Kindo
Sprint Planning
← Back to Main
Kindo × Deloitte Program

Sprint 1 Planning

First scrum cadence planning — aligning strategic decisions, team, and execution backlog

📅 Monday, June 29, 2026

Meeting Agenda

Sprint 1 Planning — 60 minutes. First formal scrum cadence session.

⚡ Context: June 24 strategy session + June 26 working session confirmed 2-week scrum cadence starting June 29. Tony steps back to biweekly sprint planning (product owner cadence). SOC for AI is the new #1 strategic priority per Kush. Agents A.6 and A.10 deprioritized.
📋 Session Flow
TimeTopicOwnerGoal
0–8 min Cadence & Ceremonies Joana Confirm scrum structure, ceremonies, artifacts
8–15 min Team & Hiring Joana / Victor Current team, Value First hires, role shifts
15–25 min Portfolio Overview Joana Agent status, priority shift, Kush decisions
25–35 min Deep Dive Sessions Tony / Joana Quarterly design sessions plan, first deep dive scope
35–52 min Sprint 1 Planning All Select backlog items, assign, define sprint goal
52–60 min Open Decisions All Blockers, needs-human items, next steps
S1
First Sprint
2W
Sprint Length
Jul 10
Sprint 1 End

Cadence & Ceremonies

Three-layer cadence: 2-week sprints, monthly portfolio, quarterly deep dives

🏃 Sprint (2 weeks)

Execution cadence. Ship code, build agents, deliver results.

PlanningMonday, Day 1
StandupsMon/Wed/Fri (team only)
Review + RetroFriday, Day 10
Tony attendsPlanning + Review only
📊 Portfolio (monthly)

Show results to executives. Video evidence + working links.

First meeting~Late July (Jul 27)
AttendeesTony, Charlie + Ron
FormatShow, don't tell (see below)
NoteSame city as quarterly deep dive to avoid over-travel
🔬 Deep Dive (quarterly)

In-person design sessions. Strategic alignment + release planning.

Duration~5-7 days
Q3 locationHouston, TX (aligned with July monthly portfolio — Ron anchor event Jul 27)
WithTeam + Ron + key engineers
OutputQuarter release plan
📐 Scrum Artifacts — Automated Delivery
The team programmatically delivers real-time sprint artifacts. No asking for status — artifacts should be self-service.
ArtifactCadenceDescriptionStatus
Sprint Backlog Updated daily Items committed for current sprint, status, blockers Sprint 1
Burndown Real-time Story points or items remaining vs time Sprint 1
Sprint Review Deck End of sprint Video evidence + working links of shipped work Sprint 1
Definition of Ready Standing Criteria for items entering sprint (see below) Defined
Definition of Done Standing Criteria for items being complete (see below) Defined
🎬 "Show Don't Tell" — What Tony Means

We're moving at 5× traditional speed. Words can't keep up. Ron didn't understand net new revenue agents until the third time Tony explained it — and Ron is the sharpest person in the room. If Ron needs three reps to absorb a verbal summary, everyone else needs more.

The old way (what we stop doing):

  • Verbal status updates: "A.5 is code complete, waiting on merge"
  • Slides with bullet points about progress
  • Written summaries explaining what happened
  • "Technical done" without "business done" evidence

The new way (what every portfolio item needs):

  • 🎥 Video walkthrough — screen recording of the feature working in production (30-90 sec)
  • 🔗 Live URL — clickable link where the stakeholder can see it themselves, right now
  • 📸 Before/after screenshots — visual proof of what changed
  • 📊 Metrics — measurable impact (e.g., "21min per alert → 5min")
Tony's original architecture: Akira AI PMO confirms requirements via video → team builds → confirms results via video. Both ends are visual. If you can't show it working on screen, it's not done enough to present. This applies to everyone — Krishna, Kush, Ron, Forge Point.
🗓️ Proposed Location Rotation (Monthly Portfolio)
MonthLocationNotes
JulyHoustonRon in Houston Jul 27 — anchor event
AugustSan FranciscoKrishna's base
SeptemberAustinTony's base
OctoberLos AngelesCharlie's base

✅ Deep dives aligned with monthly portfolio locations to avoid over-travel (Q3 → Houston, Q4 → Los Angeles).

Team & Hiring

Current team, role evolution, and Value First hiring plan

👥 Current Team
PersonCurrent RoleEvolving ToSprint Role
Tony Strategic leadership Product Owner (biweekly) Sprint planning + review only
Joana Program delivery Net new revenue agent design Scrum lead + agent designer
Victor Technical delivery Net new revenue agent design Technical lead + agent designer
Charlie Chief Architect / Agent Runtime Platform + architecture decisions Technical advisor
AI Delivery Engine Engineering & Ops AI Autonomous Delivery Partner Delivery method engine, scrum artifacts
Dukane Delivery support QA manager Output quality review
⚠️ Role Shift: Joana & Victor moving from program delivery (100 installs, training) → net new revenue agent design. Hires will backfill the program delivery gap.
🤝 Hiring Plan — Value First / Omberto
RoleCountRegionFocusStatus
AI PMO / Soft Skills 1-2 LatAm (preferred) Requirements gathering, stakeholder mgmt, verification Interviewing
Engineer 1-2 Eastern Europe or LatAm MLflow, Kindo agent configuration, integrations Planning

Process: Invoice → Charlie → Ron. Charlie vets technical candidates. LatAm for soft-skills (live meetings), Eastern Europe for code (Charlie's preference).

🔧 Engineering Availability
June Blocker: Agent Runtime team (Madison, Sean) on extended PTO through end of June. Core Kindo engineering (Brian Van's team) out for 3 weeks. Expected to normalize in July.
TeamStatusExpected BackImpact
Agent Runtime (Madison) PTO Early July 2.5 weeks PTO — multi-agent, agent features blocked
Agent Runtime (Sean) PTO Week of Jul 7 Working ~2 days/week past 2 weeks + off next week
Core Kindo (Brian Van) PTO Mid-July Core platform changes blocked
Charlie (Agent Runtime lead) Active Shipped memory prototype solo

Sprint 1 implication: Focus on soft-skills deliverables (requirements, agent design, research) that don't need Kindo eng. Engineering-dependent items slot into Sprint 2+ when team is back.

Portfolio Status

Agent portfolio with June 24 priority shift — SOC for AI is the new #1

🔴 Priority Shift (Kush, June 22): SOC for AI takes top priority. A.6 (Vitals Dashboard) and A.10 (IoT/OT Monitor) deprioritized. We must produce results faster — A.6 took 6 weeks to align, same pattern as Generative UI. If we don't capture, Deloitte builds it themselves.
🗺️ Agent Status Map
IDAgentStatusRevenue ClassBlocker
A.1Threat Monitoring PROD Contracted
A.2Threat Intel PROD Contracted
A.3Threat Hunt PROD Contracted
A.4Detection Engineering PROD Contracted
A.5CTEM BUILT ContractedDeployment pending
A.6Vitals Dashboard ⏸️ DEPRIORITIZED AllianceKush shifted to SOC for AI
A.7Quality Audit Agent REQS AllianceDesign sprint needed
A.8Cloud Security Agent PLANNED Alliance
A.9IR Agent PLANNED Alliance
A.10IoT/OT Monitor ⏸️ DEPRIORITIZED AllianceKush shifted to SOC for AI
A.11Custom Client Agents REQS AllianceShadow & document method
A.12Identity Agent → IdaaS PLANNED AllianceTim Corder engagement
A.13GRC Agent → GRC aaS PLANNED AllianceNathan Ellis engagement
NEWSOC for AI 🔴 #1 PRIORITY AllianceResearch + integration mapping
🛡️ SOC for AI — Platform Governance Scope (Reframed Jul 7)

Reframed Jul 7 based on Kishore's clarification. SOC for AI monitors known Kindo agents for governance violations — not enterprise shadow AI discovery (handled by Deloitte's detection engineering team). Platform-native: all 5 objectives consume Kindo audit logs and control plane APIs.

🎯 Objective 1 — Detect unauthorized agent deployment/modification

🔌 Objective 2 — Detect unauthorized tool/data source connections

📊 Objective 3 — Detect behavioral drift from SOPs over time

🔒 Objective 4 — Detect guardrail/policy changes

🏯 Objective 5 — Detect cross-tenant data contamination

🔌 Platform Data Sources

All governance monitoring data comes from Kindo's own control plane APIs and audit logs. No external integrations required for v1.

Endpoint / SourceData Provided
GET /v1/agents/list Agent inventory (name, creator, models, recent runs)
GET /v1/integrations/connections Integration connections (type, owner, creation date)
GET /v1/models Enabled models
Audit Log (Enterprise) All agent actions, config changes, tool invocations, DLP events
SMK/Syslog Forwarding SIEM export for external detection rules
🏯 Tenant Survival Filter (Tony's note #1): Every data source must pass: (1) tenant-scoped? (2) data stays in tenant, no egress? (3) SOC 2 Type II / BAA / DLP compatible? Cross-tenant/egress = disqualified.

Safest bet: Kindo's own control plane — all API calls and audit logs are inherently tenant-scoped. No cross-tenant risk for platform governance monitoring.
📊 Revenue Classification
$5.5M
Contracted (A.1–A.5)
$1-2M+
Alliance Net New (A.6–A.13)
$5-12M+
Upside (2-3× Expansion)

Deep Dive Design Sessions

Quarterly in-person sessions — strategic alignment + release planning with Deloitte

Origin: Tony proposed quarterly deep dives 6 months ago — modeled after what he and Ron did independently in December (7 days of uninterrupted deep thinking). Deloitte "wholeheartedly agreed" at the June 22 meeting.
🎯 Deep Dive Topics (from Deloitte meeting)

These items were categorized under "design sessions" in Tony's meeting notes. More than half of what Deloitte raised maps to these sessions.

#TopicDescriptionOwner
1 Net New Revenue Agents Design + deploy agents that generate alliance revenue (Tier 2/3 packages) Tony / Joana
2 Threat Remediation Extend A.1-A.5 into automated remediation workflows Charlie / Victor
3 Deloitte Roadmap (Azure/GCP) Cloud platform alignment and multi-cloud strategy Charlie
4 Institutional Knowledge / Memory → Skills, memory, compound learning flywheel. View IK Approach Charlie
5 AI Cyber Guard / Tower Control plane co-development Charlie
6 Lifecycle Hooks Generic lifecycle hooks — Kush says yes but NOT most important. Ship fast MVP, don't over-engineer. Deployment speed > stickiness features. Charlie
7 Workflow Acceleration Accelerate deployment cycle (time-to-value for new Kindo customers) Victor / Joana
8 SOC for AI Shadow IT discovery, AI governance, integration mapping Joana / Victor
📅 Proposed Deep Dive Calendar

Q3 2026 — First Deep Dive

Houston, TX — aligned with July monthly portfolio (Ron in Houston Jul 27). 5-7 day in-person session. Team + Ron + key engineers. Output: Q3 release plan, SOC for AI architecture, net new revenue agent designs.

Q4 2026 — Second Deep Dive

Los Angeles, CA — aligned with October monthly portfolio (Charlie's base). Review Q3 results, plan Q4 releases, expand to service lines beyond D&RaaS (Identity aaS, GRC aaS).

Sprint 1 — June 29 → July 10

First sprint: focus on soft-skills deliverables while engineering is on PTO

🎯 Sprint Goal

Validate scope ✅ → Pivot to platform governance → Deliver Governance Monitor Agent PoC

Scope validated Jul 7 — Deloitte's SOC team confirmed SOC for AI = monitoring known Kindo agents, not enterprise shadow AI discovery. Sprint pivoted to platform governance. Deliverable: working Kindo Scheduled Agent that monitors agent inventory, integration connections, and model usage against governance baselines.

📋 Proposed Sprint 1 Backlog
ItemTypeOwnerEffortDependencies / Risks
SOC for AI — Scope Confirmation
✅ DONE. Deloitte (Kishore) confirmed: shadow AI discovery = their detection engineering team. SOC for AI = Kindo platform governance monitoring. 5 objectives defined.
P0 Joana 1d ✅ DONE
SOC for AI — Capability Mapping + Odin Assessment
✅ DONE. Kindo audit logging, RBAC, DLP mapped against 5 objectives. API endpoints confirmed (GET /v1/agents/list, GET /v1/integrations/connections, GET /v1/models). Gaps identified: drift detection, policy-change alerting, cross-tenant contamination detection.
P0 Delivery Team + AI Engine 0.5d ✅ DONE
SOC for AI — Governance Monitor Agent: Design + Build v1 in Kindo
⚠️ Blocked: Enterprise access needed for API Action steps. Agent design complete: 4-step workflow (agent inventory → integration connections → models → LLM governance analysis). Unblocked once Enterprise access confirmed.
P1 Delivery Team 2d human / 2d AI-assisted ⚠️ Blocked: Enterprise access needed
🔥 Sprint 1 Meta-Risk: The entire sprint is optimized for soft-skills work because eng is on PTO. That's smart — we're shooting where we're unblocked. But if eng comes back mid-sprint and unblocks code work, do NOT mid-sprint pivot. Finish what we committed to. New engineering items go into Sprint 2 backlog.
📦 Delivery Method Alignment: Every Sprint 1 item must map to the Deterministic Outcome Package — (1) Named Owner, (2) Transcript/Intake Artifact, (3) Standing Rules, (4) Cron-Sustained Refresh, (5) Recipient's Narrative. Evidence Base 85% pattern applies: named human owner + same-day delivery + output in recipient's language.
Sprint 1 Exit Criteria

Show-don't-tell: every criterion needs evidence, not a status update.

  • Scope confirmed by Deloitte → deliverable: Kishore's 5 objectives documented + capability mapping at /capability-mapping/
  • Platform capability mapped with API-level detail → deliverable: Odin assessment integrated into capability mapping page
  • Governance Monitor Agent v1 built in Kindo → deliverable: working scheduled agent + design doc + video walkthrough (⚠️ blocked on Enterprise access)
  • Detection use case catalog (4+ use cases) → deliverable: SIEM-ready detection rules mapped to Kindo audit events
  • Automated sprint status delivery → deliverable: live sprint dashboard URL (program track)
🚫 Sprint 1 — NOT In Scope
  • Shadow AI discovery agent (Deloitte's detection engineering team owns this) — out of scope per Kishore
  • CrowdStrike/Defender endpoint integration for discovery — not needed for platform governance
  • Code shipping to Kindo platform (eng on PTO) — Sprint 2 when team returns
  • Core platform changes (Brian's team required) — blocked, not our call
  • A.6 Vitals Dashboard (deprioritized by Kush) — parked
  • Scaling Story for Ron (important but not SOC for AI) — separate workstream, not sprint backlog
  • A.7 Quality Audit design sprint (depends on Krishna scheduling) — backlog, not Sprint 1
  • Hiring decisions (interviews in progress) — parallel track
Definition of Ready (DoR)

An item can enter the sprint when ALL of these are true:

#CriteriaWhy
1Clear outcome defined — what does "done" look like in business terms, not technical terms?Victor's point: business value, not technical value
2Owner assigned — single person accountableNo orphan items
3Dependencies identified — blocked/unblocked explicitly taggedVictor's framework: shoot where we're unblocked
4Effort estimated — days, not points. Be honest.Tony needs to know what to expect without asking
5Classified soft-skill vs code — which work type? Determines who can execute.~80% soft-skills moves without Brian's team
6Passes tenant filter (if integration) — tenant-scoped? data stays in tenant? SOC 2 II / BAA / DLP?Tony's note #1: "Deloitte only"
7Fits the sprint — total committed work ≤ team capacityDon't overcommit then under-deliver
8Acceptance criteria written — how will we verify it's done?"Show don't tell" starts here
🏁 Definition of Done (DoD)

An item is done when ALL of these are true:

#CriteriaEvidence Required
1Acceptance criteria met — every criterion checked off with proofScreenshots, video, or live URL
2Business done, not just technical done — stakeholder can see and use itWorking link or deployed artifact
3Evidence attached — "show don't tell" proof in the same message as the completion claimVideo walkthrough, screen recording, API response
4Integrations pass tenant + compliance checkTenant filter results documented
5No open blockers or regressionsVerification report
6Reviewed — at least one other team member has seen the outputReviewer name + ✅/⚠️/❌
7Documented for automated status reportingUpdates programmatically
8Owner confirmed doneExplicit sign-off
Hard rule: "Code complete" ≠ done. "Waiting on merge" ≠ done. "I verified" without the actual evidence ≠ done. If you can't show it on screen, it's not done.

Product Backlog

Items that produce a built artifact (agent, integration, code in Kindo) — ranked by strategic priority

Prioritization framework: SOC for AI Platform Governance is #1 (Kush mandate, refined by Kishore Jul 7). Then net new revenue agents (alliance revenue). Then contracted platform work. Unblocked items before blocked items (self-unblocking bias).
P0 — Must Do Now 2 items
SOC for AI — Platform Governance Scope Confirmation
Kishore's team clarified SOC for AI = monitoring Kindo's own agents for drift, unauthorized changes, and data contamination — not enterprise shadow AI discovery. Five objectives defined. Capability mapping complete. Three targeted questions + four pre-built recommendations ready for Friday meeting.
✅ Capability Mapped 3 Questions + 4 Recommendations
Reframed Jul 7 based on Kishore's clarification. Shadow AI discovery handled by Deloitte's detection engineering team. Full capability mapping at /capability-mapping/
SOC for AI — Platform Monitoring Agent: Design + Build v1
Design + build a Kindo agent that monitors Kindo agents. Covers Objectives 1 (unauthorized deployment), 2 (unauthorized integrations), and 4 (policy/guardrail changes). Uses existing audit log API, agent inventory API, and integration connections API. No new platform instrumentation required.
Agent Design Platform Governance Unblocked
Replaces shadow AI Discovery Agent. Existing telemetry + audit APIs confirmed sufficient. Sprint 2 build target.
P1 — High Priority 5 items
IK / Behavioral Drift Detection — Design Session
Objective 3 (SOP compliance monitoring) is the most complex and highest-value objective. SME-driven IK approach: Engineering Lead does sessions with the agent on real scenarios, encoding judgment patterns into structured eval curriculum. Design in Sprint 2, build in Sprint 3.
IK Capture Promoted from P3 Needs: SME confirmation
A.7 Quality Audit Agent — Design Sprint
Option 1: High-bandwidth meeting with Krishna's team (~3hrs). Design the 5-agent Quality Audit package (Alert Scorer, Human Baseline Validator, Efficiency Tracker, Pool Optimizer, Audit Dashboard).
Agent Design Joana Blocked: Krishna scheduling
A.11 Custom Client Agents — Shadow & Document
Option 2: AI ingests SOPs + ride-along with analysts. Capture institutional knowledge for custom agent patterns. Start with HP deployment patterns.
IK Capture Victor Depends: analyst access
A.5 CTEM — Production Deployment
CTEM is built, needs deployment. Blocked on Kindo eng availability but should be first code ship when team returns.
Deployment Blocked: Eng PTO
Show-Don't-Tell — Video Evidence Pipeline
Establish process for capturing video evidence of working functionality. Every P0/P1 should have video proof + working URL. Tony: "We're moving too fast for words."
Delivery Mechanism Delivery Team Unblocked
P2 — Medium Priority 3 items
Cross-Tenant Contamination Detection (Objective 5)
Detect when an agent for one client cross-references data of another client. Prevention controls exist (tenant-scoped memory, cross-tenant referent checks, pinned connections). Detection layer needs building. Depends on multi-tenancy architecture confirmation from Q1.
Platform Governance Depends: Q1 architecture answer
A.8 Cloud Security Agent — Requirements
Kush's Deloitte roadmap includes Azure/GCP alignment. Design cloud security agent leveraging cloud-native tools.
Agent Design Depends: Deep Dive
A.9 IR Agent — Requirements
Incident Response automation agent. Extends threat monitoring (A.1) into response workflows.
Agent Design Depends: Deep Dive
P3 — Future / Deep Dive Topics 5 items
A.12 Identity Agent → IdaaS (Tim Corder)
Service line expansion into Identity aaS. Requires engagement with Tim Corder + Ravi.
Service Line ExpansionPhase 4
A.13 GRC Agent → GRC aaS (Nathan Ellis)
Service line expansion into GRC aaS. Requires engagement with Nathan Ellis.
Service Line ExpansionPhase 4
Lifecycle Hooks MVP
Kush: short answer is yes, but NOT most important. Ship fast MVP — don't over-engineer. Focus: deployment speed, not stickiness.
PlatformKush deprioritized
AI Cyber Guard / Tower — Control Plane
Control plane co-development with Deloitte. Design session topic.
Deep DiveCharlie
SaaS Discovery Integrations (Nightfall, Netskope, Okta)
CASB/SaaS-level AI discovery tools. No Kindo integrations today. Lower priority than endpoint-level discovery.
IntegrationP3
⏸️ PARKED 2 items
A.6 Vitals Dashboard
Deprioritized by Kush (June 22). SOC for AI takes its slot. May resurface in future sprints.
Deprioritized
A.10 IoT/OT Monitor
Deprioritized by Kush (June 22).
Deprioritized

Program / Strategic Track

Parallel track — program management work that needs an owner and date but doesn't produce a product artifact. Not sprint-rankable. Joana's track.

📋 Why this is separate: These items are essential program work — investor narratives, portfolio prep, hiring, travel planning — but they don't produce a built artifact in Kindo. They run on their own timelines with their own owners, parallel to the sprint. Mixing them into P0–P3 product lanes creates false prioritization conflicts.
ItemOwnerTarget DateNotes
Scaling Story for Ron / Forge Point Tony + Joana During Sprint 1 Capture while Tony is present — he goes biweekly after. 3-5× revenue growth narrative for Forge Point VC.
Monthly Portfolio Prep (July) Joana Ahead of Jul 27 Houston Video evidence + working links for shipped functionality.
Value First Hiring Joana / Victor Ongoing Interviews in progress. Invoice → Charlie → Ron.
Deep Dive Prep / Calendar + Budget Tony / Joana TBD ✅ Done — deep dives aligned to monthly portfolio cities (Houston Q3, LA Q4).
Automated Scrum Artifacts Setup Joana + Delivery Team Sprint 1 Configure automated sprint backlog, burndown, status delivery. ⚠️ Verify accuracy before relying on automated output.
Backlog reframed within 4 hours of Kishore's scope clarification — capability mapping, dependency analysis, and sprint impact assessment produced from existing platform knowledge. Continuous requirements mapping ensures scope changes accelerate rather than reset delivery.

Open Decisions & Needs-Human

Items requiring team input during Monday's planning

🔴 SOC for AI — Scope Confirmation (Joana → Charlie & Victor, Jun 25)

5 scope questions sent Jun 25. 2 resolved (Q1, Q3). 3 remaining (Q2, Q4, Q5) — resolve in today's planning. Source: SOC for AI — Scope & Research doc.

1️⃣
Registry check (Victor/Charlie) ✅ RESOLVED All 6 confirmed: CrowdStrike (OAUTH2_CC), Splunk (API_KEY), Datadog (API_KEY), Grafana (API_KEY), Sumo Logic (BASIC), Google SecOps (OAUTH2). Microsoft stack (Defender/Intune/Purview) covered via Microsoft MCP — tenant-side config, not new builds. Caveat: Deloitte manages their own Kindo instance — need their team to confirm what's active.
2️⃣
CrowdStrike tenant isolation (Charlie) ✅ SUPERSEDED Moot for SOC-for-AI scope — Deloitte's detection engineering team handles endpoint/shadow AI discovery. CrowdStrike integration still relevant for other agent work but no longer gates SOC for AI.
3️⃣
Microsoft path (Charlie) ✅ RESOLVED Confirmed: MS Defender/Intune/Purview are covered via Microsoft MCP (same Entra app registration flow). Not a separate build — tenant-side configuration. Doc: docs.kindo.ai/connecting-integrations/microsoft-mcp
4️⃣
First build — Governance Monitor Agent ✅ RESOLVED Kishore's Jul 7 clarification resolved this: first build = Kindo agent monitoring Kindo agents (governance, not discovery). Design complete. Blocked on Enterprise access for API Action steps.
5️⃣
Scope + control plane (both) ✅ RESOLVED Kishore's Jul 7 response resolved: SOC for AI = Kindo platform governance (5 objectives). Shadow AI discovery = Deloitte's detection engineering team. Two-layer (A/B) scope replaced by 5-objective framework. Kindo instance access still needed for PoC deployment (Enterprise access pending).
🟡 Other Open Items
📋
Sprint 1 scope — confirm or adjust proposed items Do the SOC for AI items match what we can realistically ship in 2 weeks with eng on PTO? Each item should map to the Delivery Method's 5-component Outcome Package (named owner, transcript, standing rules, cron refresh, recipient narrative). Apply Evidence Base 85% pattern.
📋
SOC for AI requirements — governance-focused Original questionnaire superseded by Kishore's 5-objective framework. New requirements for Friday call: SIEM target, audit export method, "unauthorized" definition, SOP documentation for drift baselines, tenant architecture. Owner: Delivery Team.
📋
Value First hiring — interview status Joana met with Omberto's team. Results feed into team planning but don't block Sprint 1. 1-2 AI PMO / soft-skills roles + 1-2 engineers (LatAm preferred for soft-skills, Eastern Europe for code).
📋
Agent runtime team return date — confirm July availability Madison back early July, Sean week of Jul 7. Track for Sprint 2 planning.

Evidence Base

110 days of empirical data (Mar 9 – Jun 26, 2026) — what worked, what failed, what was killed

⚡ Why this matters for Sprint 1: This sprint is the first cycle born after the March→May→June learning curve (Additive → Peak Complexity → Subtraction). Every item should operate on the validated methods, carrying none of the killed ones. Full audit: internal audit reference
85%
Kindo Dashboards (best)
40%
Pipeline Execution (worst)
7
Systems Killed (Jun 17)
📊 Initiative Grades — Measured, Not Claimed
InitiativeGradeKey Evidence
Kindo × Deloitte Dashboards 85% 2 production dashboards, daily cron, 240 deploys. Owner (Joana) + same-day loop + cron refresh
Strategic Dashboards 80% Same-day delivery pattern. Revenue Map for Ron dinner = built same day Tony directed
Kindo LMS 70% Training videos re-generated (T1-T4), next-button gating shipped. Architecture stabilized on Deloitte-specific worker.
Tony CoS Dashboard 55% 136 deploys, CI/CD working. But: DM capture unverified, Phase 2 never started
Autonomous Pipeline 40% Architecture complete (43 routes). But: 0 agents dispatched for 42+ days. Pipeline became meta-work
The 85% Pattern (what works)
  • Named human owner driving the loop (Joana, Tony)
  • Same-day delivery — directive → artifact in hours, not days
  • Output in recipient's language, not internal jargon
  • Cron-sustained refresh — keeps it alive after delivery
  • Synchronous sessions > async (4h live = weeks of async)
The 40% Pattern (what fails)
  • Autonomous pipeline without human driving = meta-work
  • AI judging AI = shared fault amplification (40-48% pass rate)
  • Process docs as probabilistic input = unreliable
  • Complexity accumulation without outcome measurement
  • Demo site proliferation — 8+ sites, most never viewed
💡 The Arc — March → May → June

March–April: Build

Pipeline architecture, 67 routes, Sprint 0, first dashboards. Every problem solved by adding.

May: Peak Complexity

6 new Pages in 3 weeks. 5 AI eval crons. 4 daily dossiers. Maximum complexity = diminishing returns.

June: Subtract

7 systems killed. 0% dossier engagement. Silence First. Human-only review. Higher signal than anything added.

🎯 Capability Resonance — What Customers Actually React To
#CapabilityWho ReactedWhen
1Thinking model / decision OSIgor: "Jarvis not Siri"May 15
2Knowledge extractionSteve Ward: "floored"May 25
3AIPMO / autonomous PMValent: SOW signed ($5K)Apr–Jun
4Dependency mappingNFL corpus proof pointApr
5Sprint planning / estimationHector: "that's money"Jun
⚠️ Guard Rail for Sprint 1: If the number of systems, crons, or processes starts climbing without outcome improvement, subtract before adding. The March→May arc proved complexity accumulation is the default — it has to be actively resisted. Max 1-5 individual changes at a time (Tony, Jun 19).

Delivery Method

Autonomous Delivery Partner — the product is the method, AI is the engine

Key insight (Tony + Victor, Jun 26): "The product is this delivery method — with AI as the engine that makes it run at the speed and quality level that a human team can't match." The AI engine transforms non-technical creative direction into shipped products. Not autonomous SDLC — Autonomous Delivery Partner.
📦 Deterministic Outcome Package — 5 Components

Every load-bearing milestone (85% of outcomes) used these 5 components. Each Sprint 1 item should map to this template.

#ComponentWhat It MeansSprint 1 Check
1 Named Owner A human on the customer side driving the loop (the "Tony" equivalent) Who is the owner for each item?
2 Transcript / Intake Artifact Customer's own words, not our interpretation Do we have source material in their language?
3 Standing Rules Deterministic rules, written — never probabilistic process docs Are the rules codified or still in people's heads?
4 Cron-Sustained Refresh Keeps the output alive after initial delivery Will this need automated updates or is it one-shot?
5 Recipient's Narrative Output framed in the customer's framework, not VtKl's internal language Are we using Krishna's/Kush's words or ours?
🔬 The Operating Pair — Appears in 5/8 Load-Bearing Milestones

#1 Owner + Same-Day Loop

The mechanism. Without a named owner driving the loop, nothing ships. Present in: Kindo dashboards, CDO thesis, Execution Plan + Revenue Map, Strategic Portfolio Design, Great Subtraction.

#11 Narrative (Recipient's Framework)

The communication layer. Without narrative framing, output ships but doesn't land. Ron needed visual revenue framing. Igor needed "Jarvis" framing. Krishna needed triage language.

🗺️ Validated at Three Altitudes

🏔️

Executive

Revenue Map for Ron dinner (May 21-23). Visual, revenue-framed.

⚙️

Strategic

Sprint plan + GANTT + questionnaire (Jun 8-9). Full planning cycle in one thread.

📊

Operational

Daily dashboard refresh via cron. 240 deploys. Deloitte logs in daily.

🧭 Role Evolution — Self-Organized, Not Designed

Each person migrated UP in altitude over 110 days. Sprint 1 should respect these altitudes.

PersonStarted AsEvolved ToSprint 1 Altitude
TonyOperator (every function)Teacher → SubtractorStrategic direction only. Biweekly.
VictorOps SupportChief Operating IntelligenceTactical execution + AI calibration
JoanaProgram DeliveryDelivery AuthorityScrum lead + agent design
CharlieBuilderPlatform ArchitectArchitecture decisions only
AI Delivery EngineEngineering ToolAutonomous Delivery PartnerExecution engine — absorbs operations
⚠️ Strategic Warning (Charlie, Jun 24)

"I don't want to do design partnership work where we equip them to build stuff that we want to build." Deloitte can execute faster than T&C on skills/memory if they know the HOW. Share the WHAT, never the HOW. This applies to all IK/memory sessions with Kush.

📋 Requirements Questionnaires — Sprint 1 Pre-Work

Tony (Jun 26): needs questionnaires for SOC for AI, A.7, A.11 — same format as A.6. Forward to Krishna who identifies who answers (resolves Adelina maternity leave gap without guessing org structure).

AgentMethodAdelina CoverageStatus
SOC for AI Questionnaire (framed as Kush's priority) Krishna routes to right person ⚠️ Dependency — generate + send early in sprint
A.7 Quality Audit Design Sprint questionnaire (~3hr session) Krishna routes to QA lead Sprint 2+ prep
A.11 Custom Agents Shadow & Document (SOPs + ride-along) Shiva/Harish (client delivery) Sprint 2+ prep
🔵 Blue Ocean (Victor, Jun 26): Agile, Scrum, SAFe — all made for humans, not AI. What T&C is building is the AI agility cycle. Greenfield. Nobody is talking about this yet. Tony isn't just talking — he has implemented and proved the value. The quarterly deep dives = "AI Big Room Planning" — release planning with AI in the room collapsing strategy-to-artifact gap to zero.