Engineering and architecture
Data readiness checklist for architecture and engineering firms
By SourceX Editorial · Updated
Short answer
A data readiness checklist for an architecture or engineering firm tests whether project records can be found, exported, linked by project number, cleared for use and stripped of personal details. Score the 20 points as met, partial or not met; a firm in the top band can start a licensing fit check or internal AI project without long cleanup.
Key takeaways
- Readiness depends on linked records, not storage volume: an RFI log tied to responses beats a full file server with no index.
- One consistent project number across Deltek or BQE, Procore and file storage is the most useful single readiness fix.
- Rights and privacy belong in the readiness score, because a tidy archive can still be closed to outside use.
- Fix gaps in systems and project logs first; rights and privacy work goes faster once records are findable.
What does data readiness mean for an A/E firm?#
Data readiness for an A/E firm means the records behind past projects can be located, exported in usable form, connected to each other and cleared for a new use. It describes the archive and the paperwork around it, not any particular AI tool.
The checklist groups 20 points into five areas: systems, project logs, rights, privacy and archives. Mark each point met, partial or not met, count a met point as one and a partial point as a half, and compare the total with the scoring band at the end.
Score from what IT and project managers can confirm today. If nobody knows the answer, mark the point not met; unknowns are usually the cheapest gaps to close.
Systems: where does each record family live? (points 1 to 5)#
The systems points test whether the firm knows which platforms hold its project history and whether that history can be exported with its links intact.
Point 5 is the one most worth testing early, even though it scores the same as the others. A system named on paper but never exported tends to hide surprises: fields that come out as internal codes, attachments that need a separate download, or history that an earlier migration trimmed.
| Point | Met when |
|---|---|
| 1. Project and financial system named | Deltek Vantagepoint, Vision, Ajera or BQE Core is identified as the record of projects, phases, hours and billing, with the years it covers. |
| 2. Project management platform named | Procore, Newforma, Autodesk Construction Cloud or a similar platform is identified as the home of RFIs, submittals and meeting minutes. |
| 3. Review tools named | Bluebeam Revu and Studio, or another markup tool, is identified, and someone knows whether markup lists can be exported. |
| 4. Models and drawings mapped | Revit and CAD files, specifications and calculation packages have a known location and folder convention. |
| 5. Export route tested | Someone has exported a sample from each system with record IDs and dates intact, not only a PDF report. |
Project logs: are decisions recorded with outcomes? (points 6 to 9)#
The project log points test whether the archive captures questions, decisions and their results, which is what makes A/E records useful for AI work. A drawing set shows the final answer; the logs show how the team reached it.
Logs kept only in email are the most common partial score. They count as met once someone can export them with the project number and dates attached, even if the export is a folder of messages rather than a structured log.
- 6. RFI logs link each question to its response, the responder and the date closed.
- 7. Submittal logs record each review action, such as approved, approved as noted or revise and resubmit, with the reviewer's comments.
- 8. QA/QC review comments are kept with their resolution instead of being discarded once the set is issued.
- 9. Change records connect each change request or ASI to the fee, cost and schedule decision that followed.
Rights: is the firm free to use the records? (points 10 to 13)#
The rights points test whether contracts and third-party terms leave room to use project records beyond the original project. A well-organized archive can still be closed to outside use if client agreements say so.
Score point 11 as partial when the flags live only in a principal's memory. The flag has to sit on the project record itself so every later export, report or AI tool can filter on it.
| Point | Met when |
|---|---|
| 10. Contract forms known | The firm knows which projects used AIA, EJCDC or ConsensusDocs forms and which used client-drafted agreements. |
| 11. Confidentiality flagged | Projects with strict confidentiality, NDAs or no-use clauses are flagged on the project record. |
| 12. Client-owned material separated | Deliverables and client-furnished information the client owns or controls are distinguished from internal working records. |
| 13. Third-party content mapped | Subconsultant drawings, manufacturer product data and platform terms that limit reuse are noted per project. |
Privacy and archives: the last seven points#
The privacy points test whether personal details can be found and removed, and the archive points test whether closed projects will still be readable after staff, systems and folder conventions change.
Buyers of licensed records also commonly ask for a written account of where data came from and what use is allowed. The Data & Trust Alliance's Data Provenance Standards, an open specification, group that kind of metadata into Source, Provenance and Use, which is a sensible model for documenting an archive.
- 14. The firm knows where personal details sit: names and phone numbers in RFIs, homeowner addresses, employee names in markups and timesheets.
- 15. Employee files, pay data and HR records are stored apart from project records and excluded by default.
- 16. Names can be replaced with consistent placeholders, so one reviewer stays one reviewer across records.
- 17. A person reviews the output after automated scanning, because detection tools miss things.
- 18. Closed projects are archived in one consistent structure, not as unlabeled zip files on an old server.
- 19. Records from retired systems, such as a previous accounting or document platform, were exported before access ended.
- 20. Projects inherited from acquired firms keep their original project numbers and contract files, so their rights can be checked.
Scoring band: what your total means#
The scoring band turns the 20 points into a next step. Treat it as a guide to sequencing work rather than a pass or fail grade.
| Score | What it usually means | Next step |
|---|---|---|
| 16 to 20 | Records are documented, linked and mostly cleared | Run a metadata-only fit check or scope an internal tool |
| 10 to 15.5 | Good records with gaps in logs, rights or privacy | Close the missing points for one or two record families first |
| Below 10 | History is scattered or undocumented | Start with points 1 to 5 and a project number standard |
Illustrative: an architecture practice scores itself#
Illustrative: a fictional architecture practice with healthcare, workplace and multifamily studios scores itself and lands in the middle band. Its Deltek and Procore records score well, but RFI responses from before its Procore rollout live only in project email folders, and nobody has flagged which clients signed strict confidentiality terms.
The COO assigns two fixes: export the email-era RFI folders with project numbers attached, and add a confidentiality flag to the project record in Deltek. Scored again for the Procore-era RFIs and submittals alone, the practice reaches the top band, so that record family is assessed first while the email-era history waits for a second pass.
How SourceX uses a readiness check#
SourceX begins with the Supply step of the SourceX five-step transaction, which overlaps closely with the systems, logs and archive points above. The fit check asks for metadata only, such as system names, years of history, record families and known restrictions, so no project files leave the firm at that stage.
The rights and privacy points feed the Rights and Preparation steps. Their results are written into a SourceX Evidence Packet, so the firm can see what was cleared, what was removed and what it approved before anything is delivered.
Frequently asked questions
Who should fill in the checklist?
The COO or director of operations usually owns it, with IT scoring the systems and archive points, a senior project manager scoring the logs and the risk manager or counsel scoring rights. Splitting it this way gets honest answers quickly, because each person scores only what they actually know.
Do we need to clean every project before starting?
No. Readiness is judged per record family and per period. A firm can be ready for RFIs and submittals from its current platform while older projects remain in cleanup. Start where the score is highest and widen the scope later.
Does a low score mean our records have no value?
No. A low score usually means the history exists but is hard to find, link or clear. Gaps such as a missing project number standard or an untested export are often fixed with modest effort, and the underlying records can still be useful once they are organized.
Should BIM models be part of a readiness review?
Map them, yes; include them in a license, cautiously. Revit models often embed client-owned design content and consultant work, so they usually sit behind a rights review. Records about the models, such as clash reports and review comments, are often easier to clear.
Is this checklist only useful for licensing?
No. The same points decide whether an internal search or drafting tool will work well. A firm that can find, link and clear its records gets better results from its own tools and keeps the option of licensing later.
Sources
- The Data & Trust Alliance's Data Provenance Standards (version 1.0.0 specification) define dataset metadata in three groups: Source, Provenance and Use, needed to enable proper dataset selection for AI model training. Source
Related resources
- QuestionWho owns enterprise data?
- QuestionDo AI labs buy code?
- InsightHow do I de-identify source code for AI training?
- InsightHow do I de-identify CAD and engineering drawings for AI training?
- InsightHow do I de-identify code reviews and pull requests for AI training?
- SolutionData partnerships between businesses and AI developers
See if your company qualifies
A short company assessment. No data uploads are needed.