Delivery manifest · No. DM-2026
Delivery manifest · No. DM-2026
siTOTis client program · March 30 – July 15, 2026
A running record of what has shipped since the AI Reporting program began — every number below comes straight from the project board and its time log.
Leg 01 · The route
The program started March 30. By March 31 an AI-reporting prototype was answering questions against your live data. By April 4 it had become an interactive product. Everything since has been widening what it can do.
| Month | Hours |
|---|---|
| March (Mar 30–31) | 16.75 |
| April | 160.50 |
| May | 203.90 |
| June | 142.75 |
| July (through Jul 15) | 115.25 |
| Total | 639.15 |
Full database structure exported and parsed; an AI answering reporting questions against your real records.
33 working queries across the schema, a live data connection, and the first interactive reporting interface.
Thirteen dashboard features matched and exceeded — chart auto-selection, exports, drill-downs.
Anyone on the team can save and re-run their own questions — stored in FileMaker itself.
AI-written HTML reports with personal data masked before anything leaves the building; itinerary output in three visual styles, including CJK translation.
Spreadsheet-style pivot reporting in review; the AI onboarding assistant — flagged by your team as the highest-leverage opportunity — in active build.
Leg 02 · Cargo delivered
Eight workstreams, each a working capability the business didn't have in March.
The engine: plain-English questions become correct queries over your database, with the right chart chosen automatically.
Per-user saved questions, stored and loaded through FileMaker — the team's own report library grows as they use it.
Full written reports produced on demand, with client personal data tokenized before any AI sees it.
Client-ready itineraries in three visual styles, including translated CJK editions for overseas travellers.
Your data answered directly in Slack — eleven named queries plus a command-line bridge into FileMaker.
Spreadsheet-style pivot and export options, matching the tooling your team asked for in June. In review.
Onboarding workflow build-out, and an AI assistant that reads client documents and fills the booking system itself.
S3 storage migration and the plumbing that keeps everything above running quietly.
Hours shown are time entries tagged to each feature. A further ~200 hours of cross-cutting build sessions, schema walkthroughs, and working meetings span all of them — the full log totals 639 hours.
Leg 03 · Beyond the manifest
Alongside the tracked program, three substantial pieces were built that never carried billed hours.
A standalone, modern rebuild of the whole system — its own application and database, developed in parallel as a forward path beyond FileMaker.
A local AI engine that lets the reporting features run entirely on siTOTis hardware — no client data sent to outside AI providers.
The project board, time-tracking, and reporting tooling that produced this page — built so the program runs transparently.
Leg 04 · The ledger
The engagement caps each two-week invoice at 80 hours. The work didn't stop at the cap.
| Measure | Hours |
|---|---|
| Invoiced (7 invoices, Apr 13 – Jul 6) | 507.45 |
| Delivered beyond invoiced | 131.70 |
| Total delivered | 639.15 |
Custom natural-language reporting engine, dashboard with saved queries and exports, Slack integration, document-ingestion pipeline — a typical small-shop quote before any AI features existed off the shelf.
The same scope shipped and in production — plus the platform rebuild, private AI service, and program tooling delivered outside the invoices.
Conventional estimate is a conservative market comparison for the delivered scope, not a quote received. Every other figure on this page is drawn directly from the project board's time entries and invoice records.
Leg 05 · Next departures
Thirty items are in active progress or review on the board today. The nearest departures: