×
WHITEPAPER
Full-Scale Agile Intelligence Deployment
77 production teams · 701 active contributors · Full deployment live since end of 2025
Created by Umut ARISOY, oNabu · Reviewed & Edited by Bülent TURHAN, Altan EVNI
+112%
Customer Focus
17.5 → 37.2 (first → last week)
+209%
Retro Improvements
16.2 → 50.0 (first → last week)
10,821
AI Jira Actions
7,358 comments + 3,463 status changes
9,893
Meeting Reports
86.5% full processing rate
Executive Summary
The Problem
Team ceremonies generate rich behavioral and operational data — but traditional reporting captures only the 4% visible layer: sprint velocity, release dates, closed tickets. The other 96% — cross-team blockers, planning debt, silence in ceremonies, erosion in psychological safety — remains trapped inside 300+ hours of weekly meetings that no executive can attend.
The Solution
oNabu deploys as a Next-Generation Management Layer — an AI agent that sits inside every ceremony and continuously: (a) transcribes, classifies, and scores team behavior across 27 dimensions; (b) detects canonical dysfunction patterns with severity thresholds; (c) writes findings back into Jira as comments and status changes; (d) issues evidence-based coaching tactics. The result is a real-time organizational nervous system that makes the invisible 96% visible and actionable.
Key Results (110-day reporting window)
- Five Space+ metrics showed measurable, defensible movement
- 9 in 10 observed ceremonies were fully processed
- 3,294 cross-team blockers surfaced
- 23 teams had no captured retrospective; ~42 had no captured sprint review
- These gaps are now quantified, named, and addressable — which is itself the product
Context
LC Waikiki — Turkey's largest apparel retailer, with 55,000+ employees, 1,287 stores in 58 countries.
1997
First store in Türkiye
2009
First international store (Romania)
2014
100th store abroad
2017
Sub-Saharan Africa & SE Asia
2020
1,000 stores worldwide
2024
15th global year
2025
58 countries, 1,287 stores
LC Waikiki has pursued a formal digital transformation program since 2019. The organization has adopted agile working principles across product, supply chain, e-commerce, and store technology domains. The deployment reported here covers the full 77-team software delivery organization — 701 active engineers, scrum masters, and leadership — running on Jira (Data Center 8) and Microsoft Teams. 76 of 77 teams operate primarily in Turkish.
From pilot to full deployment: The relationship began as a narrow pilot in October 2025 with 2 teams, expanded to 7 teams in late November, and moved to full-organisation deployment across all 77 production teams by the end of 2025. This whitepaper describes the first 110 days of full-scale operation after the organisation-wide rollout.
| Parameter | Value |
|---|---|
| Initial pilot cohort | 08 October 2025 – 29 December 2025 (2 → 7 teams) |
| Full-scale deployment live | By end of December 2025 (all 77 production teams) |
| Reporting window analysed | 29 December 2025 – 17 April 2026 (110 days) |
| Baseline window | 29 December – 12 January 2026 (first 14 days) |
| Late window | 3 April – 17 April 2026 (final 14 days) |
| Production teams | 77 (Onboarding team excluded) |
| Active contributors | 701 (574 team members + 69 scrum masters + 58 admins) |
| Integrations in use | Jira DC 8 & Microsoft Teams |
| Observation volume | 11,039 meetings captured · 9,554 fully processed (86.5%) |
Goals
Transparency
Make the invisible 96% of team behavior visible without manual reporting overhead.
AI Coaching
Observe teams as a virtual coach, utilizing AI to provide mentoring and support.
Early Warning
Detect cross-team dependencies, planning debt, and ceremony skipping before sprint failures.
Measurement
Generate quantifiable behavioral metrics that leadership and teams can act on each sprint.
Automation
Automate Jira hygiene so engineering time shifts from documentation to delivery.
Problem
At organizational scale, agility breaks down because leadership cannot see inside the ceremonies.
Visibility Gaps
4,343 scrum ceremonies were observed — no single stakeholder could attend more than a fraction. Critical decisions surfaced in speech, not in Jira or status reports.
Dependency Propagation
Cross-team dependencies — the #1 cause of delivery slippage — were discussed once in a meeting and then lost. No structured mechanism existed to aggregate them.
Decision & Execution Delays
Action items from retrospectives lived in personal notes. The feedback loop between signal and action measured in sprints, not days.
Unmeasured Team Dynamics
Psychological safety, balanced contribution, self-management, and motivation were entirely unquantifiable and invisible to leadership.
Baseline Dysfunctions (first 14 days)
| Signal | Value | Interpretation |
|---|---|---|
| Dysfunction alarms fired | 31 → 152/week | Detection ramp as teams activated |
| Speaking no-show rate | 22–30% | 1 in 4 ceremony slots silent at start |
| Retrospective reports | 2 reports/week (77 teams) | Not yet captured or institutionalized |
| Sprint review reports | 2 reports/week (77 teams) | Not yet captured or institutionalized |
| AI Jira comments | 2 → 26/week | Baseline adoption curve confirmed |
| Dependency surfacing | 14 dependencies/week | Hidden blocker rate invisible pre-deployment |
Approach
Two hypotheses defined before the full-scale rollout, with measurable success criteria.
H1 — THE VISIBILITY HYPOTHESIS
Deploying oNabu across all 77 teams for 110 days will surface organizational signals that are currently invisible — blockers, ceremony gaps, behavioral patterns, and dysfunction trends — at a volume and granularity that traditional reporting cannot match.
H2 — THE AUTOMATION HYPOTHESIS
oNabu's AI-generated write-backs to Jira will exceed 5,000 actions in the reporting window at a minimum 80% confidence, proving that high-volume hygiene work can be automated without introducing low-quality noise.
Success Criteria
| # | Criterion | Target | Result | |
|---|---|---|---|---|
| 1 | Meeting coverage | ≥ 75% processed | 86.5% | ✓ |
| 2 | Dysfunction detection | ≥ 5,000 alarms | 31,444 | ✓ |
| 3 | AI Jira actions | ≥ 5,000 high-confidence | 10,821 | ✓ |
| 4 | Behavioral metric movement | ≥ 1 metric > 10% | +112% | ✓ |
| 5 | Delivery metric movement | ≥ 1 signal improves | Heartbeat +6.5% | ✓ |
| 6 | Hidden blocker surfacing | ≥ 500 dependencies surfaced | 3,294 | ✓ |
| 7 | Team coverage | ≥ 70/77 teams active | 75/77 | ✓ |
| 8 | Ceremony visibility | Make presence/absence measurable | 23 zero-retro & ~42 zero-review surfaced | ✓ |
What we agreed NOT to claim
- Sprint velocity: sprint_metrics table was empty; could not baseline.
- Dependency resolution rate: oNabu surfaces blockers; resolution is owned by teams.
- Dollar ROI: FTE-cost field was not populated. Financial modelling is directional.
Implementation
All 77 production delivery teams enrolled. 701 active users across three role types.
574
Team Members
Individual contributors
69
Scrum Masters
Team leads & admins
58
Company Admins
Leadership, RTE, PMs
Integrations
| System | Use | Bidirectional? |
|---|---|---|
| Jira Data Center 8 | Work item context, write-back target | Read / Write (with Human in the Loop) |
| Microsoft Teams | Ceremony transcription, speaker diarization, chat & channels | Read |
Platform capability note
Timeline
Initial Pilot
Oct 8 – Nov 28, 2025
2 teams onboarded; platform tuning; integration shakeout
Pilot Expansion
Nov 29 – Dec 14, 2025
Expanded to 7 teams; Scrum Master enablement iteration
Full Rollout
Dec 15 – 28, 2025
All 77 teams onboarded
Phase 1 — Ramp
Dec 29 – Jan 26
First 31 alarms → 1,387/week; 2 reports → 234/week
Phase 2 — Signal
Jan 27 – Mar 1
Detection peaks; 2,400+ alarms/week; backlog spike identified
Phase 3 — Outcome
Mar 2 – Apr 13
Customer Focus +112%; Retro Improvements +209%; stabilising
Review & Readout
Apr 14 – 17
Executive readout, metric review, renewal scoping
How It Works
oNabu operates as a three-layer pipeline: Continuous Capture → Discovery & Scoring → Action & Write-Back.
STEP 01
Capture
- –Meeting audio & participation (Teams)
- –Chat & collaboration context
- –Jira issues, transitions, sprint state
- –Calendar & scheduling metadata
STEP 02
Discover & Score
- –27 possible dimensions (14 active)
- –Meeting summaries & daily reports
- –Speaking analysis & sentiment
- –Dependency detection & dysfunction classification
STEP 03
Write Back
- –Jira comments (context-rich)
- –Status changes (automated)
- –Coaching tactics (individual & team)
- –Escalation alarms (leadership)
Active Dimensions
| Dimension | What is produced | Volume |
|---|---|---|
| Meeting summaries | Risks, decisions, action items, open questions | 4,628 reports |
| Daily reports | Attendance, speaking share, engagement, blockers | 3,737 reports |
| Speaking analysis | Per-speaker share, engagement, no-show, sentiment | 48,961 events |
| Dependency detection | Cross-team / external blockers with type & severity | 3,294 events |
| Dysfunction classification | 14 of 27 codes fired, severity-scored | ~75,000 events |
| Psychological safety | Per-team safety trend with weekly deltas | 4,655 events |
The 80/20 Principle — applied to Jira
Key Insights
Findings from live deployment data, grouped by the kind of visibility oNabu created.
7.1 — Hidden blockers became visible at scale
oNabu surfaced 3,294 cross-team and external dependency events, peaking at 359 in a single week. 130 were classified high-severity. Dominant categories: review dependencies (1,001), spec dependencies (572), and access/integration blockers (357).
7.2 — Ceremony gaps were larger than anyone knew
Missing Sprint Retrospectives
Missing Sprint Reviews
Whether skipped or held offline, these ceremonies were invisible to leadership. oNabu makes the gaps addressable. That is the product.
7.3 — Build-to-spec → customer-pull shift
+112%
Customer Focus Score
Team Customer Focus (Space+ code 204) rose from 17.5 to 37.2 over five measured weeks. Monthly averages: February 21.30 → March 36.54 (+71.5%). Teams began opening sprint reviews with customer-facing outcomes instead of ticket-close statistics.
7.4 — Voice equity started to form
Team Speaking Share moved from near-zero to 16%. Dominant speakers (>50% of speaking time) grew from 0 to 54 by week of April 6 — the platform began identifying voice-concentration patterns that retrospectives alone cannot surface at 77-team scale.
7.5 — First-cohort fatigue surfaced as an early warning
Psychological Safety declined — and that is actionable
7.6 — Planning debt spike was caught in February
Week of February 23: 8,012 Product Backlog PBI Readiness events detected — normal weeks were 666–2,020. The spike revealed systemic planning debt: items entering sprints without acceptance criteria, estimates, or dependency mapping.
7.7 — Learning culture formed where ceremonies were held
+209%
Retro Improvements
Team Retro Improvements (code 221) moved from 16.2 to 50.0. Monthly avg: Feb 20.42 → Mar 32.26 (+58%). Participating teams progressed from ritual compliance toward genuine retrospective practice.
7.8 — The observability paradox
The most defensible story is not the metrics that went up. As coverage expanded, Psychological Safety and Discipline moved negatively. That is not a failure — it is what happens when an organisation stops measuring the easy 4% and starts measuring the hard 96%. The decline is specific, team-named, and early enough to act on.
Impact
Measured across four dimensions: operational, delivery, behavioral, and financial.
Operational Impact
86.5%
Ceremony Coverage
9,554 / 11,039 meetings processed
3,294
Blockers Surfaced
130 high-severity dependencies
6,454
Coaching Tactics
668 tactic sessions, individual + team
Delivery Impact
−2.4%
Lead Time
2.91 → 2.84 (improved)
+6.5%
Heartbeat
67.2 → 71.6 ceremony stability
2.35→5.20
Velocity Trend
Upward but not claimed (incomplete baseline)
Behavioral Impact
+112%
Customer Focus
17.5 → 37.2
+209%
Retro Improvements
16.2 → 50.0
Stable
Motivation & Engagement
67.5 → 66.6 (no burnout)
Enabled
Voice Equity
Speaking share + dominant speaker ID
Financial Impact (Directional)
Directional — not invoiced
- Jira hygiene automated: 10,821 AI actions × ~3 min each ≈ 540 engineering hours saved.
- Directional dollar range: ~US$10K–20K in recovered engineering capacity (110 days).
- Efficiency signal: Space+ FTE module identified ~5% efficiency loss across 77 teams.
- Prevented sprint failure: 3,294 surfaced blockers represent avoided sprint failures. A single failure costs 10–15 engineer-weeks.
Before vs. After
Baseline (first 14 days) vs. final window (last 14 days) of the full-scale reporting period.
| Dimension | Before | After | Δ | |
|---|---|---|---|---|
| Ceremony observation (reports/wk) | 14 | 616 | 44× volume | ▲ |
| Dysfunction alarms / week | 31 | 2,446 | 79× detection | ▲ |
| AI Jira comments / week | 2 | 421 | 210× automation | ▲ |
| Customer Focus (code 204) | 17.54 | 37.20 | +112% | ▲ |
| Retro Improvements (code 221) | 16.18 | 50.00 | +209% | ▲ |
| Heartbeat (code 226) | 67.19 | 71.55 | +6.5% | ▲ |
| Customer Lead Time (code 222) | 2.91 | 2.84 | −2.4% | ▲ |
| Speaking Share (code 216) | 0.10 | 16.07 | ~0 → 16 | ▲ |
| Motivation & Engagement (code 201) | 67.46 | 66.57 | −1.3% (stable) | → |
| Psych Safety (code 202) | 79.06 | 69.28 | −9.8 pts | ▼ |
| Discipline & Self-Mgmt (code 203) | 47.64 | 42.58 | −5.1 pts | ▼ |
| Dependencies surfaced / wk | 14 | 110 | ~8× surfacing | ▲ |
| Teams with zero retros | 2 observed in 14d | 23 of 77 | Made visible | ● |
| Teams with zero sprint reviews | 2 observed in 14d | ~42 of 77 | Made visible | ● |
Two trends moved negatively (Psych Safety, Discipline). Both declines are now quantified and team-specific, which makes them actionable.
Lessons Learned
✓What Worked
Full-org scope
Portfolio-level patterns (backlog debt, missing-retro concentration, dependency hotspots) only emerge at full scale.
High-confidence automation
Jira write-back at 0.80+ confidence. Zero complaints about hallucinated Jira noise in 110 days.
Invisible integration
Teams did not have to change how they worked. Integration was invisible to engineers.
Local-language handling
76 of 77 teams operated in Turkish. Transparency about error margins built trust.
Two-way adoption
Value flowed top-down (leadership citing findings) and bottom-up (SMs acting on tactics) simultaneously.
⚠Limits & Open Gaps
Delivery baseline — empty sprint_metrics
Delivery velocity could not be baselined. Future deployments should validate telemetry during onboarding.
Financial inputs — null FTE-cost
Financial modelling was directional. Customers should populate FTE-cost before deployment start.
Alarm pipeline gap
Late changes in detection pipeline affected Code 2 and Code 5 alarm behaviour. Documented and scoped.
Resolution closure
oNabu surfaces blockers but does not currently track resolution. Closing this loop is a roadmap item.
Zero-ceremony teams
Teams with no retros/reviews cannot be scored. Future: a dedicated scoring track for ceremony-absent teams.
Adoption Factors
Scrum Master Engagement
The 69 TEAM_ADMINs were the adoption bottleneck. Where SMs engaged early, their teams moved metrics.
Top-down Pull
Once leadership cited oNabu findings in portfolio reviews, scrum masters began acting on them.
Localisation
Reports generated in Turkish were read. English-only would not have been. Localisation was adoption-critical.
Conclusion
oNabu's deployment was not an improvement exercise. It was a visibility exercise.
The organisation did not become more agile because oNabu told it to be. It became legible — measurable, comparable, and governable at portfolio scale — because oNabu instrumented what was already happening and wrote the signal back into systems leadership already uses.
CORE VALUE
oNabu is a Next-Generation Management Layer for agile organizations
At 77-team scale, no human can attend every ceremony, read every transcript, or cross-correlate every blocker. oNabu does — continuously, at ≥ 0.80 confidence, writing findings back into Jira so that signals do not die in meeting memory. The product is the compression of 11,000 meetings into decisions leadership can act on in the current sprint, not the next quarter.
Where it fits best
Enterprises running 15+ agile teams on Jira, Azure DevOps, Zoom/Teams/Meet, and Slack — where manual telemetry has already broken down.
Organizations in active digital-transformation programs where agile adoption is uneven and leadership lacks a common view of ceremony execution.
Multilingual enterprises — oNabu handles non-English ceremonies natively, which most US-origin observability tools do not.
Organizations where the cost of a missed sprint is material and surfacing cross-team blockers one week earlier changes the quarter.
Recommended Next Step
Based on the deployment results, the recommended next step for LC Waikiki is a continuation scope covering:
Cut-off date for numeric findings: April 17 2026.
All numeric findings derived from oNabu data sources and analytics snapshots.
· Next-Generation Management Layer · onabu.ai