Engineering and architecture
Engineering firms and AI in 2026: how to read the ACEC and BST surveys
By SourceX Editorial · Updated
Short answer
Engineering firm AI surveys in 2026, including ACEC Research Institute's business sentiment research and BST Global's AI and data report, track strategy, investment and barriers. Broader 2026 surveys add sourced numbers: adoption is wide but shallow, and data quality is the top reported obstacle. None ask whether a firm's own project records have outside value.
Key takeaways
- Engineering-specific surveys track AI strategy, investment and barriers; read the original releases for their figures and samples.
- RSM's 2026 middle-market survey found data quality and availability was the top inhibitor to AI deployment.
- Census survey data shows larger firms report higher AI use than smaller ones.
- No major survey asks A/E firms whether their project records could be licensed to AI developers.
- The practical response for a mid-sized firm is the same either way: name data owners, fix standards and settle rights.
What do the ACEC and BST surveys measure?#
The ACEC and BST surveys measure how engineering and design firm leaders see AI: whether they have a strategy, where they are investing and what holds them back. ACEC Research Institute runs recurring business sentiment research among engineering firm leaders, and BST Global, which makes ERP software for AEC firms, publishes research on AI and data in the industry.
Both are useful because they ask firm leaders rather than technology vendors. Both also have limits worth stating plainly: respondents self-report, samples lean toward firms engaged enough to answer, and the questions treat firms as buyers and users of AI. Read the original releases for exact figures, sample sizes and question wording before quoting them in a board deck.
Across engineering-focused research and the wider surveys below, one theme repeats. The constraint is less access to tools than the state of a firm's own data, standards and skills.
What do broader 2026 surveys add?#
Broader 2026 surveys add sourced numbers that engineering leaders can use as context. They are not engineering-specific, so treat them as the backdrop for A/E findings rather than a substitute.
These numbers describe adoption, not results. Bain's September 2026 work on private equity found little correlation between AI spend and value at most portfolio companies, a useful caution for any firm tempted to measure progress by the number of tools it has bought.
| Source | Finding | What it suggests for an A/E firm |
|---|---|---|
| RSM middle-market AI survey, released July 21, 2026, 1,030 U.S. and Canadian leaders | 86% have partially or fully integrated AI into operations, but only 36% have AI fully embedded across core processes; 73% use generative AI. | Experimenting is common; embedding AI in project delivery and finance is where firms separate. |
| RSM 2026 survey, inhibitors | Data quality and availability was the top inhibitor (34%), ahead of security and privacy (30%), legacy systems integration (28%) and talent gaps (28%). | Cleaning project, time and document data is the first AI investment, not the last. |
| U.S. Census Bureau Business Trends and Outlook Survey, period ending May 3, 2026 | 32% of firms with 100 to 249 employees and 37% of firms with at least 250 employees said they used AI. | Mid-sized firms are not late; many peers are still deciding. |
| McKinsey, The State of AI in 2026 | 40% of respondents from large organizations report scaling AI agents, up from 27%, while smaller organizations stayed flat at 22%. | Large clients and competitors may move faster on agents, which raises expectations of their consultants. |
- 01Your work records
- 02Personal details removed
- 03You approve the release
- 04Licensed to AI developers
Where engineering firms say AI gets stuck#
Engineering firms find that AI gets stuck on their own information more than on software. Project knowledge sits across Deltek or BQE for time and billing, Bluebeam markups, Revit and Civil 3D models, file servers full of calculations, and email threads that hold the real decisions.
Each of these is a records problem before it is an AI problem. That is why surveys reporting data quality as a barrier and surveys reporting slow scaling are describing the same thing from two angles.
- Inconsistent project numbering across the ERP, file shares and document control.
- Calculations and QA/QC comments stored as PDFs with no link back to the project record.
- Professional liability concerns about AI drafting anything that will carry a seal.
- Client contracts that restrict reuse of deliverables or require confidentiality.
- Few senior staff with time to curate data while billable targets stay in place.
What the surveys do not ask#
The surveys do not ask whether an engineering firm's records have value to anyone outside the firm. They treat the firm as an AI user, while AI developers increasingly look for expert, real-world work records to train and test models on professional tasks.
For A/E firms those records include RFI logs with responses, submittal reviews, QA/QC comments with resolutions, change records and proposal-to-project histories. Their value depends on whether records connect a question to a decision and an outcome, on privacy preparation and on rights. Ownership of instruments of service and client confidentiality terms decide what is in scope, and they are assessed project by project with counsel.
What the findings mean for a mid-sized engineering firm#
For a mid-sized engineering firm, the findings point to a short, practical agenda that pays off whether or not AI tools change again next year.
None of this needs a large budget. It needs a decision about who owns each system's data and a calendar for keeping it clean, written into someone's job rather than added to everyone's.
| Survey signal | Action for the firm | Owner |
|---|---|---|
| Data quality is the top barrier | Standardize project numbers, phase codes and folder structures across ERP and file servers | COO and IT lead |
| Adoption is wide but shallow | Pick two workflows, such as proposal drafting and QA/QC tracking, and measure them properly | Principal in charge of operations |
| Talent and skills gaps persist | Name a data owner for each core system and give them non-billable time | CEO |
| Security and privacy concerns rank high | Adopt an AI use policy covering client data, sealed work and approved tools | General counsel or risk manager |
| No survey prices the records | Inventory record families and years of history using metadata only | CFO |
Illustrative: a regional engineering firm reads the surveys#
Illustrative: the CEO of a fictional regional water and transportation engineering firm reads the year's AI surveys before a board retreat. The firm runs Vantagepoint and Bluebeam, with a shared drive organized by office rather than by project, and several AI pilots that nobody has measured.
Instead of buying another tool, the board approves a records project: one project numbering scheme across systems, QA/QC comments captured in a log rather than only on marked-up PDFs, and an inventory of record families and history by system. The pilots get clean inputs, and when a private equity suitor later asks about data, the CEO answers with an inventory instead of a guess.
Where SourceX fits#
SourceX helps established firms decide whether their operational records can be licensed to AI developers, using the SourceX five-step transaction: Supply, Rights, Preparation, Approval and Delivery. The first step is a metadata-only fit check, and no files are shared during the initial assessment.
Records are assessed with the SourceX Enterprise Data Value Framework, a SourceX-developed methodology with qualitative ratings. Domain expertise, human-generated signal and rights raise value, while preparation cost and privacy burden reduce net value. SourceX does not publish prices, and its dataset rights are set out in the signed supplier agreement; records are licensed, not sold, and the firm keeps ownership.
Frequently asked questions
Where can I read the ACEC and BST survey results?
Go to the publishers directly: ACEC Research Institute posts its sentiment research and BST Global posts its AI and data research. Use the original release for figures, sample sizes and question wording, and note the survey date, because AI questions change from one wave to the next.
Do these surveys apply to architecture firms too?
Partly. BST Global's research covers AEC firms broadly, while ACEC focuses on engineering firm leaders. Architecture firms share many of the same barriers, such as scattered project data and liability concerns, but their records, contracts and client relationships differ enough that findings should be read with care.
Should a mid-sized firm hire an AI lead because of these surveys?
Hire for a defined job, not a trend. Many firms get further by naming data owners for Deltek, document control and QA/QC, then bringing in outside help for one specific workflow. A dedicated AI role makes sense once there are several workflows to run and maintain.
Why would AI developers want engineering records?
Models that assist with professional work need examples of how experts handle real questions: an RFI and its answer, a review comment and its resolution, a change and its approval. Such records are scarce in public data. Whether a given firm's records qualify depends on rights, privacy preparation and how well the records connect.
Does licensing records conflict with using AI internally?
No. Internal AI use and licensing are separate decisions. A license grants defined use of prepared copies for an agreed term, while the firm keeps ownership and keeps using its records. The same data cleanup supports both.
Sources
- RSM's 2026 middle-market AI survey of 1,030 senior leaders in the U.S. and Canada, released July 21, 2026, found 86% have partially or fully integrated AI, only 36% have AI fully embedded across core processes, and 73% use generative AI; data quality and availability was the top inhibitor (34%), followed by security and privacy (30%), legacy systems integration (28%) and talent gaps (28%). Source
- Census reported that in the period ending May 3, 2026, 32% of firms with 100 to 249 employees and 37% of firms with at least 250 employees said they used AI. Source
- McKinsey's State of AI 2026 report found 40% of respondents from large organizations report scaling AI agents, up from 27%, while smaller organizations stayed flat at 22%. Source
- Bain's September 2026 piece says that for most portfolio companies there is little correlation between AI spend and value. Source
Related resources
See if your company qualifies
A short company assessment. No data uploads are needed.