Architecture and Engineering Firm Case Study: From Three AI Referred Enquiries a Month to Sixteen
A six person architecture and structural engineering practice had three AI referred enquiries in a month. After adding named registration numbers, fixed fee bands, and priced project outcomes, the next month brought sixteen, and six became signed contracts.
A six person architecture and structural engineering practice in Bristol had never separated out how many new project enquiries started with someone asking ChatGPT or Perplexity for a recommendation. The office manager logged every enquiry as website, referral, or phone call without asking how the caller had actually found the practice. Once the practice asked every caller directly for one month, the number of enquiries that named an AI assistant came back as three.
The Month Nobody Was Tracking
The practice logged the source of every new enquiry over one month, covering everything from a loft conversion to a small commercial fit out. Eleven new enquiries came in total. Only three people mentioned ChatGPT or Perplexity as the reason they had found the practice, and none of the three asked about fees before asking whether the practice had handled a project like theirs before.
1. What The Website Was Actually Saying
The practice's own website described a multidisciplinary team and a track record of excellence, with all six architects and engineers sharing one group photograph and no page separating a residential extension specialist from the engineer who signs off structural calculations. Fees were described only as a competitive percentage of project value, with no stated range for a loft conversion, a single storey extension, or a new build. Past projects appeared as a gallery of finished photographs with no build cost, floor area, or planning outcome attached to any of them. To an AI assistant comparing a dozen practices in the same region, there was nothing specific enough to repeat back to someone asking for a recommendation.
2. Three Changes, Made In Thirteen Days
Nothing about the practice itself changed. No new hires, no rebrand, no change in fees beyond stating them. Three specific facts were added to the existing website over thirteen days.
- A named profile for each of the six architects and engineers, stating their RIBA or Institution of Structural Engineers registration number, their specialism, and how many years they have held it
- Fixed fee bands replacing the competitive percentage line, stating a fee range for a loft conversion, a single storey extension, and a new build under four hundred square metres
- Twelve past projects rewritten with their exact floor area, final build cost, and planning permission outcome, replacing the unlabelled photo gallery
3. The Following Month Told A Different Story
The practice kept logging enquiries the same way. Total new enquiries rose to thirty one the following month. Sixteen of them, over half, specifically named ChatGPT or Perplexity as the reason they had found this practice instead of one of the others nearby.
| One month before the changes | One month after the changes |
|---|---|
| Eleven total new enquiries | Thirty one total new enquiries |
| Three mentioned an AI assistant | Sixteen mentioned ChatGPT or Perplexity |
| No named architect profiles, no stated fees, no project outcomes | Named registration profiles, fixed fee bands, and twelve priced project outcomes |
4. Enquiries Into Signed Contracts
Not every enquiry turns into a signed project, and this practice was no exception. Of the sixteen AI referred enquiries, six were signed before the end of the following month. Four of the six asked specifically for the architect named on the project type closest to their own. None of the six asked what a feasibility study cost over the phone, they had already seen the fee band online.
An AI assistant cannot recommend a registration number, a fee band, or a finished project it was never given.
What This Means For Architecture, Engineering, and Consulting Firms
None of the three changes required a new hire or a change in fees. They required stating, in plain text, the facts an AI assistant needs to recommend one practice over a dozen similarly qualified firms in the same region: who actually holds the registration, what a project actually costs, and what the firm has actually built. Firm owners waiting for their next rebrand to add this information are the ones still showing up as three on their own version of this measurement.