PTS: inspection, zero to production
Hired to design, build and maintain a system that turned field inspection from Excel spreadsheets into a reliable digital product.
What I found
The inspection operation ran on Excel: loose photos, measurements and defects typed by hand, outdated spreadsheet versions and reports assembled manually. The result was rework, error risk and no consolidated view of the operation.
I started by observing a real inspection in the field — without interfering — to map the actual workflow, exceptions included. From there, I separated business rules, states, data and automation points, and designed a system simple enough for the inspector to use in the field and robust enough to scale.
What I built
- Offline-first mobile app for field capture: photos, measurements, defects and evidence — working without internet, syncing automatically afterward.
- Web admin dashboard with real-time operational analytics, work queue, review workflow (creation, return, cancellation, no-show) and report generation.
- Backend and data-reliability rules — phase-based ownership to prevent conflict between field and back office, and consistent sync.
- AI and AI-assisted development to speed up documentation, QA, testing and the automation roadmap (checklists, standardized observations, defect classification).
Technical sheet: capability implemented
- Field capture
- Photos, measurements, defects and evidence recorded in the app without depending on a connection on site.
- Synchronization
- Automatic upload when the connection returns, with consistency guaranteed between field and back office.
- Ownership by phase
- Each stage has an owner in the system, which prevents field and back office from editing the same report at the same time.
- Review flow
- Creation, return, cancellation and no-show are states of the report — not arrangements made over chat.
- Work queue
- What is pending, with whom and for how long, visible in the admin dashboard.
- Operational analytics
- Quality and productivity consolidated in real time, with no spreadsheet in between.
- Reports
- Generated from what was captured: the report structure arrives already filled in by the field record.
- AI in development
- Applied to documentation, QA and testing, and to the automation roadmap — checklists, standardized observations and defect classification.
Role: Product Lead & Solution Architect, on contract, since Sep/2025. The system is in production and continuous operation — uptime, support, telemetry and preventive maintenance — remains my responsibility. No line in this sheet is an estimate: gains, percentages and projections stay in the diagnostic, under the estimated-impact label.
How the system works, in the order it runs
Field — the inspection is recorded where it happens
View full sizeField inspection with the mobile app. An illustration generated from the app's real layout, to protect the client. App — the work queue runs without internet
View full sizeMobile app — inspection list, offline-first. Back office — the same record reaches the dashboard consolidated
View full sizeExecutive dashboard — consolidated quality in real time.
Screens with demo data; client identity preserved.
What changed in the operation
Capability implemented What used to be a manual Excel process became a system in production, with standardized field capture, traceability and real-time managerial visibility. I remain responsible for ongoing operation — uptime, support, telemetry and preventive maintenance.
Estimated impact Estimated productivity gains and rework reduction were modeled in the proposal and can be shared on request.
What this costs your operation
Let's find this cost together, one step at a time. Each step covers one part of the operation and shows what it costs before opening the next one. Steps 1 to 5 use no number you haven't entered yourself; every assumption sits in step 6, visible and editable.
Fields marked * are required to move on to the next step.
The work you pay for
Start with what you can see: how many inspections your operation does a month, how long each one takes and what an hour of the field inspector's time costs.
The time the inspector spends at the client, from the start of the visit to the end.Visible costa month in the fielda year, in your currencyThis is the work that creates value — it's what you pay for and what's in the budget. Keep this number in mind. Everything that comes next adds to it, outside that budget.
Fill in the three fields to continue.
What happens after the visit ends
The inspector leaves the client with a full notebook. How much more time does it take to turn that into a report? Typing, attaching photos, checking, closing.
Hidden cost · 1 of 3a montha year, in your currencyComplete the previous step.
What the back office checks
Before it reaches the client, every report goes through someone. How long does that check take — even when the report arrives right?
If you leave it blank, I use the inspector's hourly cost.Hidden cost · 2 of 3a montha year, in your currencyComplete the previous step.
What comes back wrong
A missing photo, a measurement, a signature. How often does it happen, and how much time does it take to fix?
Asking for the missing photo, waiting for the reply, adjusting or redoing.Optional
It doesn't go into any calculation — it shows up in step 6 as your turnaround baseline.Hidden cost · 3 of 3a montha year, in your currencyComplete the previous step.
The total nobody tracks
In your operation these three costs live apart, which is why nobody has ever added them up. Together, they come to this. No assumption goes in here — only what you entered.
—hours a month outside the inspectionconsolidation, review and rework—a year in your currency, for that time aloneat the hourly cost you entered—inspections would fit in the inspector's hoursa straight conversion of hours into visitsComplete the previous steps.
What changes with the system
Up to here it was diagnosis. From here on it's simulation: these are assumptions, not measurements. You choose how much you believe each of the three stages can be cut, and the result changes with it.
How much can each stage be cut?
The three percentages below are my assumptions. The real measurement depends on the baseline being built right now — once it's complete, these numbers stop being assumptions. Drag each bar to whatever makes sense for your operation: the further right, the bigger the cut.
10%90%70%I set gain 1 lower because it's the most gradual: it grows as the team works the app into the rhythm of the visit.
The three labels below separate what exists from what is arithmetic: Capability implemented is what the system already does today, with no assumption behind it. Estimated impact is a number that depends on the assumptions above. Result being measured is what I'm still building, so it has no number.
Gain 1 The inspection gets more efficient
The form is ready on the phone, a photo goes in the moment it's taken, a measurement is typed only once.
Estimated impact—Gain 2 The report leaves the site almost finished
What the inspector fills in on site already is the report. This is where the cut is biggest: instead of rebuilding everything after the visit, it only needs to be closed and sent.
Capability implemented—Gain 3 Review gets shorter and fewer reports come back
The app won't let a visit close with a blank field or a missing photo, so fewer incomplete reports reach the back office — and what does arrive is simpler to check.
Capability implementedI'm adding AI to this stage to speed up the check. It's work in progress — so this gain is not added to any number on this screen.
—The work ends when the visit ends
No evenings of typing, no Saturdays on spreadsheets. I leave it without a number on purpose: it's an effect on people, not on the bill. Even so, it's the first thing inspectors mention.
Capability implementedends with the visitReport turnaround becomes a tracked number
Capability implementedstarts being measuredSeeing the defect that keeps coming back
With the history organized you can see patterns by batch, shift and supplier. I'm still building this indicator.
Result being measuredno numberIf these hours become savings
a year, in your currencyFill in the previous steps.
If they become more inspections
more a month, with the same field teamFill in the previous steps.
The two boxes read the same time saved in two different ways and do not add up. Savings count consolidation and review; capacity counts the time saved in the inspection and in consolidation — only the inspector's time turns into visits. Consolidation hours show up in both because they can become one thing or the other, not both.
Complete the previous steps.
Nothing you type leaves your browser. There's no server, no sign-up and no storage — close the tab and the numbers are discarded. If you'd like to talk about them, the button below opens an email with your values already written in. It's only sent if you click send.
What this diagnosis shows
The hidden cost of the time spent on consolidation, review and rework, split between inspector and back office and calculated only from the data you entered.
What this simulation estimates
How much of that time can be cut in each of the three stages, under editable percentages — not a promise of results.
What still needs to be measured
The cut actually achieved after rollout, taking into account adoption, how the operation is set up and how exceptions are handled.
What is not being claimed
That the savings are validated or the return guaranteed. Without a baseline and measurement after rollout, everything here is an estimate.
Does your operation need a system like this?
Also open to AI Product Lead roles, in Brazil and abroad.