Skip to main content

Chatbot (BOT Squad) — Program Roadmap

Delivery / Program Management — team-level rollup across the team's initiatives.

Portfolio

InitiativeTarget quarterHealthDRIImplementorsQATimeline
Autonomous AI Agent2026-Q3At RiskDimas Fauzi Hidayatfarras.herdicho, samudra.linggartito.amandatimeline
AI Agent Knowledge2026-Q3Dimas Fauzi Hidayateko.aprianto, samudra.linggartito.amanda, izzultimeline
AI Agent Testing2026-Q3Dimas Fauzi Hidayatfarras.herdicho, baghiz.firdausa, eko.aprianto, samudra.linggartito.amanda, izzultimeline
Bot-vs-Human Traffic Split (A/B)2026-Q3Dimas Fauzi Hidayattimeline
Mekari Action (AI Agent)2026-Q3Dimas Fauzi Hidayattimeline
Native Integration (AI Agent)2026-Q3Dimas Fauzi Hidayatagus.suparman, baghiz.firdausatito.amanda, izzultimeline
Qontak Action (AI Agent)2026-Q3Dimas Fauzi Hidayateko.aprianto, agus.suparman, baghiz.firdausatimeline
Unified Agent Quality Scorecard2026-Q3Dimas Fauzi Hidayattimeline
AI Agent Idle Action2026-Q3Dimas Fauzi Hidayat
AI Agent Monitoring2026-Q3Dimas Fauzi Hidayat
Multi-WABA Impact2026-Q3Dimas Fauzi Hidayat
AI Agent Impact Report2026-Q3Dimas Fauzi Hidayatagus.suparman, baghiz.firdausatito.amanda, izzul
Meta Business Agent Integration2026-Q3Dimas Fauzi Hidayat
AI Spam Gatekeeper2026-Q3Dimas Fauzi Hidayat
Automation Flow Builder2026-Q3Dimas Fauzi Hidayat
AI Agent Memory2026-Q4Dimas Fauzi Hidayat
Branch Bot Response — Customer/Contact Field2026-Q2CompletedDimas Fauzi Hidayatagus.suparman, samudra.linggarizzul
Chatbot 26Q3 Engineering Initiatives (bucket)2026-Q3Backlogagus.suparmanagus.suparman, eko.aprianto, samudra.linggar, baghiz.firdausatito.amanda, izzulREADME
Chatbot Q3 2026 Code Readiness (bucket)2026-Q3BacklogPuji Triwibowobaghiz.firdausa, samudra.linggar, farras.herdichoREADME

Priority

PriorityInitiativeDepends onNote
P1Autonomous AI Agent — Phase 2: AI-Assisted Refinement (Refiner)Design: Ready (Wulan's live "Refine" prototype; Figma to follow). Blocking Data/ML refine-skill-pack dep (see External deps below); a possible duplicate P4 "Iteration" item (Qontak_AI_Agent_Research_Report.docx, likely Phase 4) is left unranked pending PM confirmation.
P2AI Agent Impact Report — Phase 1: Live Impact ReportDesign: Not yet — PRD header marks Figma as pending; a plain-language wireframe and concept prompts exist as references only. Depends on Data/BI's per-organization activity-log datamart (blocking).
P3AI Agent Testing — Phase 1: Historical ValidationDesign: Ready — design is finished and linked in the PRD. Depends on the AI platform squad's batch shadow-inference capacity (TPM/RPM limits) for the shared answer-evaluation engine (blocking).
P4AI Agent Knowledge — Phase 1: Conversation HistoryDesign: Partial — an early design exists in the repo designer; needs a designer revisit. Depends on the AI/ML (Data & AI) team's PII-masking engine, conversation-training service, and training-status webhook that turn past chats into searchable knowledge (blocking).
P5Unified Agent Quality Scorecard — Phase 1: Settings & RubricDesign: Not yet — concept prompts are prepared for the designer. Depends on AI platform (DSAI)'s scoring-engine metric definitions, but this dependency is explicitly non-blocking for Phase 1 per its own PRD's dependency table — it becomes blocking starting Phase 2.
P6Unified Agent Quality Scorecard — Phase 2: Auto-Scoring & In-RoomDesign: Not yet — concept prompt is prepared for the designer. Depends on AI platform (DSAI)/engine team's SkillPack evaluator output contract (blocking).
P7Unified Agent Quality Scorecard — Phase 3: Analytics ReportDesign: Partial — the dashboard is built in the analytics tool (Metabase reuse); the report host page/nav entry around it still needs design. Depends on Data/BI (Data / Analytics)'s unified Metabase dashboard + warehouse pipeline (blocking).
P8Mekari Action (AI Agent)Backlog-tier (2026-07-02 append).
P9Native Integration (AI Agent)Backlog-tier append (2026-07-02). Scoped Sprint-1 pull-in only — Google Calendar OAuth/booking + Google OAuth app-verification submission; portfolio rank unchanged (see sprint 01).
P10Qontak Action (AI Agent)Backlog-tier (2026-07-02 append).
P11AI Agent Idle ActionBacklog-tier (2026-07-02 append).
P12AI Agent MonitoringBacklog-tier (2026-07-02 append).
P13Bot-vs-Human Traffic Split (A/B)Backlog-tier (2026-07-02 append).
P14Multi-WABA ImpactBacklog-tier (2026-07-02 append).
P15Meta Business Agent IntegrationBacklog-tier append (2026-07-12), placeholder rank pending PM confirmation. Draft: Phase 1 (MBA-aware foundation) ungated; Phases 2–3 gated on MBA general availability + Indonesia service-message pricing.
P16AI Spam GatekeeperBacklog-tier append (2026-07-16), placeholder rank pending PM confirmation. Draft motivated by Meta's 2026-10-01 per-service-message billing change (see bifrost/meta-whatsapp-pricing-oct-2026, Epic BIF-8755).
P17Automation Flow BuilderBacklog-tier append (2026-07-17), placeholder rank pending PM confirmation. Draft: desk-validated GO (vs Sleekflow/Respond.io/WATI/Trengo/Gallabox), pre-PRD — repositions the bot tree into a unified conversation+automation engine; core risk is a live-traffic runtime migration needing an engineering feasibility spike before write-prd.
P18AI Agent MemoryCEBEBacklog-tier append (2026-07-22); 2026-Q4 target (only Q4 row — ranked last by quarter, not importance). ANCHOR + Phase-1 PRD both READY; hard CEBE + DSAI blockers (see Cross-initiative / External deps below); Design N/A (Phase 1 flag-only).

Adhoc / Interrupt lane

Bucket (Epic)🔴 Expedite (may preempt)📅 Fixed-Date🟡 Standard (pull-in fill)⚪ IntangibleTracker
Chatbot 26Q3 Engineering Initiatives (Epic BOT-4620)A1 (Ruby/Rails Phase 1), A3 (pdfjs-dist SCA remediation)A2 (Rails 8.1 / Ruby 4.0 Phase 2), A4 (subscription flag gating), A5 (Aegis no-response alert → auto fallback), A6 (stuck KB Processing 5h alert)tracker
Chatbot Q3 2026 Code Readiness (Epic BOT-4500)A1 (MCP), A2 (testing), A3 (hooks), A4 (agent skills), A5 (internal tooling)tracker

Stories are classed, not ranked; each bucket's tracker + ## Expedite log is canonical and this table mirrors it. Estimates remain TBD. Open: A3's 🔴 Expedite has no ## Expedite log entry — DRI to authorize retroactively.

Portfolio progress & forecast

Nine of the thirteen BOT initiatives read 0% done with effort_*/forecast TBD (no DRI-confirmed man-days). Four are sized (rfc-task-breakdown-derived, no sign-off yet) — figures in the table below. No implementor is over capacity; see capacity.md.

Headline: AI Agent Testing (P3) overflows 2026-Q3 (p85 2026-10-25) — a real sequencing conflict, per-person schedules in status 2026-07-14. AI Agent Knowledge stays TBD (RFC review HOLD on cross-team blockers).

% done is live for Autonomous AI Agent Phase 2 (51% as of 2026-07-20, see status) and, from Sprint 1's close, for AI Agent Testing — 7% (2.5 / 37.5 man-days as of 2026-07-27, see status); a floor, since 5 merged issues carry no breakdown task-id. Others stay (board-done issues don't yet join breakdown tasks). The Engineering Initiatives bucket reads 1 of 6 done (17%) as of 2026-07-27 (A3 pdfjs SCA remediation BOT-4661 merged, chatbot-fe PR #1556), effort unsized.

Branch Bot Response — Customer/Contact Field is a shipped 2026-Q2 row recorded retroactively on 2026-07-27, excluded from the Q3 narrative — 100% (11/11 stories, DRI-attested bookkeeping, not per-Story code evidence). See status 2026-07-27.

Autonomous AI Agent is At Risk despite BOT-4191 at 80/80 Done (Phase 1): production-deployed but sandbox-gated, with unresolved staging runtime errors blocking GA — see status 2026-06-25. Phase 1 man-days were never recorded, so the % done below covers Phase 2 only.

Initiative% doneEffort (total man-days)p85 forecastConfidence
Autonomous AI Agent51% (12.5 md)24.5 (Phase 2 only)2026-08-18medium
AI Agent Knowledge13TBDmedium
AI Agent Testing37.52026-10-25medium-high
AI Agent Impact Report33.52026-09-18medium
Bot-vs-Human Traffic Split (A/B)TBDTBD
Mekari Action (AI Agent)TBDTBD
Native Integration (AI Agent)TBDTBD
Qontak Action (AI Agent)TBDTBD
Unified Agent Quality ScorecardTBDTBD
AI Agent Idle ActionTBDTBD
AI Agent MonitoringTBDTBD
Multi-WABA ImpactTBDTBD
Meta Business Agent IntegrationTBDTBD
AI Agent MemoryTBDTBD
Branch Bot Response — Customer/Contact Field (2026-Q2, shipped)100% (11 of 11 stories done)TBD (never sized)— (rolled out 2026-07-09)
Chatbot 26Q3 Engineering Initiatives (bucket)17% (1 of 6 stories done)TBDTBD
Chatbot Q3 2026 Code Readiness (bucket)0% (0 of 5 stories done)— (unsized bucket)

QA Lanes

InitiativeLaneBasisAs of
Autonomous AI AgentBreal-time chat flow + multi-org data isolation (Lane B triggers)2026-06-26
AI Agent KnowledgeBthin spec; triggers unknown — defaults to B2026-06-29
AI Agent TestingBno Lane-B trigger; no E2E specs yet — entry bar unmet2026-06-29
Bot-vs-Human Traffic Split (A/B)Brouting/AB experiment; no E2E specs yet — entry bar unmet2026-06-29
Mekari Action (AI Agent)Bsecurity: cross-product SSO & HMAC auth2026-06-29
Native Integration (AI Agent)Bmoney-critical: Midtrans payments; out-of-reach: third-party OAuth2026-06-29
Qontak Action (AI Agent)Bmoney-critical: deal/ticket creation2026-06-29
Unified Agent Quality ScorecardBanalytics/scoring; no E2E specs yet — entry bar unmet2026-06-29
AI Agent Idle ActionBno Lane-B trigger; no E2E specs yet — entry bar unmet2026-06-29
AI Agent MonitoringBreal-time: live supervisor alerts / push2026-06-29
Multi-WABA ImpactBout-of-reach: external Meta WABA platform2026-06-29
Meta Business Agent IntegrationBout-of-reach: external Meta MBA platform / Thread Control API; real-time inbox handover2026-07-12
Chatbot 26Q3 Engineering InitiativesBscaffolded conservative default — production tech-stack work requires a human QA gate once a story is defined2026-07-13
Chatbot Q3 2026 Code ReadinessBconservative validation default; QA Scope and ownership are declared per story when QA participates2026-07-24
AI Agent MemoryBreal-time: memory read on the live agent response path; AI-dependent output (team policy); no E2E specs yet2026-07-22
Branch Bot Response — Customer/Contact FieldBreal-time: CDP fetch on the live conversation path (2–3s SLA); security/PII: read-only customer attributes must not be logged2026-07-27

Lane A: 0 of 16 initiatives (target 20–50% in Lane A). Every classified BOT initiative has a Lane-B trigger and/or no E2E specs yet, so the Lane-A entry bar (100% E2E, spec-mapped) is unmet across the board.

Team policy: initiatives that depend on AI services stay Lane B — AI-powered features are non-deterministic/probabilistic and need human judgment to verify alignment. Targeted automation is still added for specific testcases (e.g. CRUD configuration & functionality, UI changes, hide/show validation, permission validation).

Capacity summary

See capacity.md for the full per-implementor table.

  • No implementor is over capacity. The binding constraint is Baghiz Firdausa (FE) at 0.55×; five of eight active/draft initiatives are unstaffed, so headroom exists once external blockers clear. Per-person figures: capacity.md.
  • Chatbot 26Q3 Engineering Initiatives is intentionally omitted from numeric load: its stories have no named implementor or DRI-confirmed effort yet. Add its allocation only when a specific story is staffed and pulled under the bucket rule.
  • Code Readiness is also omitted from numeric load because its work is unsized. Its current contributors are mirrored in capacity as TBD; QA is added only for stories with QA Scope.

Cross-initiative dependencies

  • AI Agent Memory (P18) is blocked by Customer Event-Based Engine (CEBE) (CDP squad, TF-3302) — the Central DB + Retriever/query layer must be live in staging before Memory's BUILD gate G0; Memory is CEBE's named flagship downstream consumer ("Agentic AI reads CEBE memory — Q4 2026"). Tracked in ai-agent-memory/README.md frontmatter depends_on: [customer-event-based-engine-cebe].

External & cross-squad dependencies

InitiativeBlocking dependencyOwner (squad / vendor)Status
Autonomous AI AgentUpstream refine-skill-pack suggestion-engine endpoint (Phase 2 Refiner)Data / ML🔴
Autonomous AI AgentStructured guardrails array on the AI-service setup-agent contract (+ migration of existing setup agents) — without it Refine cannot split the merged guidance string back into fields; DSAI committed 2026-07-21 to pick it up, no dateAI platform (DSAI)🔴
AI Agent Impact ReportPer-organization activity-log datamart the report reads fromData / BI🔴
AI Agent Impact ReportPer-participant room assignment-source tracking (AAA · round robin · assign by bot · get chat · handover · manual from room · contact assignee) — needed for the unintended-room metric; not stored in any table today (note)Chat squad (Burhanudin Hakim)🟡
AI Agent TestingBatch shadow-inference capacity (TPM/RPM limits) — shared answer-evaluation engineAI platform🔴
AI Agent KnowledgePII-masking engine, conversation-training service + training-status webhookAI/ML (Data & AI) team🔴
Unified Agent Quality ScorecardSkillPack engine evaluator output contract (Phase 2 Auto-Scoring)AI platform (DSAI) / engine team🔴
Unified Agent Quality ScorecardUnified Metabase dashboard + warehouse pipeline (Phase 3 Analytics Report)Data / BI (Analytics)🔴
AI Agent MemoryCEBE Central DB + Retriever/query layer live in staging (TF-3302) + this phase's consumer requirements landed in the CEBE RFC (query ≤1–2s, DSAI auth, AI-consumer PII allow-list)CDP squad (Task Force)🔴
AI Agent MemoryCEBE retriever client + context injection + memory telemetry in the response envelope — signed contract is BUILD gate G0AI platform (DSAI)🔴