Time & Motion Analysis · POC #1
AI Time & Motion: Screen Recording Deep-Dive
34 activities classified. 28 context switches detected. VA/NVA/NNVA breakdown across 8 applications — extracted automatically from a 47-minute screen recording. No manual tagging. No analyst hours.
7.8/ 10
Above-Average EfficiencyHigh Context-Switch Rate34 Activities Classified71% Value-Adding Time
This 47-minute screen recording shows above-average productivity (industry benchmark: 65–70% VA). The primary drag is context-switching — 28 task transitions in 47 minutes (1 every 1 min 41 sec), which is 1.8× the knowledge-worker average. Peak focus block: 14 min 22 sec of uninterrupted design work.
Time Classification · VA / NNVA / NVA
Value-Adding (VA)
71%
33 min 33 sec
Figma — Design Review14 min 22s
Jira — Requirements8 min 15s
Google Docs — Spec Writing6 min 48s
GitHub — Code Review4 min 8s
Necessary Non-Value-Adding (NNVA)
18%
8 min 30 sec
Slack — Coordination3 min 44s
Gmail — Status Updates1 min 28s
Notion — Meeting Notes1 min 24s
File Navigation1 min 54s
Non-Value-Adding / Waste (NVA)
11%
5 min 15 sec
Social / Off-Task Browse3 min 8s
Extended Idle (>2 min)2 min 7s
Productivity Metrics · 12 Measured
Value-Adding Time Ratio
71%
Above industry benchmark (65–70%). Primary VA activities: Figma design review, Jira requirements, spec writing, code review.
Context-Switch Rate
28 switches
28 transitions in 47 min = 1 switch every 1 min 41 sec. Industry average: 15–20 switches per session. Flag: HIGH.
Peak Focus Block
14 min 22s
Longest uninterrupted VA run: 09:14–09:28, Figma design review. Deep work quality: GOOD.
Idle Detection Accuracy
100%
All idle periods (>30 sec with no input) correctly flagged and timestamped. Longest idle: 1 min 47 sec at 09:52.
Application Classification
8 apps / 34 acts
Figma, Jira, GitHub, Slack, Gmail, Google Docs, Notion, Chrome. 34 distinct activity types across all apps.
Communication Overhead
18%
Slack + Gmail consuming 18% of session. Recommended limit for deep-work roles: 10–15%. Actionable reduction available.
Off-Task Browse Time
6.6%
3 min 8 sec of identifiable off-task browsing across 2 separate periods. Primary distraction: non-work content at 09:38 and 10:01.
Task Completion Rhythm
Fragmented
Average uninterrupted task run: 1 min 41 sec. No single task completed in one session except Figma review. Multitasking pattern detected.
Efficiency Score vs Benchmark
+8.5%
71% VA vs 65% industry benchmark for knowledge workers. Score boosted by extended Figma focus block and structured Jira workflow.
Transition Cost Estimate
~37 min/wk
Extrapolating the 28-switch rate across a 6-hour productive workday (5 days) yields ~37 min/week lost to context-switch restart cost.
SOP Step Coverage
89%
8 of 9 defined workflow steps observed and correctly sequenced. Missing: peer-review checklist completion before GitHub PR submission.
Activity Classification Confidence
94%
94% of second-level activity classifications made with high confidence. 6% flagged as ambiguous (tab switching without visible content change).
Application Time Breakdown · 8 Apps Detected
Fi
Figma
14 min 22s
Ji
Jira
8 min 15s
Gd
Google Docs
6 min 48s
Gh
GitHub
4 min 8s
Sl
Slack
3 min 44s
Cr
Chrome (off-task)
3 min 8s
Gm
Gmail
1 min 28s
No
Notion
1 min 24s
Activity Timeline · Key Events
09:02
Session open — Jira sprint board
Opened Jira, reviewed 3 open tickets for sprint context. Classified: VA — Requirements Analysis.
Jira09:07
Context switch #1 — Jira → Slack
Switched to Slack mid-ticket-review. Responded to 2 messages. First of 28 switches in the session.
Slack09:10
Off-task browse — non-work content
Chrome opened to non-work content for 1 min 44 sec. First idle/distraction period.
Chrome09:14
PEAK FOCUS — Figma design review begins
Entered Figma. 14 min 22 sec of continuous, uninterrupted design review — longest focus block of the session. Multiple artboard iterations.
Figma09:28
Focus block ends — switch to Gmail
Figma review ended; context switch to Gmail. Sent 1 status update. Interrupts the longest VA streak.
Gmail09:31
Google Docs — spec writing
Started writing PRD section. 4 min 22 sec before first interruption. Classified: VA — Documentation.
Google Docs09:35
Context switch — Docs → Jira (ticket link)
Left Docs to check a Jira ticket number. Returned 38 sec later. Avoidable with dual-monitor or better note structure.
Jira09:38
Off-task browse — second distraction period
Chrome non-work content, 1 min 24 sec. Second distraction event in the session.
Chrome09:41
GitHub — pull request review
4 min 8 sec of code review. Left inline comments on 3 lines. Classified: VA — Code Review. SOP step: PR review ✓.
GitHub09:47
Slack — extended coordination thread
2 min 11 sec in Slack, 6 message exchange. Classified: NNVA — Coordination. Recommend async over sync for this type.
Slack09:52
Extended idle — 1 min 47 sec
No keyboard or mouse input detected for 1 min 47 sec. Possible bathroom break or away-from-desk. Flagged as NVA — Idle.
09:54
Jira — requirements finalization
Closed 2 tickets, updated acceptance criteria on 1. 4 min 27 sec. Classified: VA — Requirements. SOP step: ticket close-out ✓.
Jira09:58
Notion — meeting notes update
1 min 24 sec updating shared meeting notes from a prior standup. NNVA — required but not directly value-adding.
Notion10:01
Off-task browse — session end
Final 40 sec of session: non-work content in Chrome before recording ended. Classified: NVA.
ChromeAnalysis Output Quality
Activity Classification Confidence
94%
94% of second-level activity types classified with high confidence from visual context alone.
Timeline Precision
Second-level
Every activity transition timestamped to the second. No sampling — full session coverage.
VA/NVA Categorisation
Auto
VA/NNVA/NVA assignment made without any manual configuration. AI infers from activity type and context.
Context-Switch Detection
28 found
All application and task transitions detected. Sub-30-sec micro-switches also captured separately.
SOP Compliance Coverage
89%
8 of 9 workflow steps observed. Missing: peer-review checklist before PR submission.
Report Completeness
100%
Full VA/NVA breakdown, activity timeline, app usage map, bottleneck analysis, and recommendations generated.
AI-Generated Recommendations
CRITICALContext-switch rate (28/47 min) is 1.8× industry average. Block 90-minute deep work windows in calendar with notifications off. Target: reduce to ≤15 switches per session.
HIGHOff-task browsing accounts for 6.6% of session (3 min 8 sec across 3 events). Introduce a distraction blocker during core hours. Recovery cost far exceeds the browse time itself.
HIGHSlack and Gmail consume 18% of session combined. Batch communication to 3 fixed windows (9am / 12pm / 4pm). Async-first policy for threads that don't require real-time response.
HIGHFigma→Gmail context switch at 09:28 broke the session's only 14-minute focus block. Identify what triggered this interruption and set a minimum focus-block rule of 25 minutes (Pomodoro baseline).
MEDIUMFile navigation overhead (1 min 54 sec, 4% of session) suggests non-optimal folder or bookmark structure. Audit frequently-accessed files and pin or template-link them in a quick-access panel.
MEDIUMSOP gap: peer-review checklist not completed before GitHub PR submission (step 9 of 9). Add a pre-submit checklist prompt in GitHub PR template to close the SOP compliance gap from 89% to 100%.
Audio Intelligence · Meeting Synthesis & Decision Memory
Beyond screen activity, VidForgeX captures and transcribes audio from recorded sessions — attributing every statement, decision, and action item to a named speaker. The AI synthesises a context brief after each meeting: who was present, what was decided, who owns what, and what risks were flagged. This brief persists — so future queries can reference it without re-watching the recording.
Live Transcript · Speaker-Attributed
AM
We need to decide on the data pipeline approach before Thursday. I'm leaning toward the event-driven model — fewer moving parts once it's live.
SR
Agreed. Let's go with event-driven. Amir owns the schema definition by Wednesday EOD. I'll handle the consumer side.
JK
One concern — we haven't accounted for schema versioning when consumers are already live. That's a breaking change risk if we move fast.
AM
Good catch. I'll include a versioning strategy in the schema doc. James reviews before Wednesday merge.
Decisions & Action Items Extracted · AI-Attributed
| Decision / Action | Owner | Due | Type |
|---|---|---|---|
| Adopt event-driven pipeline architecture | Sara R. | Immediate | Decision |
| Define event schema with versioning strategy | Amir M. | Wed EOD | Action |
| Review schema doc before merge | James K. | Wed EOD | Action |
| Build consumer-side pipeline integration | Sara R. | Thursday | Action |
| Schema versioning risk — breaking change if consumers live | James K. | — | Risk |
Auto-Generated Context Brief
Sprint planning — 18 Jun 2026 · 09:00–09:47. Three attendees: Amir M. (backend lead), Sara R. (team lead), James K. (infra). Key outcome: team aligned on event-driven architecture for the data pipeline. Amir owns schema definition by Wednesday EOD including a versioning strategy after James flagged breaking-change risk for live consumers. Sara owns consumer-side integration, targeting Thursday delivery. James reviews the schema doc before the Wednesday merge. No unresolved blockers — one risk logged and assigned.Next touchpoint not scheduled; follow-up recommended Thursday morning to confirm Wednesday deliverables landed.
Context Memory · Persists Across Sessions
5 extracted
Every decision and action item stored — queryable by owner, date, or topic in future sessions.
3 speakers
AI attributes every statement to a named participant. Ask "what did Sara commit to?" across 6 months of recordings.
1 flagged
Risks remain open until a follow-up session marks them resolved. Unresolved risks resurface in future context briefs.
Instant
"What did we decide in June about the pipeline?" — answered in seconds, no re-watching. The brief is the memory.
AI Agent Chat · Context-Aware Queries Against This Session
AI Chat — ask_about_session()
Grounded in this session's analysis · context retained across turns
tma-screen-1.mp4 · 47:18
Which app is causing the most productivity drag, and why?
U
ask_about_session(sessionId="tma-screen-1", q="top drag app")
→ 88ms
Chrome (off-task) is the primary drag — not by time alone (3 min 8 sec, 6.6%), but because each browse event interrupts a VA streak and incurs a 10–15 min cognitive restart cost. The 09:10 browse broke an early Jira flow; the 09:38 one interrupted Docs spec writing mid-sentence. Slack is the secondary drag: 8 messages across 2 threads consumed 3 min 44 sec and triggered 6 context switches — each one fragmenting VA runs that averaged only 1 min 41 sec.
If we eliminate the off-task browse entirely, what would the projected efficiency score be?
U
ask_about_session(sessionId="tma-screen-1", q="projected score no browse")
→ 61ms · context from turn 1 reused
Removing the 3 browse events (3 min 8 sec, 3 distinct interruptions) and crediting restart recovery time would raise VA from 71% → estimated 76–78% and reduce context switches from 28 to ~22. Projected score: 8.4–8.6 / 10 — moving from above-average into the top quartile for knowledge workers. The Figma focus block at 09:14 already demonstrates this analyst can sustain deep work; the browsing is a habit pattern, not a structural bottleneck.
Context carried from turn 1 — browse events and switch count not re-fetched.
Can you compare this session's context-switch rate against other sessions I have uploaded?
U
get_person_progression(scope="org", metric="context_switches")
→ cross-session lookup
Upload multiple sessions to unlock cross-session benchmarking — the AI retains every session's analysis and can answer questions like this across your entire recording history. Once you have 3+ sessions, it surfaces trends: improving VA ratio, recurring distraction windows, SOP drift over time, and individual vs team patterns — all from natural-language queries, no dashboard-building required.
Ideal for Operations, Engineering Leaders & Lean Teams
VidForgeX TMA replaces manual stopwatch studies with continuous, automated measurement — from screen recordings, factory floor cameras, and warehouse footage. Get a custom demo with your own recordings.
[email protected]Mohamed Fazaary · RaynX.ai
Frequently Asked Questions
What does VidForgeX TMA detect in screen recordings?
VidForgeX TMA classifies every second of the recording into activity categories (Value-Adding, Necessary Non-Value-Adding, Non-Value-Adding), identifies the active application, detects context switches, measures cycle times for repeated tasks, flags idle periods, and computes a productivity efficiency score. All automatically — no tagging, no manual review.
What is Value-Adding vs Non-Value-Adding time?
Value-Adding (VA) time is spent on work that directly contributes to outputs the customer or stakeholder cares about — in this case design, analysis, and code review. Necessary Non-Value-Adding (NNVA) covers activities required to coordinate work (email, Slack, meetings). Non-Value-Adding (NVA) is pure waste: distraction browsing, extended idle, and avoidable rework.
What is a context switch and why does it matter?
A context switch is a transition from one task or application to another requiring cognitive re-orientation. Each switch incurs a restart cost — research estimates 10–23 minutes to fully re-enter deep work. This session recorded 28 switches in 47 minutes (1 switch every 1 min 41 sec), which is 1.8× the knowledge-worker industry average and a primary productivity drag.
Can VidForgeX TMA analyse any type of screen recording?
Yes. VidForgeX TMA works with any screen recording format (MP4, MOV, MKV, WebM). It adapts to any application mix — developer workflows, design tools, CRM systems, ERP dashboards, spreadsheet-heavy finance work, or remote customer service sessions. The AI infers activity type from visual context, not from pre-configured app lists.
How is the efficiency score calculated?
The score (7.8/10 in this POC) is a composite of: VA time ratio (71%), context-switch frequency penalty, idle duration penalty, and focus-block quality (longest uninterrupted VA run as a fraction of total session). Industry benchmark for knowledge workers is 65–70% VA time; this session is above average but flagged for high switch frequency.