source: .forge/discovery/meeting-notes/2026-07-22-kickoff-debrief.md · generated — edit the markdown, not this file

Empire Teas CRM — Pre-Kickoff Question Responses: Debrief

Date: 22 July 2026 Source: Client responses to the “CRM & AI Initiative — Pre-Kickoff Questions” document (sent to Empire Teas, 22 July 2026) Purpose: Map every client answer against the live PRD open questions (project-prd-signals.md), explain what each answer means, and flag which ones need a real follow-up conversation before we can lock designs around them.

This file is the audit record for the 22 July response round. Update — 2026-08-02: the confirmed answers have since been folded into project-prd.md (Rev 4) and the open items recorded in project-prd-signals.md as ◐ partial · ASSUMED; ERP and AI-hosting were settled by manager decision on 2026-07-28 (see Part 1 items 2 and 5). The PRD + signals are now the authoritative status; this debrief is retained as the record of the response round, annotated below where an item’s status has since changed.


Part 1 — Follow-up conversation needed, ranked

1. RESOLVED — “Auto or Manual” review option (OQ-16). The client wrote: “Correct / Also its better to have a selection option (Auto or Manual).” Originally ambiguous: “Auto” could mean routing work to reviewers (harmless) or AI acting without human approval (which would break the locked governance invariant “every AI action is a reviewable draft approved by a human”). Settled 2026-07-28: the 28 July closing document states, as a fixed note, that every AI-drafted action always requires human approval; the Auto/Manual choice governs how work is routed to a reviewer, not whether approval happens. Recorded as assumption OQ-16 (◐ partial); the human-approval invariant is intact.

2. RESOLVED (accepted) — ERP (OQ-2). Originally flagged as a scope unknown: “start with the current ERP” was a real decision, but we didn’t know which ERP it was or whether it had an API. Settled by manager decision 2026-07-28: the current ERP is IFS. We integrate the CRM→ERP handoff against IFS assuming its APIs are available, and fall back to manual ERP entry if they are not available by integration start. The new ERP (TBD, targeted ~2027) is deferred. Risk R-01 → accepted; not asked in the 28 July closing document.

3. “Per client and per standard” (OQ-42) — too terse for the most load-bearing pricing decision. The document itself calls this “the single most load-bearing element of the pricing model,” and the answer is four words. Best-guess read: “per standard” = priced per grade (confirms grade as a required price-key axis), “per client” = genuine buyer-specific list prices, not just a quote-time discount. If that reading is right, it’s a bigger schema than currently designed — the current model only keys on product × destination market, with no buyer dimension. Needs an explicit confirmation before the F26 schema is finalized.

4. Certifications (OQ-9) — only half the question was answered. “Spreadsheet” tells us where buyer-required certs are tracked, but the other half — which certifications each Empire product/grade actually holds — is the literal prerequisite the PRD states the cert-mismatch feature “cannot be built without.” We should ask for the actual spreadsheet rather than re-ask the question in prose.

5. RESOLVED (accepted) — external AI / data privacy (OQ-15). Originally the highest open architectural risk (a “no” would force a self-hosted-model rebuild of the entire AI layer). Settled by manager decision 2026-07-28: no on-prem/self-hosted LLM; AI runs on secure, reputable cloud infrastructure (Anthropic Claude), with the specific data-handling details presented for the client’s formal approval at the contract stage. Risk R-02 → accepted; the self-hosted-model spike (SP-ARCH-001) stands down unless contract approval fails. Not asked in the 28 July closing document.


Part 2 — Section-by-section walkthrough

A. Systems & data privacy

# OQ Answer What it means Status
1a OQ-2 “Target Jan 2027” (new ERP) Resolved (accepted), 2026-07-28. Current ERP confirmed as IFS (manager decision). New-ERP name/timeline (~2027) deferred. Accepted — R-01
1b OQ-2 “Yes. We can start with the current ERP” Resolved (accepted), 2026-07-28. Integrate against IFS assuming APIs available; manual-entry fallback if not. Accepted — R-01

B. People & roles

# OQ Answer What it means Status
2 OQ-3 “Correct” Confirms the four roles (Sales Rep, Sales Manager, R&D/tea taster, Admin), including that tea tasters do use the system. Minor ambiguity: unclear if “Correct” also confirms the optional executive read-only view, or just the core four. Resolved, one minor clarifying ping worth sending
3 OQ-16 “Correct / Also better to have Auto or Manual selection” See Part 1, item 1. Follow-up needed — high priority
4 OQ-39 “Tea room team setting” A specific team (“Tea Room”) owns weekly pricing, not individual reps. Worth confirming whether “Tea Room” is the same group as the assumed R&D/tea-taster role or needs its own permission slice. Mostly resolved, small role-mapping check
4 — “Can extract from the Tea auction API” New information: there may be a real auction API available. Worth a dedicated follow-up — this could upgrade our currently “deferred, manual-only” auction-reference feed into something automatable much sooner. See combined note with Q10 below. New — needs scoping, not just logging

C. Pricing

# OQ Answer What it means Status
5 OQ-33 “Until the next auction / should have an option to update manually for exceptional cases” The question explicitly asked whether a quote’s validity is the same clock as the price or a separate one; the client gave one unified answer, not two, and no 30-day figure at all — read as one clock for both: a quote is valid until the next auction, same as the published price. This drops our original ~30-day quote-validity assumption. It also fits the domain better: a 30-day-old quote would be dangerously stale against a weekly-moving auction baseline. The manual-override request matches a “refresh price” action already planned in the pricing research — no new work there. Resolved — single clock, quote validity = until next auction
6 OQ-42 “It’s Per client and per standard” See Part 1, item 3 — the most consequential ambiguity in the whole batch. Follow-up needed — highest priority
7 OQ-43 Working-assumption text highlighted (confirmed, not written out separately) Confirms the assumption as stated: bulk is quoted FOB Colombo in USD, with other currencies handled as display conversions only. Currency is not a separate price-key axis — no need for per-currency maintained price rows. Resolved
8 OQ-44 “Depends on the client, mostly FOB” Confirms FOB Colombo as the default, but CIF genuinely happens for some clients — not purely a display conversion. Current design (FOB base + per-quote cost-build-up lines) already accommodates this, so low risk. Resolved, no schema change needed
9 OQ-45 “Yes” Confirms per-kg price does vary with order size, but doesn’t say whether that’s formal published tiers (a real schema dimension) or rep discretion (handled as the existing discount field). This changes the schema either way. Follow-up needed
10 OQ-46 “Online Auction system” Prices are recorded/communicated via an online auction system today, not Excel/memory. Combined with the Q4 mention of a “Tea auction API,” this is worth a dedicated conversation — it could mean real migration-seed data and even a live price feed are available, not just manual entry. Resolved as answered, but opens a real opportunity worth scoping

D. Quotes, samples & products

# OQ Answer What it means Status
11 OQ-9 “spreadsheet” Only answers the buyer-cert-tracking half. Product→certification mapping (the actual prerequisite) is still missing. Follow-up needed — request the file directly
12 OQ-10 (no answer given) Payment terms and whether they belong on the quote: unanswered. Not answered — needs re-ask
13 OQ-11 (no answer given) Sample-round count: unanswered. Low stakes — system supports unlimited rounds regardless. Not answered — low priority re-ask
14 OQ-20 “Correct” Own-brand goods sell through this pipeline. Product/Deal need a brand attribute — the prototype’s no-brand assumption no longer holds. Resolved — schema addition needed

E. Channels & where the data lives

# OQ Answer What it means Status
15 OQ-14 “individual rep inboxes, and info@empire.lk” Two channels confirmed: individual mailboxes plus a shared address. Worth a quick check on who monitors info@empire.lk for triage routing. Resolved
16 OQ-24 “Microsoft 365” Confirms provider. Combined with Q15, gives the full mailbox picture. Whether Pipedrive’s email sync is already on and whether historical mail is usable for AI context: this is a system-config fact, so it will be confirmed via the Pipedrive audit rather than asked of the client. Resolved — sub-item routed to audit, not a written question
17 OQ-4 “Mostly Excel sheets, WhatsApp threads, personal email” Clear and useful. The earlier concern about whether “personal email” means an accessible M365 account vs. a genuinely private one turned out to be moot: F20’s email integration is per-user OAuth login, not a bulk migration, so the company never needs special access to any account regardless of ownership — each rep just connects their own inbox. No follow-up needed. Fully resolved

F. Three defaults

# OQ Answer What it means Status
i OQ-21 “OK” Migration approach confirmed: migrate everything, preserve IDs, AI-triage after. Resolved
ii OQ-38 (no answer given) Which destination markets get differentiated pricing: not answered. The market list itself will come from the Pipedrive data pull regardless. Not answered — needs re-ask
iii OQ-40 (no answer given) Pricing/margin approval authority (assumed Sales Manager+): not explicitly confirmed. Not answered — needs re-ask

Part 3 — Full open-questions ledger

Everything currently tracked in project-prd-signals.md, so nothing is hidden:

Status OQ IDs What’s happening with them
Resolved by this response OQ-3, OQ-4, OQ-14, OQ-20, OQ-21, OQ-24, OQ-33, OQ-39, OQ-43, OQ-44, OQ-46 Answers land cleanly (minor sub-clarifications noted above, not blocking)
Accepted by decision — not a client question OQ-2, OQ-15, OQ-16 Update 2026-08-02. OQ-2 → current ERP is IFS; integrate assuming APIs, manual fallback (R-01 accepted). OQ-15 → secure cloud AI, no on-prem, approval at contract (R-02 accepted). OQ-16 → human approval on every AI action is mandatory; Auto/Manual governs routing only.
Sent in the 28 July follow-up; assumption in place pending the client’s reply OQ-9, OQ-10, OQ-11, OQ-38, OQ-40, OQ-42, OQ-45, OQ-47 Recorded as ◐ partial · ASSUMED in signals and in the proposal’s Assumptions section; OQ-42 (buyer-specific pricing) is the load-bearing one.
Reclassified to “we’ll find this ourselves” (Pipedrive audit) OQ-1, OQ-5, OQ-6, OQ-7, OQ-8, OQ-12, OQ-13, OQ-17, OQ-22, OQ-23, OQ-25, OQ-26, OQ-27, OQ-28, OQ-29, OQ-30 Deliberately not asked in writing — confirmed 2026-07-27 to be pulled directly from the Pipedrive export/walkthrough, not asked of the client
Deliberately deferred (gates a post-V1 feature) OQ-18, OQ-19 Not blocking anything right now
To be asked live, not in writing OQ-41 The open-ended “what would you miss about Pipedrive” catch-all backing the “lose nothing” promise. Decided 2026-07-27: better answered in conversation than in writing — will be raised directly during the Pipedrive audit walkthrough rather than included in the follow-up document.

Appendix — Full source document, questions and answers exactly as sent/received

Reproduced verbatim from the “CRM & AI Initiative — Pre-Kickoff Questions” document sent to Empire Teas on 22 July 2026, with the client’s answers as returned (shown in bold, inline, immediately after each question). HTML entity artifacts from the original copy (&, ') have been cleaned to plain characters; no wording has been altered.

ALTRIUM · FOR EMPIRE TEAS (PVT) LTD

CRM & AI Initiative — Pre-Kickoff Questions

22 July 2026

Ahead of our project kickoff, this document sets out the questions where we need Empire Teas’ input to design the new CRM and its AI capabilities around the way you actually work. It is longer to read than to answer: many items are one-line confirmations. Wherever we hold a working assumption we have stated it openly — correcting us is just as valuable as confirming.

How to respond. Brief written answers ahead of the kickoff are ideal, even partial ones — whatever remains, we will simply work through together in the session. Some questions may need input from different people (sales, pricing, IT, legal); feel free to circulate the relevant sections.

A. Systems & data privacy

1. Your ERP landscape — current and new We understand you run an ERP today and that a new ERP is on the way. Three parts: a. Which new ERP is planned, on what timeline, and will it expose an API? Target Jan 2027 b. Should we integrate with the current ERP first and move the integration later, or keep a well-managed manual handoff until the new ERP lands? Yes. We can start with the current ERP

Why we ask: This decides whether the handoff of a won deal to your ERP is automated from the start or begins as a clean manual step. Version 1 ships with the manual path either way, so nothing here blocks the early phases.

B. People & roles

2. Who will use the CRM Which teams and user types will actually work in the system? We currently assume four roles — Sales Rep, Sales Manager, R&D / tea taster, and Admin. Should QA/compliance, export-documentation coordinators, or management (read-only) also have access? Correct

Our working assumption — please confirm or correct: The four roles above, plus possibly an executive read-only view. The R&D / tea-taster role in particular needs your validation.

3. Who reviews AI output AI in this system drafts and recommends; people approve. Who should review and approve AI outputs such as triaged leads and drafted replies — one designated person, or each rep for their own records?

Our working assumption — please confirm or correct: Each rep reviews their own, with manager oversight — but this is your organizational call. — Correct / Also its better to have a selection option (Auto or Manual)

4. Who owns the weekly bulk price Who sets and publishes your weekly bulk prices — one central pricing person or role, or does each rep price their own deals? Tea room team setting

Why we ask: This shapes the weekly price-update workflow and who holds pricing permission in the system.

Our working assumption — please confirm or correct: A centralised weekly publish aligned to the Colombo auction cadence — but your actual practice decides. Can extract from the Tea auction API

C. Pricing

These questions let us finalise how prices are structured in the system. They are the most detailed section — ideal for whoever owns pricing day to day.

5. How long a price holds When you set a bulk price, how long does it hold — one week / until the next Colombo auction, around 30 days, or negotiated per deal? And is a quote’s validity the same clock, or a separate one? Until the next auction / Should have an option to update manually for exceptional cases

Our working assumption — please confirm or correct: Two separate clocks: the published price holds roughly one week (auction cadence), while an issued quote stays valid for about 30 days.

6. What exactly carries a price When you set “the price of a bulk tea,” what is it attached to — a specific grade (for example, a Uva BOP1), a broader product family, or is it set ad hoc per deal? And within one destination market, do two buyers ever receive different list prices for the same tea, or is any buyer-level difference handled as a discount at quote time? It’s Per client and per standard

Why we ask: This fixes the single most load-bearing element of the pricing model — what a price row in the system actually points at.

7. Currency Are all bulk prices quoted in USD, or do some markets get EUR, GBP or another currency? If multi-currency: is each maintained as its own price, or converted from the USD number?

Our working assumption — please confirm or correct: Bulk is quoted FOB Colombo in USD, with other currencies as display conversions.

8. What the published price includes Is the weekly price you set an FOB Colombo number, with freight and insurance added per deal on top? Or do you maintain separate CIF prices per destination? Depends on the client, Mostly fob

Our working assumption — please confirm or correct: FOB Colombo, with per-deal cost build-up for freight and insurance — this is what our current design assumes.

9. Volume-based pricing Does the per-kg price change with order size (for example, a trial pallet versus a full container)? If yes — are there formal volume tiers, or is it rep discretion? Yes

Why we ask: Formal tiers become part of the price structure; discretion stays as a discount field at quote time.

10. How weekly prices are maintained today When this week’s auction results come in, how does the new price actually get recorded and communicated today — an Excel sheet built from broker reports, an email, or simply someone’s experience? Online Auction system

Why we ask: Whatever exists today becomes both the seed for migrating price history and our reference for designing the weekly price-publishing workflow.

D. Quotes, samples & products

11. Certifications Which certifications apply to which products? And how is each buyer’s required certification set tracked today — spreadsheet, CRM field, or memory? spreadsheet

Why we ask: The system will automatically flag certification mismatches between a buyer’s requirements and a quoted product — it needs both sides of that mapping, and the answer tells us where to migrate it from.

12. Payment terms What payment terms are standard per market (LC, open account, DA, DP)? Should payment terms appear as a field on the quote, or are they handled entirely in the ERP?

Our working assumption — please confirm or correct: A simple payment-terms field on the quote, with detailed LC administration remaining with the ERP and export team.

13. Sample rounds How many rounds of samples does a typical private-label deal go through before approval?

Why we ask: A quick one — the system supports unlimited rounds either way; your answer calibrates reminders and expectations.

14. Own-brand sales Do your own-branded finished goods (Thurson, Hyson, Akina, Ceylon Wild Harvest) sell through this same B2B pipeline, or does it carry only bulk and private-label business? Correct

Why we ask: A yes/no that decides whether products and deals carry a brand attribute in the data model.

E. Channels & where the data lives

15. Where inquiries arrive Where do buyer inquiries land today — a shared mailbox, individual rep inboxes, the website form, WhatsApp? individual rep inboxes, and info@empire.lk

Why we ask: This determines where the AI inquiry-capture connects first.

16. Mailbox setup Do reps send from individual mailboxes or a shared team mailbox — and on which platform (Gmail, Outlook / Microsoft 365, on-premise Exchange, other)? Microsoft 365

Why we ask: Email integration is standard for the major platforms; anything less common we simply scope a little earlier.

17. Sales data outside the CRM What sales data lives outside Pipedrive today — Excel sheets, WhatsApp threads, personal email, physical files? Nothing needs to be tidy; we just need to know where to look. Mostly Excel sheets, WhatsApp threads, personal email

Why we ask: This is the one thing no system audit can see. It tells us whether the migration scope is complete, so nothing your team relies on is left behind.

F. Three defaults — confirm or adjust

For the following, we propose a sensible default. A one-word “confirmed” is enough; equally, tell us if reality differs.

i. Migration approach We migrate everything from Pipedrive, preserve the original record IDs for auditability, and let AI triage flag junk after migration — rather than dropping records beforehand. Nothing is silently lost. — OK

ii. Destination markets We will derive the candidate market list from your Pipedrive data, then ask you only the part that is genuinely yours to answer: which of these markets do you price differently for the same tea? (Working list: EU–Northern, EU–Western, Middle East, East Asia, North America, Eastern Europe.)

iii. Pricing approval Quote pricing and margin approval sits with the Sales Manager and above. Confirm, or name who actually holds this authority today.

Generated from .forge/discovery/meeting-notes/2026-07-22-kickoff-debrief.md by .claude/lib/render-docs-html.js. Markdown is canonical.