How it works
Six steps. One loop. Nothing left to guesswork.
Step through one report from a payment terminal: what happens on the device, what happens in Field, and who gets to decide at each point.
- Screenshot
- Logs · last 30 min
- Device state
- Card numbers → [PAN]
Step through it
One report, start to finish.
Use the tabs or the arrow keys. Each step shows the device, Field and the decision maker side by side.
- On the device
- The cashier taps the floating Report a problem button, picks a category if they want (Card reader, Printer, Network, App frozen, Other) and carries on serving.
- In Field
- Nothing yet: the device was dormant. Its only traffic so far was a small signed check-in.
- Who decides
- The person at the counter, or an authorised person who triggers a remote capture.
- Screenshot
- Logs · last 30 min
- Device state
- Crash & error events
- Note: “Card reader”
1 tap · no form · works offline, uploads when it can
The evidence
What leaves the device, and what never does.
A report is small, redacted on the device and sent only when someone reports or an authorised person asks.
Sent with a report
- Screenshots of what the person saw (payment screens are blanked)
- The recent logs from the apps you choose
- The device's state: battery, storage, network, printer and reader status, app versions
- Crashes and errors around the report
- The one-tap category and an optional note
Never sent
- Card numbers, track data and PINs
- Secrets and passwords
- Anything at all, until someone reports or an authorised person asks
Your policy decides
Every proposed fix gets one of four answers.
AI proposes. Your policy decides, from how disruptive the fix is, how sure the diagnosis is and what the device is doing.
Runs on its own
Low-impact fixes your policy allows, on a device that is idle.
Waits for people
One approver for disruptive fixes, two different people for destructive ones.
Goes to an engineer
Not sure enough, or the evidence looks suspicious. A person reads the report first.
Never runs
Your policy forbids it, the device is protected, or the fleet has had enough changes for now.
How much runs on its own, set per customer
OffSuggest onlyApproval for everythingAutomatic for low riskAutomatic
Start with suggestions only. Move up when the record shows you the fixes would have been right.
The fixes
Named fixes. Never a shell.
Field runs only fixes from its signed library: restart an app, re-sync configuration, reset the network, roll back an update, reboot. The more disruptive a fix, the more people have to agree. See what Field can fix.
Never during a payment
A fix that changes something waits until no payment is in progress. If Field cannot tell, it waits and tells you it deferred.
Verify, or roll back
Every fix checks that it worked. If it did not, it is undone, the ticket says so and a person takes over.
Staged across a fleet
A fix for many devices goes to one first, then a few, then the rest, and stops on its own if results start failing.
And then it learns Rolling out
Your engineers teach it. The next fix is faster.
After a fix, an engineer confirms the AI’s root cause or corrects it. Confirmed causes become known issues, so the next report with the same fault skips the guesswork. The policy and the approvals above still decide what runs.
- 01AI proposes a causeRanked, with the log lines behind it.
- 02An engineer confirms or corrects itOne click, or the right answer.
- 03It becomes a known issueWith the fix that worked.
- 04The next one is fixed first timeMatched on arrival; your approval rules still apply.
Across the fleet, the same knowledge powers clusters, regression watch after a rollout and a monthly savings report. See fleet intelligence
See the loop on your own terminal.
See TickFlow Field on a real PAX terminal in 30 minutes: a report, a redacted bundle, a diagnosis and a signed fix, end to end.
Or write to hello@tickflow.co