Stage & Offer Structure — PRIME PRODUCTS AI Transformation

Eleven stages spanning the 12-month program (../02-program/roadmap-12-month.md) plus ongoing support. All pricing figures are placeholders (€[X]); all payment splits are indicative pending commercial agreement. Roles only — no named individuals.

Commercial principles

1. Stage-gated. Every stage ends with checkable acceptance criteria signed at a steering gate (../02-program/governance-model.md).
2. Each stage independently valuable. The client keeps a usable asset at every gate; no stage’s value is hostage to buying the next.
3. Pause-able. PRIME PRODUCTS may pause between stages without penalty; TPL re-baselines the roadmap on resumption.
4. Hardware is pass-through. Recommended stance: TPL procures on PRIME PRODUCTS’ behalf at cost with full invoice transparency — no TPL margin on hardware. TPL’s value is selection, integration, and operation (Stages 3 and 6), which keeps incentives clean.
5. Fixed fee for bounded stages, retainer for ongoing work. Change requests via the decision log, priced separately.

Summary table

#	Stage	Months	Duration	Pricing model	Key deliverable
1	Discovery & requirement analysis	M1–M2	2 mo	€[X] fixed fee	Validated requirements register + process inventory
2	Strategy & roadmap design	M2–M3	1 mo	€[X] fixed fee	12-month roadmap, KPI framework, governance model
3	Technical architecture & infrastructure planning	M2–M3	1.5 mo	€[X] fixed fee	Approved architecture + hardware bill of materials
4	Obsidian knowledge base setup	M3–M4	2 mo	€[X] fixed fee	Live vanos-brain vault, seeded, with templates & governance
5	Mission Control website setup	M3–M4	1 mo	€[X] fixed fee	vanos.tpl.one live behind SSO
6	AI infrastructure deployment	M4–M5	2 mo	€[X] fixed fee + hardware pass-through	On-prem LLM platform in production
7	Department pilots	M5–M7	3 mo	€[X] fixed fee (per pilot wave)	3–4 department pilots with measured results
8	AI agents & automation implementation	M6–M10	5 mo	€[X] per agent tier / fixed per wave	Production agents per registry
9	Training & adoption	M5–M11	continuous	€[X] fixed fee (program)	Trained user base, champions network
10	Full rollout & optimization	M8–M12	5 mo	€[X] fixed fee	All 15 departments live; KPI targets met or accounted
11	Ongoing support & continuous improvement	M13+	ongoing	€[X]/month retainer	SLA-backed operation + improvement cadence

---

Stage 1 — Discovery & requirement analysis

• Objective: understand PRIME PRODUCTS as it actually works — needs, processes, systems, data, pain points — and produce a validated requirements base for everything downstream.
• Scope: all 15 departments, all 3 sites (Piraeus HQ, Akti Miaouli, Perama); SoftOne ERP and M365 landscape review; interview program; process inventory; data source inventory.
• Activities: structured interviews per department; process walk-throughs (incl. warehouse/logistics floor time); systems & data audit; requirement capture per ../03-discovery/requirement-analysis-framework.md; validation workshops.
• Deliverables: requirements register (../../02-Requirements/_index.md); process inventory; data source inventory; discovery synthesis report.
• Timeline: M1–M2, 8 weeks.
• Client participation: CEO 2 h/wk; each department head 3–4 h total (interview + validation); IT lead 4 h/wk; a named coordinator (PMO counterpart) 8 h/wk.
• Acceptance criteria: ≥90% of identified departments interviewed; requirements register with ≥1 validated requirement per department, each with owner and status; process inventory sign-off by department heads; synthesis report accepted by steering.
• Dependencies: signed engagement; interview access granted; NDA in place.
• Optional add-ons: customer/supplier interviews; deep data-quality audit of SoftOne.
• Pricing: €59,075 — fixed fee, priced 2026-07-03 in Offer TPL-PRIME-2026-001 (15% program-launch discount off the computed €69,500). Payment: 40% on start / 60% on acceptance.

Stage 2 — Strategy & roadmap design

• Objective: turn discovery into a committed, CEO-owned 12-month plan with measurable outcomes.
• Scope: transformation strategy; 12-month roadmap; KPI framework; governance model; risk baseline; use-case portfolio prioritization.
• Activities: prioritization workshops (value × feasibility); roadmap drafting; KPI definition with reporting/BI; governance & steering design; risk workshop.
• Deliverables: ../02-program/roadmap-12-month.md; ../02-program/kpi-framework.md; ../02-program/governance-model.md; ../02-program/risk-register.md baseline; prioritized ../08-ai-portfolio/ai-use-case-pipeline.md.
• Timeline: M2–M3, 4 weeks (overlaps Stage 1 tail).
• Client participation: CEO 3 h/wk (this stage is CEO-led by design); steering members 2 h/wk; department heads one 2-h prioritization workshop each cluster.
• Acceptance criteria: roadmap approved by CEO at steering; KPI framework with baseline values or explicit baselining plan; governance model ratified (roles filled); top-10 use cases prioritized with sponsors.
• Dependencies: Stage 1 accepted.
• Optional add-ons: board-level strategy briefing pack; competitor/market AI scan for the maritime supply sector.
• Pricing: €[X] — fixed fee. Payment (indicative): 50% on start / 50% on acceptance.

Stage 3 — Technical architecture & infrastructure planning

• Objective: a build-ready design for the on-prem AI platform (~50 daily users, Greek + English) and its integration landscape.
• Scope: platform architecture (LLM serving, RAG pipeline, auth via Entra ID, monitoring); hardware sizing & bill of materials; network/security design with PRIME PRODUCTS IT; SoftOne and M365 integration approach; data protection review.
• Activities: sizing workshops; model evaluation plan (Greek-language capability being the critical test); vendor/hardware option comparison; security review with IT; architecture documentation.
• Deliverables: ../05-technical/architecture-overview.md; ../05-technical/rag-and-data-pipeline.md; hardware BoM with pass-through quotes; integration design; security & data-protection note.
• Timeline: M2–M3, 6 weeks (parallel to Stage 2).
• Client participation: IT lead 6 h/wk; infrastructure/network staff 4 h/wk; compliance officer 2 h total (data protection review).
• Acceptance criteria: architecture approved by steering and PRIME PRODUCTS IT; BoM approved for procurement; security review signed by IT and compliance; Greek/English model evaluation criteria agreed.
• Dependencies: Stage 1 data/system inventory; IT access for assessment.
• Optional add-ons: disaster-recovery design; extended penetration-test scoping.
• Pricing: €[X] — fixed fee. Hardware itself: pass-through at cost (procured in Stage 6). Payment (indicative): 50% on start / 50% on acceptance.

Stage 4 — Obsidian knowledge base setup

• Objective: stand up vanos-brain as the company’s knowledge system and the curated core of RAG.
• Scope: vault per ../06-knowledge-base/obsidian-vault-blueprint.md; Gitea + sync setup; template library; governance (knowledge manager, champions); seeding with discovery outputs.
• Activities: vault + git infrastructure setup; template implementation (../06-knowledge-base/obsidian-template-library.md); migration of Stage 1 outputs (processes, glossary, org/site profiles); knowledge manager onboarding; first two department hubs built as exemplars.
• Deliverables: live vault with folder structure, templates, tag registry; seeded content (process inventory, glossary ≥50 terms, company profile); governance pack; orphan-report automation.
• Timeline: M3–M4, 6 weeks.
• Client participation: knowledge manager (newly named) 8 h/wk; 2 exemplar-department champions 3 h/wk; IT 2 h/wk (Gitea hosting).
• Acceptance criteria: vault operational on company infrastructure with git sync working for ≥5 users; all templates installed and used at least once; ≥100 seeded approved notes; knowledge manager can run the weekly quality cycle unaided.
• Dependencies: Stage 1 outputs; knowledge manager named; Gitea host available.
• Optional add-ons: bulk migration of legacy SharePoint/Word documentation; Greek/English glossary expansion sprint.
• Pricing: €[X] — fixed fee. Payment (indicative): 40% on start / 60% on acceptance.

Stage 5 — Mission Control website setup

• Objective: vanos.tpl.one live as the program’s single pane of glass, from the earliest possible moment.
• Scope: static site per ../07-mission-control/mission-control-blueprint.md; SSO via Entra ID; CI/CD from the vault repo; initial page set per ../07-mission-control/mission-control-page-specs.md.
• Activities: site build (Astro), proxy/SSO configuration, CI pipeline, data-file schemas for KPIs/readiness/hardware, page implementation, mobile QA with the CEO’s actual device.
• Deliverables: live site behind SSO; content pipeline (commit → live); all Phase-1 pages (dashboard, roadmap, risks, decisions, requirements, KPIs); editor guide for the project team.
• Timeline: M3–M4, 4 weeks (parallel to Stage 4).
• Client participation: IT 2 h/wk (Entra app registration, group claims); CEO 1 h (dashboard usability review); steering 1 h (access-tier sign-off).
• Acceptance criteria: site reachable at vanos.tpl.one with Entra SSO; role tiers enforced (board tier cannot see budget page); commit-to-live under 10 minutes; CEO signs off dashboard readability on mobile.
• Dependencies: Stage 2 artifacts to display; Entra tenant access; DNS for tpl.one.
• Optional add-ons: additional page waves (agents tracker, readiness matrix) ahead of schedule; automated weekly email digest.
• Pricing: €[X] — fixed fee. Payment (indicative): 50% on start / 50% on acceptance.

Stage 6 — AI infrastructure deployment

• Objective: the on-prem LLM platform in production — serving ~50 daily users in Greek and English, fed by the curated vault corpus.
• Scope: hardware procurement (pass-through), installation, platform stack deployment (LLM serving, RAG pipeline, chat UI, SSO, monitoring), vault ingestion hook, model selection per Stage 3 criteria, burn-in.
• Activities: procurement & delivery management; rack/burn-in; stack deployment; Gitea post-receive → ingestion pipeline; Greek/English model evaluation & selection; security hardening; load test at 50-user profile; pilot-user smoke test.
• Deliverables: production platform; ingestion pipeline live against vanos-brain; monitoring dashboards; operations runbook; model evaluation report.
• Timeline: M4–M5, 8 weeks (hardware lead time is the critical path).
• Client participation: IT lead 6 h/wk; facilities for rack/power/cooling as needed; 10 pilot users 1 h each (smoke test).
• Acceptance criteria: platform passes 50-concurrent-user load test; RAG answers cite vault notes; only status: approved notes retrievable; SSO + sensitivity filtering verified; uptime ≥99% over a 2-week burn-in; runbook accepted by IT.
• Dependencies: Stage 3 accepted; Stage 4 vault seeded (corpus to ingest); hardware delivered.
• Optional add-ons: GPU headroom for future fine-tuning; DR replica; extended observability stack.
• Pricing: €[X] — fixed fee (services). Hardware: €[X] pass-through at cost, invoiced on procurement. Payment (indicative): services 30% on start / 40% on platform-live / 30% on acceptance; hardware 100% on order (pass-through).
• Recommended stance restated: no TPL margin on hardware — transparency here buys trust for everything else.

Stage 7 — Department pilots

• Objective: prove value fast in 3–4 departments with measured before/after results, generating the pull for full rollout.
• Scope: pilot wave in departments chosen at Stage 2 prioritization (candidates: sales quoting, customer service, procurement, compliance documentation); assistant workflows + first prompt libraries; KB hub completion for pilot departments.
• Activities: per pilot — baseline measurement, workflow design, prompt library, hub + SOP completion, 4-week supervised use, results measurement, go/no-go for scale.
• Deliverables: per-pilot results report (time saved, quality, adoption); department prompt libraries in vanos-brain; pilot playbook for rollout reuse; updated ../08-ai-portfolio/ai-use-case-pipeline.md.
• Timeline: M5–M7, 12 weeks (staggered starts).
• Client participation: per pilot department — head 2 h/wk, 3–5 pilot users 3 h/wk of in-flow usage plus 1 h/wk feedback; champion 4 h/wk.
• Acceptance criteria: each pilot has a baseline and a measured result against it; ≥70% of pilot users active weekly by pilot end; at least 2 of the pilots meet their KPI improvement target; go/no-go decision recorded per pilot in the decision log.
• Dependencies: Stage 6 platform live; pilot departments’ KB hubs at agreed coverage.
• Optional add-ons: additional pilot department in the same wave; customer-facing pilot (e-commerce PPE content) with extra review controls.
• Pricing: €[X] — fixed fee per pilot wave. Payment (indicative): 40% on wave start / 60% on wave results acceptance.

Stage 8 — AI agents & automation implementation

• Objective: move beyond assistant usage to production agents and automations that remove work, not just speed it up.
• Scope: agents from the prioritized registry (../08-ai-portfolio/ai-agent-registry.md) — candidates: Meeting Scribe, tender-document assembler, provisioning-order intake, report drafter for reporting/BI; each with spec, guardrails, human-in-the-loop points.
• Activities: per agent — specification (agent template), build, guardrail testing, supervised pilot, production cutover, runbook; integration work against SoftOne/M365 where required.
• Deliverables: production agents with specs and runbooks in vanos-brain 07-ai/agents/; registry updated; per-agent metrics wired to Mission Control agent tracker.
• Timeline: M6–M10, waves of 2–3 agents.
• Client participation: per agent — business owner (role) 2 h/wk during build/pilot; IT for integration credentials/API access as needed; affected users for supervised-run feedback.
• Acceptance criteria: per agent — passes its guardrail test suite; supervised run with zero unhandled critical failures over 2 weeks; business owner signs the runbook and human-in-the-loop design; metrics visible on Mission Control.
• Dependencies: Stage 6 platform; per-agent system access (SoftOne API/exports, M365 Graph); relevant department onboarded.
• Optional add-ons: additional agents beyond the contracted wave (priced per tier: simple / integrated / multi-step); custom SoftOne integration middleware.
• Pricing: €[X] per agent tier, or €[X] fixed per wave. Payment (indicative): 30% on spec approval / 40% on supervised-pilot start / 30% on production acceptance.

Stage 9 — Training & adoption

• Objective: a workforce that uses the platform well by habit — the program’s adoption KPIs are earned here.
• Scope: role-based curriculum (basic AI literacy for all ~109 staff; power-user track; champion track; leadership track); champions network; “document as you answer” habit installation; ISO 21001/29993-aligned training documentation in 11-training/ and 15-compliance/.
• Activities: curriculum design; train-the-champion sessions; department training waves synced to rollout; office hours; adoption nudges via Teams; completion tracking feeding Mission Control.
• Deliverables: curriculum & materials (in vanos-brain, both languages as needed); champions network (1/department); training calendar; completion & proficiency reporting.
• Timeline: M5–M11, continuous alongside Stages 7–10.
• Client participation: HR 2 h/wk (scheduling, records); each employee ~6 h total (basic track); champions 2 h/wk sustained; department heads open/close their department’s wave.
• Acceptance criteria: ≥90% of staff completed basic track; champion active in all 15 departments; training completion visible per department on Mission Control; training documentation meets the company’s ISO doc-control requirements.
• Dependencies: Stage 6 platform (hands-on training needs the real system); Stage 7 playbooks.
• Optional add-ons: e-learning packaging of the curriculum; external certification prep for power users.
• Pricing: €[X] — fixed fee for the program-length training track. Payment (indicative): 30% on curriculum acceptance / 40% at mid-program checkpoint / 30% on completion criteria met.

Stage 10 — Full rollout & optimization

• Objective: every department live on the platform and knowledge system, KPI targets met or honestly accounted for, program hand-over-ready.
• Scope: rollout to all remaining departments using the Stage 7 playbook; KB coverage to target across all 15 hubs; performance/cost optimization of the platform; KPI closure review; program handover pack.
• Activities: department onboarding waves (champion-led, TPL-supported); prompt library expansion; platform tuning (latency, cost/query, model updates); gap-report-driven documentation sprints; final KPI measurement; handover to Stage 11 operating model.
• Deliverables: all 15 departments onboarded; KB coverage report vs targets; optimization report; program closure report with KPI results; handover pack (runbooks, credentials process, governance docs).
• Timeline: M8–M12, 20 weeks overlapping Stages 8–9.
• Client participation: each remaining department head 2 h/wk during their wave; champions 4 h/wk during their wave; IT 4 h/wk; steering monthly gates.
• Acceptance criteria: ≥40 of ~50 target daily active users sustained over the final month; all 15 department hubs live at agreed coverage thresholds; program KPI scorecard published on Mission Control with met/missed accounting; handover pack accepted by IT and the knowledge manager.
• Dependencies: Stages 6–9 in flight; department availability (respecting seasonal peaks in ship supply).
• Optional add-ons: second-year roadmap design; advanced analytics/BI integration phase.
• Pricing: €[X] — fixed fee. Payment (indicative): 30% on start / 40% at mid-rollout gate / 30% on program acceptance.

Stage 11 — Ongoing support & continuous improvement

• Objective: keep the platform healthy, the knowledge fresh, and the improvement loop turning after program close.
• Scope (retainer): platform operations SLA (monitoring, patching, model updates, incident response); knowledge-governance support to the knowledge manager; improvement cadence (monthly review of gap reports and ../08-ai-portfolio/ai-use-case-pipeline.md backlog); quarterly steering review; agreed bucket of improvement hours.
• Activities: monthly ops report + improvement review; quarterly model/benchmark refresh; on-call per SLA; small enhancements from the bucket; annual SOP-review support.
• Deliverables: monthly ops & improvement report on Mission Control; SLA compliance record; maintained runbooks; quarterly roadmap refresh note.
• Timeline: M13 onward, 12-month renewable term, 3-month notice.
• Client participation: IT 2 h/wk; knowledge manager 4 h/wk (their ongoing role, TPL-supported); steering 2 h/quarter.
• Acceptance criteria (per quarter): SLA met (uptime, response times as contracted); improvement bucket consumed against prioritized backlog with sign-off; quarterly review held and minuted.
• Dependencies: Stage 10 handover accepted.
• Optional add-ons: higher SLA tier (extended hours); dedicated improvement sprints; new-department or new-company (group) extension.
• Pricing: €[X]/month retainer (base SLA) + €[X]/month improvement-hours bucket. Payment (indicative): monthly in advance; improvement hours reconciled quarterly.

---

All fees €[X] and payment splits are placeholders, indicative only, and subject to the master services agreement. Hardware, licenses, and third-party services are pass-through at cost unless explicitly stated otherwise.