Sales

Pilot wave 1 (M5). Impact: high. See the hub for the shared rhythm.

Current state (assumptions to validate)

Sales at PRIME PRODUCTS is segment-based quoting under time pressure. A vessel calls at Piraeus for hours or a few days; the agent or owner sends a requisition list (often an Excel or PDF with IMPA codes, sometimes free-text item descriptions in mixed English); sales must turn it into a priced, stock-checked quote fast enough to win the call, then convert to order before the sailing deadline. Alongside vessel provisioning, the team handles safety equipment/PPE sales (including the “Protect & Defend” e-commerce channel), UKHO Admiralty chart and e-navigation sales, and formal tenders for defense and public-sector customers — large document packages with compliance matrices and strict submission deadlines.

Assumptions — to validate in discovery: team size ~10–15 across segments; quoting done in SoftOne with heavy Excel side-work for IMPA-list matching and price adjustments; customer communication almost entirely by email (Outlook); tender assembly is manual document collation; no CRM beyond SoftOne customer records; product knowledge for ~tens of thousands of SKUs lives in a few senior heads. Pain points: slow RFQ turnaround on long requisition lists, error-prone manual IMPA-to-SKU matching, lost quotes with no follow-up loop, tender-response effort concentrated on one or two people.

Target AI-first operating model

By M12 an incoming requisition list is parsed by an agent within minutes: items matched to SoftOne SKUs via IMPA code and description, price and stock pulled live, a draft quote produced for the salesperson to review, adjust margins, and send. Free-text and ambiguous lines are flagged for human resolution — the salesperson works the exceptions, not the whole list. Tender teams start from an extracted compliance matrix and AI-drafted response sections grounded in the certification and past-tender library. Every quote gets an automated follow-up draft if unanswered. Humans own pricing decisions, customer relationships, tender sign-off, and all sending.

AI use cases

Use case	Pain addressed	Data needed	Complexity	Impact	Pilot
RFQ-to-quote drafting agent (parse requisition, match SKUs, pull SoftOne price/stock, draft quote)	Slow turnaround, manual re-keying	SoftOne items/prices/stock, IMPA catalog, historical quotes	H	H	Y
IMPA-code product matching from vessel requisition lists (incl. fuzzy description matching)	Error-prone manual matching across huge SKU range	Product catalog with IMPA mappings, item descriptions EN/GR	M	H	Y
Tender compliance-matrix extraction from tender documents	Manual reading of 100+ page tender packs	Tender PDFs, certification library, past tenders	M	H	N
Tender response section drafting grounded in past submissions	Effort concentrated on few people	Past tender responses (SharePoint), ISO certificates, company profile docs	M	M	N
Quote follow-up and lost-quote analysis drafts	No systematic follow-up loop	Quote status in SoftOne, sent-mail history	L	M	N
Customer briefing pack before a vessel call or visit (history, open items, credit status)	Prep time, fragmented knowledge	SoftOne customer/order history, email history	L	M	N
Chart/e-nav quote assistance (folio and edition lookup)	Specialist knowledge bottleneck	Admiralty catalog data, licensing terms	M	L	N

Process transformation opportunities

• RFQ intake: from inbox-and-Excel to a single intake pipeline (shared mailbox → parsing agent → draft quote in SoftOne), with exception queue instead of full manual processing.
• Quote follow-up: from ad hoc to a scheduled, agent-drafted follow-up cadence with win/loss capture.
• Tender production: from document scramble to a maintained tender library (certificates, boilerplate, past answers) that agents draft from.
• Pricing consistency: codify margin rules per segment/customer so the drafting agent applies them and flags deviations, rather than each salesperson improvising.

Required data sources

• SoftOne: items, price lists, stock per warehouse, customers, quotes/orders history (module coverage — assumption, to validate).
• IMPA catalog and internal SKU↔IMPA mapping (likely Excel today).
• Outlook shared/individual sales mailboxes; SharePoint tender and certificate folders.
• Historical quotes and won/lost outcomes (partly Excel).
• Admiralty product catalog for the navigation segment.
• E-commerce product data for PPE.

Potential AI agents

• Quote Drafting Agent — turns a requisition list into a draft SoftOne quote; human approves pricing and sends.
• Product Matching Agent — maps requisition lines to SKUs (IMPA + fuzzy description); human resolves flagged lines.
• Tender Analysis Agent — extracts requirements/compliance matrix and drafts response sections; bid owner approves every section.
• Quote Follow-up Agent — drafts follow-up emails on aging quotes; salesperson approves each send.

Automation opportunities

• n8n flow: shared RFQ mailbox → attachment extraction → Product Matching Agent → draft-quote notification in Teams.
• Scheduled weekly report: open quotes by age/value per salesperson.
• Tender-deadline calendar automation from tender-portal notifications.

Required integrations

• SoftOne read (items, prices, stock, customers) and quote-write — see integrations.
• M365: Outlook (shared mailbox), SharePoint (tender library), Teams notifications.
• E-commerce platform product/stock sync (read).

KPIs

Aligned with the KPI framework (adoption / efficiency / business layers).

KPI	Baseline	M12 target
Median RFQ-to-quote turnaround (vessel segment)	TBD (discovery M2)	−50%
Requisition lines auto-matched without human correction	TBD	≥80%
Quotes with follow-up within 48h	TBD	≥90%
Quote win rate (vessel segment)	TBD	+3 pp
Sales assistant weekly active users	0	100% of team
Tender response prep effort (person-days per tender)	TBD	−40%

Risks

• Wrong SKU match reaches a customer quote → mandatory human review of flagged lines; confidence thresholds; no auto-send.
• Margin/pricing errors from stale SoftOne data → live reads only, price-list freshness checks.
• Senior salespeople bypass the assistant (“faster by hand”) → co-design pilot with the top quoter; measure and publish turnaround gains.
• Confidential customer pricing exposure across the team → role-scoped access to price data in the assistant.

Training needs

• All sales staff: assistant basics, RFQ pilot workflow, exception handling (M4–M5).
• Tender team: tender-analysis workflow, grounding and citation checking (M7).
• Sales management: KPI dashboard, pipeline review with AI outputs (M8).

Deliverables

• RFQ-to-quote pilot (M5) and production workflow (M9).
• Curated IMPA↔SKU mapping dataset (with BI, as RAG/matching source).
• Tender library structure on SharePoint + Tender Analysis Agent.
• Quote follow-up automation and sales KPI dashboard.

12-month execution milestones

Month	Milestone
M1–M2	Discovery interviews; RFQ samples collected; quoting process mapped; baselines measured
M3	IMPA↔SKU mapping data assessed and cleanup plan agreed
M4	Assistant onboarding; sales staff trained
M5	Wave-1 pilot: RFQ-to-quote drafting on one segment (vessel provisioning)
M6	Pilot evaluation; matching accuracy tuning; go/no-go for expansion
M7	Expand to all segments; tender-analysis pilot on one live tender
M9	Quote Drafting + Product Matching Agents in production; follow-up automation live
M11	Win/loss analytics with BI; pricing-rule codification review
M12	KPI review vs targets; handover to line management