The Starting Menu
The seed call output a seven-item menu of candidate AI use cases plus a built-in pick — voice site diaries — and the rest of this strategy exists to test whether that pick is right.
The call summary described a business where “large amounts of project information exist, but much of it is fragmented across documents, drawings, Procore, emails, Excel files, photos, site records, and people’s personal knowledge.” From that, it distilled seven near-term use cases:
| # | Use case | What it does | Money signal |
|---|---|---|---|
| 1 | Voice/WhatsApp site diaries | 4pm prompt → voice note → transcribed daily diary (labour, materials, plant, delays, events) | Only ~30% “do them properly” today |
| 2 | Tender document ingestion | Reads tender pack, extracts deliverables, flags missing info, generates RFIs and estimating actions | Faster, more complete estimating |
| 3 | AI cost planning from history | Queryable “data room” of past jobs — “What did we pay for this flooring before?” | Months → minutes |
| 4 | Programme/status dashboard | Programme linked to real progress; live view across all active projects | Directors “lose the ability to stay close” as the company grows |
| 5 | Live O&M manual | Builds the handover file continuously from day one, not as end-of-job paperwork | Removes painful late-stage compilation |
| 6 | Assisted drawing takeoff | Reads drawings, measures quantities, produces BoQ; possible Bluebeam integration | Replaces manual measurement labour |
| 7 | Materials/asset tracking | QR identity per component, design → procurement → delivery → install | Surfaces risks like furniture arriving after a hotel is finished |
On the count, because the strategy depends on it. Seven items, framed as use cases — candidate places AI might help, not a diagnosis. Later summaries compress this to six or call them “problems”; both are wrong. The count stays seven because the whole argument is over which can lead to defensible money.
The seed’s own pick was use case 1. Per the summary: “the clearest first prototype is likely automated site diaries from structured voice prompts… without requiring full BIM integration or expensive site-capture technology.”
Open question. Is the menu’s own favourite — the voice diary — the right first bet, or a trap dressed as a quick win? That question drives the pivot.