Standard — every customer
Pass/fail on the criteria you select per scenario, completion status, full attempt history, CSV export.
How we score
We score the decision, not the shot.
AGOGE VR records what your officer saw, said and did — second by second — and turns it into a report a trainer signs. Pass/fail on the criteria you choose. The full picture when you want it.
Built for four different readers
| Reader | Their question | The page’s answer |
|---|---|---|
| Training officer / FTO | “Does this help me coach, or just grade?” | The timeline + labeled AI notes + your sign-off. Coaching mode by default. |
| Ops / owner (security co.) | “Will my client accept this as training?” | Pass/fail on criteria you choose, per role. Exportable. |
| Risk / counsel / insurer | “Can I prove this guard trained to a standard on this date?” | A timestamped, human-signed record. “Prove it,” not “did it.” |
| LE command | “Is this the thing our union is fighting?” | No. AI interprets; your trainer decides. Nothing about an officer is auto-scored into a file. |
Three axes. Never one number.
Every session is read against three questions, not one blended score that hides which one someone missed.
Decision
Was the force option right for what was known at that moment — not in hindsight.
“Held fire while the hand was concealed — force matched what was known at 00:13.”
Communication
Did they identify themselves, give clear direction, and try words before tools.
“First command given at 00:02, no self-identification before it.”
Tactics
Distance, cover, positioning, and whether they called it in.
“Cover reached at 00:07; distance never closed inside the reactionary gap.”
What a session captures
A trainer can’t watch six things at once. The system records all of them, in order.
What they saw
the cue, the moment the object was confirmed
What they said
every command and radio call, timestamped
What they did
position, movement, tool selection
When
every one of the above, in order, down to the tenth of a second
How the report is built
Record
The headset logs every event with a timestamp. Nothing is scored on the headset.
Measure
Intervals, distances, counts. Plain math, repeatable, no opinion.
Check
Your criteria, your thresholds, your role profile → pass or fail.
Interpret
The AI reads the evidence and writes what it sees, labeled as the AI's reading, citing the rule it applied and the timestamp.
Decide
Your trainer confirms, adjusts, or overrides — with a reason — and signs. That signature is the record. Not the AI.
Standard vs Pro Reporting
Pro Reporting
The evidence timeline, the AI reading, the trainer-signed record, reaction intervals against the scenario’s decision window, trends over time, roster view.
Sold through Defencify — ask them for it.
Built for the job you actually have
A security officer who keeps distance, gives direction, and reports it did the job. A patrol officer with a duty to act is held to a different line. The same run can pass under one profile and fail under the other — and the report tells you which yardstick it used.
What we deliberately don’t do
- No single blended score — it hides the officer who talks well and shoots wrong.
- No “faster is better” — early is a flag, same as late.
- No automated verdict on a person — the AI interprets; a human decides.
- No heart-rate zones or wrist biometrics — they don't hold up, and some misread by skin tone.
See it
One real scenario run, fully interactive. Every number below is fabricated for demonstration.
SAMPLE DATA — PREVIEW. Not a real trainee, not real data.
After-Action Report · AGOGE VR
Federal Lobby (SEC-002) — M. Webb
Role profileSecurity — unarmed
Attempt1 of 1 · first exposure
Run2026-08-24 · 14:02 · 0:24 duration
LoadoutRadio · OC · no firearm
Mode: CoachingUnreviewed AI draftengine 1.0 · rules SEC-v0 · prompt p3
Result Standard · every customer
Pass/fail is computed from the criteria your admin included for this scenario and role profile. Every line points at the moment in the run that decided it. No AI is involved in this block.
PASS
5 of 6 included criteria met · rule: allow 1 miss| Criterion | Result | Evidence |
|---|---|---|
| Engagement was appropriate (contact inside the protected post; disengage remained available) | Met | 00:00–00:07 |
| Identified self as security before first command | Not met | 00:01.8 — first utterance was a command; no "security" in transcript |
| Verbal direction given before any tool use | Met | 00:01.8, 00:12.1 |
| Reactionary gap ≥ 10 ft while unengaged | Met | min 11.4 ft at 00:06.9 |
| No draw / OC before a weapon was confirmed | Met | 00:13.2 — no tool drawn at any point |
| Reported to dispatch within 120 s of resolution | Met | 00:21.3 (8.1 s) |
| Shots on target ≥ 60% | Recorded | Excluded by your admin — no firearm in this loadout. Still recorded. |
Evidence timeline Pro
What the trainee saw, what the subject did, what the trainee said and did — in order, with the derived intervals drawn against the scenario's expectation window. Early and late are both flags; nothing here rewards speed.
Time to first verbal
1.8 s
Cue → cover reached
6.9 s
Time to decision
4.1 s
Resolution → radio report
8.1 s
Expectation window for this profileTrainee's interval — within windowEarly / late
00:00.0SceneSubject enters lobby; right hand concealed in jacket pocket. (threat cue: concealed hand)
00:01.8Trainee said"Sir, I need to see your hands." · ASR 0.94
00:05.0SubjectSubject stops, hand stays in pocket. "I've got ID in here, relax."
00:06.9Trainee didSidesteps to the pillar (cover); distance 11.4 ft, does not close.
00:01.8AIInterpretation: first utterance is a command with no self-identification (SEC-COMM-01). pending trainer confirmation.
00:06.9CriterionReactionary gap maintained (min 11.4 ft ≥ 10 ft).
00:11.0SubjectSubject begins withdrawing hand slowly.
00:12.1Trainee said"Slowly — let me see it." · ASR 0.91
00:13.2SceneObject visually confirmed: wallet, not a weapon. (threat state → none)
00:13.2CriterionNo draw / no OC before a weapon was confirmed.
00:14.0Trainee didHolds position; directs subject to place wallet on the counter.
00:16.5SubjectSubject complies, places wallet on counter.
00:21.3Trainee didRadio: "Post 2 to base, contact in lobby, subject compliant, no weapon."
00:21.3CriterionReported to dispatch within 120 s of resolution.
00:24.0SceneScenario ends: verbal success outcome.
Three axes Pro · AI interpretation + trainer record
Never one blended number. Under each axis the AI's reading of the evidence is labeled as exactly that — an interpretation, citing the rule it applied and the timestamps it used — until a trainer confirms, adjusts (one band, with a reason), or adds to it.
Decision
ProficientAI interpretation — pending trainer confirmationnote · confidence high
Force option stayed at "none" while the hand was concealed (00:00–00:13.2) and the weapon was never confirmed. The choice matched the threat state at every timestamp.
Rule applied:
SEC-DEC-02 No force option before a threat is confirmed; disengage-and-report is a correct outcome (CA BSIS Power to Arrest).AI interpretation — pending trainer confirmationnote · confidence high
Decision latency 4.1 s sat inside this profile's window (3.0–6.0 s). No anticipation flag — nothing was drawn, so the 0.38 s floor does not apply.
Rule applied:
ENG-01 Draw-to-shot faster than ~0.38 s implies anticipation (Blair 2011).Communication
EmergingAI interpretation — pending trainer confirmationdeficit · confidence high
The first words at 00:01.8 were a command. "Security" does not appear anywhere in the transcript. A subject cannot tell who is giving the order.
Rule applied:
SEC-COMM-01 Identify as security before directing a member of the public; never imply peace-officer status.AI interpretation — pending trainer confirmationdeficit · confidence med
Both commands were direct and specific (“I need to see your hands”, “Slowly — let me see it”): 2 of 2 in direct form. No reason for contact was stated at any point.
Rule applied:
SEC-COMM-03 State the reason for contact when time allows; direct commands over vague ones.AI interpretation — pending trainer confirmationnote · confidence high
No profanity, no informal address. Volume +6 dB against this trainee's own baseline at 00:12.1 — reported as a within-subject change, not a stress claim.
Rule applied:
COMM-PRO-01 Profanity flag (perceived-excessive-force finding); volume delta is descriptive only.De-escalation rubric · 5 of 14 items (DePICT-style, 0–3)
Approach & containment
Time & distance management
Calming paralanguage
Transparent decision-making
Active listening
Tactics / Procedure
ProficientAI interpretation — pending trainer confirmationnote · confidence high
Cover reached 6.9 s after the cue; minimum distance 11.4 ft, never closed. Head orientation stayed on the subject for 92% of the active-threat window (orientation, not gaze — no eye tracking on this headset).
Rule applied:
SEC-TAC-02 Create distance and use cover before contact; hold position while a hand is concealed.AI interpretation — pending trainer confirmationnote · confidence high
Radio report 8.1 s after resolution, well inside the 120 s post-incident timer, with subject status and "no weapon" stated.
Rule applied:
SEC-TAC-05 Report to dispatch within the post-incident window; include subject status.Trainer record Pro
Not yet reviewed. This report is an AI draft. It becomes the training record when a trainer signs it. Exports, PDFs and the LMS receive the signed version only.
Trend · Decision axis
M. Webb across 4 attempts of Federal Lobby (SEC-002). First-exposure runs are the honest number; repeats are partly recall.
61
69
90
88
Flagged: a repeat attempt of the same scenario appears in this trend.
Roster comparison
| Trainee | Percentile | Flag |
|---|---|---|
| M. Webb | 82th | Standout |
| D. Price | 71th | — |
| E. Vasquez | 64th | — |
| A. Thompson | 58th | — |
| R. Chandler | 51th | — |
| K. Nguyen | 44th | — |
| J. Alvarez | 37th | — |
| S. Okafor | 29th | — |
| T. Brandt | 12th | Needs attention |
Every pass/fail is computed from the session’s event stream and is recomputable. Items marked AI interpretation are model output pending trainer confirmation.