[ TL;DR ]
More engineering time. Less work assembling the record.
A revised brief arrives. The client asks whether the design still works. A meeting agrees the next steps. Before an engineer can answer, someone has to compare documents, find the relevant correspondence and establish what has already been decided.
CSM built a demonstration workspace that connects those tasks. It prepares a source-linked change register, drafts query replies from the project record and turns meeting notes into proposed actions for review. Engineers can check the evidence and make the decision with the relevant information already brought together.
The example follows Riverside House, a fictional building project for Northline Engineering. A change from 60 to 90 occupants exposes a mismatch in the design record. That single issue becomes a reviewed finding, an update to an existing action, a carefully qualified client reply and a scope review for the project manager.
The opportunity is to return capacity to experienced engineers, make outstanding decisions easier to see and identify additional work while the client conversation is still current.
• Less preparation: reduce repeated searching, comparison and retyping around each decision.
• Connected evidence: keep the source passages with the finding, reply and follow-up.
• Clear ownership: named people review the work, approve updates and take responsibility for the next step.
[ The Challenge ]
A small change can create a large coordination task
The client changes one requirement in a brief. The engineer needs to establish where it differs from the agreed baseline, which design assumptions it may affect and whether the team has already opened an action to deal with it. That can mean moving between specifications, schedules, emails and meeting records before any technical assessment begins.
The cost is more than the time spent comparing files. If the change remains disconnected from the rest of the record, a client reply may repeat an outdated assumption, a meeting may create a duplicate action, or additional engineering may begin before anyone considers the agreed scope.
Good information management depends on knowing both the revision of a document and its status. A newer version is not automatically approved for every use. UK BIM Framework guidance explains how revision and status information support the common data environment workflow. The same practical distinction matters when AI gathers evidence for an engineer. Source: UK BIM Framework
[ What CSM Built ]
Three connected workflows around one project record
The Engineering Coordination workspace sits within the CSM AI Operating System. It brings the project’s selected source documents, change register, queries, meeting records and actions into one review process.
01
Change review. Compare the revised brief with the baseline and prepare differences, discrepancies and source references for an engineer to check.
02
Query preparation. Gather relevant project records and draft a reply that distinguishes known facts from questions still awaiting assessment.
03
Meeting follow-up. Separate decisions, actions and open questions, then propose updates to the project register.
The proof of concept uses seven prepared fictional document extracts and prepared outputs. It demonstrates how the workflow fits together. A client implementation would require the source connections, permissions and AI behaviour to be configured and tested against the firm’s actual records.
[ Change Review ]
Find the difference and prepare the engineering review
A useful change register gives the engineer more than a list of edited sentences. It shows the previous requirement, the revised wording, the related project record and the question that now needs a decision.
In the demonstration, the reviewer compares the baseline brief with the revision, cross-checks the selected design record and existing actions, and prepares findings with their source passages. The engineer checks those findings before any register update is approved.
[ A Revised Brief at Riverside House ]
The brief says 90 occupants. The design record still says 60.
The revised Level 3 brief produces three changed requirements and one conflict in the prepared example. The central issue is occupancy: the project schedule retains an assumption of 60 people while the revised brief specifies 90. The workspace presents the discrepancy alongside the three source passages behind it.
01
Check the finding. An engineer confirms that the changed requirement and the mismatch have been represented correctly.
02
Find the existing work. Action A-014 already covers the area. The reviewer updates that action, preserving its identity and history.
03
Give the assessment an owner. The reviewer assigns Daniel Hayes and chooses the due date. Both are explicit decisions within the fictional scenario.
04
Keep the technical question open. The record shows that occupancy has changed. Whether the ventilation design accommodates the revised use still requires an engineering assessment.
This is where the value becomes tangible. The engineer starts with a specific discrepancy, its evidence and the related action, instead of reconstructing the issue across several files. The review can move towards the assessment that matters.
[ Answering Client Queries ]
Prepare a useful reply without making an unsupported commitment
The client asks whether the ventilation design can accommodate the revised occupancy. Answering requires more than finding the latest brief: the engineer needs the design assumptions, earlier correspondence and any assessment already on record.
The query assistant brings together five cited records in the prepared example and drafts a reply. It identifies the changed occupancy and a missing detail: which days the higher occupancy applies. Design suitability remains unconfirmed because the required assessment has not yet been completed.
An illustrative reply captures the distinction: “The revised brief increases Level 3 occupancy from 60 to 90. We still need to assess the ventilation design against that requirement. Please confirm which days the higher occupancy will apply so the assessment can use the intended operating pattern.”
The engineer checks the cited facts, resolves any conflicting information and reviews the commitment in the reply. Editing the draft removes its earlier approval. Nothing is sent from the proof of concept; the engineer takes the approved wording into the firm’s normal correspondence process.
That separation reflects an established coordination principle. Autodesk’s RFI guidance describes distinct stages for drafting, review, answering and closure, with additional coordination where required. A request for information is a technical question requiring clarification; preparing a draft is one stage in resolving it. Source: Autodesk
[ Meetings and Actions ]
Carry the decision into the work that follows
The meeting workflow reads authorised notes or a transcript and prepares decisions, actions, client questions and scope-review items for confirmation. Transcript timestamps let the reviewer check what was actually said.
In the example, the meeting record proposes an update to A-014 rather than opening the same work again. The reviewer approves that update. Dates that were never agreed remain visible as missing, so a polished set of minutes does not create a false commitment.
The resulting overview separates the assigned assessment, outstanding client clarification and scope review. People can see what remains open and who needs to act, with the earlier evidence still attached. The intended saving comes from less retyping and fewer repeated enquiries about the same issue.
[ Scope and Commercial Review ]
Make additional engineering visible while the change is current
A revised brief may create work beyond the original appointment. A new requirement can mean another assessment, a design revision or further coordination. When that work is absorbed informally, the project manager may only discover its extent after the engineering time has been spent.
The demonstration connects the occupancy change with the appointment and prepares a scope-review handover. It brings the evidence and questions together so the project manager can establish whether further assessment or design work falls within the agreed service.
No fee is proposed in the example. The extent of the additional work has not yet been assessed, and a changed requirement does not by itself establish entitlement to an additional fee. The value is an earlier, better-informed commercial conversation before the team commits to further work.
[ Results ]
One change carried through to evidence, ownership and follow-up
The demonstration connects three workflows around one fictional project. A revised brief leads to a change register, a documented conflict, an update to an existing action, a client-reply draft, a meeting record and a scope-review handover.
Less comparison and searching. The engineer receives a prepared register with source references. A client pilot would measure the effort saved in finding, comparing and checking representative revisions.
Better-prepared client replies. The draft brings the relevant facts and unanswered questions together. Review would check whether each statement is supported and whether technical commitments are appropriate.
Clearer follow-up. Decisions lead to reviewed actions with named owners. A pilot would assess duplicate actions, missing responsibilities and the effort needed to keep the register current.
Earlier scope visibility. Potential additional work reaches the project manager with its evidence. Review would establish whether that improves the timing and clarity of scope discussions.
These are demonstrated workflow capabilities and intended benefits. Northline Engineering and Riverside House are fictional, and the proof of concept uses prepared interpretation, drafting and extraction outputs. It does not establish measured client savings, live integrations or independently validated AI performance.
The next step is a pilot on real project records. Compare the existing process with the new one across revisions, queries and meetings, including human review, corrections and any extra capture work. That establishes the net time returned, the quality of the outputs and whether wider implementation is justified.