Backing without a spotter, post moves
(31 of 122 vehicle incidents)
ConfidentPresent in 3 critical-severity incidents.
What would help Is a spotter actually available on a post move, and does the crew closest to the work agree that it is?
EMS safety intelligence
Your reports document what happened. They can't tell you what keeps happening.
Incident intelligence for EMS — built by someone who used to write these reports.
Most safety tools were built to document an incident. They were not built to investigate it. chatIR reads the messy incident history you already have and returns the conditions underneath it — what repeats, where the evidence runs thin, and what is worth testing next.
Take the ninety-second audit → see what your form can't capture → decide if a diagnostic's worth it.
One incident narrative · report 1042
as the crew member wrote it
Backed into the bay door. We were running late getting back in service
Four conditions, now countable across all 1,847 records
1
Time pressure
131incidents
present in all five locations
2
Solo task completion
96incidents
appears in all five locations
3
No spotter present
144incidents
the condition that appears most often
4
Practice described as normal
22incidents
called normal by the crew
Illustrative example Fictional incident, narrative, and figures.
Under your current reporting this is one row, typed MVC, closed the same week. The four conditions in it were never fields, so they were never counted — and the twelve other reports describing the same collision of repositioning alone and a handoff never met each other.
Two instruments
One asks better questions. One reads the answers.
The AI Interviewer
Every report becomes a real conversation
A form asks what happened. The interviewer asks what the conditions were — and follows up while the crew member still remembers. Ten to fifteen minutes, not a two-minute checkbox exercise, and never accusatory.
This is what puts contributing conditions on the record in the first place. Nothing downstream can find what was never captured.
Read more →The analysis engine
Reads a whole history at once
It maps conditions across every record you have, finds what repeats and what travels together, and reports the population and the evidence limits alongside each finding.
It runs on the messy history you already have — which is why the diagnostic works before you have changed a single thing about how you report.
Read more →The platform exists to hold these two together: to route a report through review, keep findings tied to their records, and carry what you decide into the next quarter. Both are built on one rule — the model proposes, a person decides. Read the method →
The same export, read twice
Nothing new goes in. Something else comes out.
This is the whole argument for starting with a diagnostic rather than a rollout. One export, twenty-four months, no change to how anyone reports.
What your report says
Incidents, sorted by kind
- Vehicle/Rig
- 604
- Employee injury
- 421
- Patient handling
- 388
- Near miss
- 232
- Other or uncategorised
- 202
Five buckets. Every one answers what kind of incident it was — a question you already knew the answer to.
What chatIR reads from the same rows
Conditions, sorted by what repeats
- A spotter was unavailable
- 144
- Time pressure
- 131
- Backing the vehicle
- 118
- Solo task completion
- 96
- Rails or restraints left down
- 74
68 of them on a post move; never once on a roadway call
present in all five locations, heaviest on post moves
54 of them on a post move
43 of them at a residence, and it appears in all five locations
38 at a residence, never once on a post move
Twenty-one conditions in total. Every one answers what was going on around the person — and that is the question that predicts the next one.
Same export, same day, no new reporting. The left column is what the record was filed as. The right is what it was holding — including the 202 rows your categories had nowhere to put.
Stop filing your incidents. Interview your data.
What comes back
The surfaces a report actually renders.
The interview, and one sheet of the analysis it feeds. Shown with representative data — the confidence marks and critical flags are the same components a delivered report renders.
Conditions are sorted into five dimensions — SPECS: System, Process & Policy, Equipment & Load, Crew, Scene.
The AI InterviewerField intake · no login required
Incident Report
A few follow-up questions
Answer the questions below to help us document what happened. This should only take a few minutes.
AI Interviewer
We're going to ask about the conditions around this incident — what your environment was like, what you had to work with, what was happening with your crew. We're not looking for who made a mistake. We're looking for the full picture, because that's where the patterns are.
Interview began
AI Interviewer
You said you were backing into the bay. Was anyone spotting you, and if not, what was going on that made that the reasonable call at the time?
You
No spotter. My partner was inside finishing the handoff and I was trying to get us back in service before the next call. Honestly that's how we do it most nights.
AI Interviewer
That's useful, and it isn't a criticism — it sounds like repositioning the unit alone is just what happens while your partner is inside. At that point in a call, is a second person usually free to spot, or is moving it solo the only realistic option?
Type your answer…
The analysis · PatternsDiagnostic report · sheet 1
Vehicle operations · 122 incidents scored
Backing events cluster where a spotter was unavailable
Mirror and clearance contact, night shift
(26 of 122 vehicle incidents)
Established
What would help Does anything about the route, the lighting or the hour make this predictable rather than unlucky?
Stretcher lock failure reported after the fact
(9 of 122 vehicle incidents)
Directional
What would help Thin recurrence — worth watching rather than acting on. Would a pre-shift check surface it earlier?
Ways to begin
Start with a flashlight. Not a contract.
Four steps, each complete on its own. Most organizations start at the first one and decide from what they see.
Step one
Instrument audit
Free
Step two
Operational Incident Diagnostic
$4,500 · Starting price
Step three
The Cycle
90 days · scoped
Step four
The platform
Scoped to the organization
Scope is confirmed before money or data changes hands. chatIR does not currently accept PHI — deidentified operational data only, and the current boundary is written down.
Guides
The thinking, written down.
Investigation
How to run an EMS incident investigation
A safety officer who has just been handed the job and has no procedure to follow.
Method
Contributing conditions vs. root cause
Anyone who has been told to find the root cause and suspects it is the wrong instruction.
Data
What your incident data is not telling you
A manager with a year of clean-looking reports and a nagging feeling that they are not saying much.
Bring the record. Keep the judgment.
Send one deidentified export. You get back what your reporting has been holding all along — and a conversation about what to do with it.