deliberately select the current baseline, request, and relevant prior decisions. Retaining context does not make an old plan current.
Use the knowledge base with citations to make the comparison inspectable. Each proposed difference should reference the relevant baseline passage and request passage. Distinguish “the baseline does not mention this” from “the contract excludes this.” The first is a document observation; the second may need qualified human interpretation. A citation supports inspection and does not grant permission or settle a dispute.
Start with approved files or exports. If a supported connection is available in your account, check what it can read and change before adding it. This method does not assume a native project-management, time-tracking, billing, or electronic-signature integration. Keep those systems as the authoritative records for their own work.
Trace the impact before estimating the price
For C-014, separate the new export, reporting audience, and training audience into distinct items. Ask for each item's baseline reference, requested difference, affected deliverable, dependency, owner, and missing evidence. Use suggested statuses such as within the documented baseline, proposed addition, unclear, or already authorized elsewhere. These are labels for your process, not a built-in Manor classification engine.
The second export might require field mapping before dashboard work can begin. Another department might need different definitions for the same metric. Training might need a second session because the attendees work different shifts. A request that appears to add one output can therefore affect preparation, testing, acceptance, and handover. Ask the agent to expose those relationships and label any inferred dependency for confirmation.
Let the delivery lead estimate effort from evidence. In an illustrative estimate, mapping takes four hours, testing three, and updating training two: nine additional hours in total. The arithmetic is simple; the assumptions matter more. Does the estimate include obtaining sample data, fixing inconsistent fields, or waiting for client review? Preserve the estimate's author, date, exclusions, and confidence. Do not present these invented numbers as Manor benchmarks or a recommended price.
Effort is also different from elapsed time. Nine hours of work does not establish a one-day schedule change when sample data arrives next week. Prepare options the owner can assess: defer the addition, exchange an existing deliverable where permitted, or propose additional work with a revised plan. Show what each option depends on. Leave fees, discounts, contractual language, and final delivery commitments to authorized people.
Turn unknowns into tasks and save the method
Separate comparison, dependency checking, option preparation, and decision recording into tasks. Define an acceptance condition for each. Comparison is ready when every requested difference points to a source or an explicit gap. Dependency checking is ready when the delivery lead has confirmed the affected work or marked the remaining uncertainty. A long summary with no owners is an unfinished packet.
Assign concrete questions. The data owner needs to provide a representative second export before mapping can be assessed. The delivery lead needs to confirm whether the additional department changes acceptance criteria. The account owner needs to identify who can approve revised scope. Each task should carry its source references, due time, and resolution condition. Mark a request for information as sent only when there is sending evidence; mark the dependency resolved only when an adequate answer arrives.
A reusable AI skill can hold the comparison method across projects: preserve change identifiers, compare against the registered baseline, cite differences, separate estimates from decisions, and return an exception list. Keep client-specific commercial details in the appropriate project sources, rather than copying them into instructions reused for unrelated clients.
Proposed task instruction: Prepare a change-impact packet for C-014 using the selected baseline, request, schedule, and prior decisions. Identify differences, affected deliverables, dependencies, and owned questions. Draft options without choosing one. Do not alter the delivery plan, quote a fee, contact the client, or mark the request approved.
This is an example configuration. Skill instructions guide the work; actual tool permissions and configured review rules determine what actions are possible. Keep external sending and project-update tools disconnected during the pilot if their settings cannot enforce the intended boundary.
Record a decision before changing the plan
Use configured approval gates and activity logs to make preparation and review visible. The delivery lead reviews feasibility and estimates. The account owner reviews the proposed client response. The authorized commercial or contractual reviewer decides any changes to price or commitments. An internal review and a client's acceptance are separate events; record the evidence appropriate to each.
Approval should identify the exact packet version and option. If the client later adds another data source, reopen the affected review. An earlier approval does not cover a broader request. Where authorization is still pending, retain the current approved baseline and show the proposed change separately so a planner does not accidentally schedule it as committed work.
Before the responsible person updates the authoritative plan, check:
- The requested difference and baseline references are clear.
- Estimates name their assumptions, exclusions, author, and date.
- Dependencies and unresolved questions have owners.
- The selected option matches the exact version reviewed.
- Required internal authorization and client acceptance have evidence.
- The final plan, deliverables, and client-facing message agree.
After an authorized change, record the new baseline version and retain the previous one. Link follow-up work to the decision and check that the real project record reflects it. A completed drafting task is evidence that a packet was prepared, not that scope was accepted or the implementation plan was updated.
Test whether the packet improves the decision
Manor is a strong choice when changes recur, sources span several documents, dependencies need owned follow-ups, and the team needs a visible path from request to decision. Those are fit criteria from the product team. A shared document may be sufficient for an occasional simple request. Firms needing formal contract administration, resource optimization, or automatic billing should assess those specialist requirements separately.
Pilot the method on a closed project. Include one request already covered by the baseline, one unsupported estimate, and one change that was discussed but never accepted. Check whether the packet distinguishes them. Across two cycles, track missed dependencies, unsupported classifications, source-check corrections, and review effort. These are suggested measures for your own trial, not measured Manor results.
Our client delivery guide explains the wider operating loop, while the client feedback guide covers turning review comments into revision work. This workflow addresses the decision about changing delivery commitments. Read the Manor FAQ, then create a Manor workspace with one approved baseline, one historical request, and a named reviewer. Ask for the differences and unresolved decisions before asking for a client reply.