Epic board

Progress against the original plan (roadmap P0–P12 + pre-program) · refreshed by the board-refresh agent · last refresh 2026-07-15

7 active · 0 done

Foundations

PRE — Commercial & kickoff baseline pre-M1 (Jul–Aug 2026)

Stage-1 offer drafted, priced and discounted; awaiting TPL C-suite review, then issue to Prime Products

in progress 70%

P0 — Preparation & governance baseline M1

Charter, governance model and board outline drafted; CEO sign-off pending

in progress 20%

P1 — Discovery & requirements baseline M1–M3

Method + 5 questionnaires ready; REQ-001…014 captured; pilot clarification call held (cylinders dept) validates the interview approach — general department discovery not yet started

in progress 20%

P2 — Process map & AI opportunities baseline M2–M4

Method defined in discovery framework; waiting on discovery

not started 0%

Platform

P3 — Infra architecture baseline M2–M3

Technical blueprint drafted (assumption-flagged); site survey pending

in progress 10%

P4 — Hardware procure & deploy baseline M3–M6

Gated by board budget approval (target end M2) and HQ server-room survey

not started 0%

P5 — LLM stack implementation baseline M4–M7

Stack chosen in blueprint (vLLM, Qdrant, n8n; Llama-3.3-70B + Krikri-8B)

not started 0%

Knowledge & Visibility

P6 — Obsidian KB design & rollout baseline M2–M9

Vault live with conventions and program docs; raw-source layer + automated ingest pipeline shipped (ADR-012/013); first Prime Products department profile captured (cylinders)

in progress 15%

P7 — Mission Control site baseline M1–M12

Live ahead of baseline at vanos.tpl.one behind Access — nav regrouped into dropdowns as page count grows; still doubles as the pilot's working dashboard, plus docs, graph (growth-replay timeline), tasks, wiki-chat, board

on track 50%

Value Delivery

P8 — Department pilots baseline M6–M8

Pilot offer priced, pending TPL C-suite review; first sample order-email batch (6 of 20–50) received and fully distilled, revealing the four-tier order-brokerage chain and a hard mailbox-ingestion constraint — still awaiting the team-details handoff and remaining sample volume

in progress 25%

P9 — Agents & automation baseline M7–M10

Use-case pipeline + agent registry drafted

not started 0%

P10 — Training & change mgmt baseline M4–M11

Change/training plan drafted

not started 0%

P11 — Full rollout baseline M9–M11

not started 0%

P12 — Optimization & governance baseline M11–M12

not started 0%

Deviations from the original plan

2026-07-03 Cost (PRE) Stage-1 fixed fee changed €69,500 → €59,075 (15% program-launch discount), payment split 40/60. Source: offer TPL-PRIME-2026-001.
2026-07-03 Scope (P7) Wiki-chat added — conversational access to the knowledge base on the Mission Control site (REQ-011/ADR-008: Bedrock Opus 4.6, no retrieval DB, OKD-hosted). Not in the original WBS; delivered same day.
2026-07-03 Timing (P7) Mission Control delivered ahead of baseline — v1 live pre-M1 (baseline had v0 launching in M1).
2026-07-02 Timing (P6) Knowledge-base vault bootstrapped pre-M1 (baseline start M2) — program docs already live in it.
2026-07-03 Assumption (all) program start month still unconfirmed — M1 ≈ September 2026 pending CEO sign-off; all calendar dates shift with it.
2026-07-06 Timing (P8) Pilot work started pre-M1 — department selected (cylinders/email-order desk), REQ-014 captured and validated, and the CEO clarification call held and ingested. Baseline had pilot department selection at M4 and pilots launching M6–M8.
2026-07-06 Scope (P6) Raw source layer (raw/) and an automated ingest pipeline added (ADR-012/013) — source material (e.g. call transcripts) is committed for provenance and auto-distilled into the wiki via a detect → distill → PR flow. Not in the original WBS.
2026-07-11 Cost (P8) First pilot wave priced — €51,000 fixed fee (15% program-launch discount applied), payment split by phase. Baseline Stage 7 pricing was an unset placeholder. Source: offer TPL-PRIME-2026-002.
2026-07-15 Assumption (P8) First sample order-email batch (6 of 20–50 requested) received and fully distilled — reveals Prime's order-entry process as a four-tier email brokerage chain (customer/shipowner-manager → Prime → manufacturer → port agent) with two deterministic quote↔order match keys, but confirms a hard technical constraint: the M365 search connector cannot expose nested email-attachment content, so automated mailbox ingestion needs full Graph API attachment access (or IMAP/EWS) plus PDF/OCR extraction, not the connector path originally assumed. Source: discovery note.

Rendered from epic-board.md at every deploy — the board changes by commit (nightly agent refresh or /board-refresh on demand), never here. Baseline months are frozen to the 12-month roadmap; M1 ≈ Sep 2026, pending kickoff.