Questions for Empire Teas
Prepared by Altrium for Empire Teas (Pvt) Ltd · first issued 21 August 2026 Companion to the client-facing PRD, the timeline, and the dependency list. Revised 19 November 2026: Q10 — we have chosen how newsletters work, so we can start building them now. Please confirm the choices, or tell us what to change. Revised 17 November 2026: no work waits on these answers any more; each open question is built on its stated assumption, which you can correct at UAT. Older change notes are under Change history at the end. Revised 12 November 2026: Q9 — loading an exhibitor list before a show is now in this version, in Phase 1; the badge photo is still needed.
We have enough detail to start building. These are the points that are not yet 100% clear to us, and where guessing would mean building the wrong thing and changing it later.
The list is deliberately split in two:
- Part 1 — questions only you can answer. These need a decision or a fact from someone at Empire Teas.
- Part 2 — things we will find out ourselves. These are answered from your Pipedrive exports, or asked of you directly where the exports cannot answer them. You do not need to prepare answers, and no Pipedrive access or review session is needed (updated 26 October 2026).
Each question says why we are asking and what we will assume if we do not hear back. The assumptions are safe, sensible defaults — but if an assumption is wrong, correcting it later costs more than telling us now.
Updated 17 November 2026: no work waits on these answers any more. Where an answer has not arrived, we build on the assumption written under the question, and you can correct any of them at UAT. The three once marked urgent (⚠️) are kept marked, so you can see which mattered most.
Part 1 — Questions only you can answer
Who can do what
Q1. What should each of the six roles hold on day one, and who reports to whom? ⚠️
Updated 7 September 2026. Roles are now configurable: an Administrator can create roles and choose their permissions, and the six roles confirmed on 3 August — Sales Rep, Sales Manager, R&D / Taster, Administrator, Tea Room (pricing), Executive / Management — are our recommended first set, not built-in templates (decided 11 September 2026, one role per person, no template layer to learn). Updated 24 September 2026: there are no built-in roles at all. The system starts with one ordinary role, IT Admin, held by one named account; your Administrator signs in with it and creates every other role, with us alongside at go-live. Each area of the system now has two permissions — view and manage — plus a few named permissions that stand alone (publishing prices, approving a below-floor quote, one for each page under Settings — users, roles, the tag pages and email templates, since 30 October 2026 — and your own mailbox). So this question is “what should each role hold on day one”, which you can adjust afterwards. For the first four we have a detailed proposal. For Tea Room and Executive / Management we only have a one-line description.
The specific gaps:
- Tea Room set and publish the weekly bulk prices. Should their role also let them see accounts, deals and quotes — even just to read them — or only the pricing screens?
- Executive / Management are read-only. Read-only across everything, including quotes and customer contact details?
The reporting line adds four questions of its own:
- Who reports to whom? A list of managers and their reports (or an org chart) lets us load the reporting line at go-live. A manager’s “team” view follows this list.
- Does “team” reach all the way down? If a Director’s team includes a Manager, should the Director also see that Manager’s representatives’ records, or only their direct reports’? Decided 24 September 2026: all the way down.
- May a manager edit their team’s records, or only see them? If we do not hear: managers can edit and reassign their team’s deals and leads, but never read a team member’s personal mailbox. Updated 24 September 2026: this is a setting on the role, so either answer works. No role can ever read another person’s mailbox — that is a fixed rule. Updated 27 September 2026: a manager can do on their team’s records exactly what their role allows on their own — with Manage on an area they edit their team’s records; with only View they read them. Seeing the team’s records while editing only your own is no longer offered.
- Records nobody owns. Updated 30 September 2026: in this version every lead has an owner —
inquiries to
info@are forwarded to a rep, who creates the lead (see Q6), and the migration gives every imported lead to one named person, who hands them on. Should a lead ever be left without an owner, it is visible only to roles with the “all records” permission for leads, through an “Unassigned” filter on the lead list; a manager who sees only their team does not see it. Previously (24 September): a migrated Pipedrive lead whose owner had no CRM account would have had no owner. Before that: we proposed such leads be visible to every manager and listed as unassigned until someone takes them. - Who can hand a record to a new owner. Added 30 September 2026: anyone who can manage the record — its owner or a colleague it was shared with (whose role manages that kind of record), a manager above either of them, or a role with “all records” — can hand it over, share it and remove its shares.
- When a manager leaves, should their team move automatically to the manager above, or should the system ask you to reassign them first? Decided 24 September 2026: it asks first — a person cannot be deactivated while anyone still reports to them.
Why we are asking: we are building the permission system now, and it decides what every user sees on every screen. The Who Can Do What companion document (revised 24 and 27 September) carries the full proposal, role by role, with tick boxes — its Section 2 covers the reporting line and sharing, and every assumption changed on 24 or 27 September is marked as such. If we do not hear: Tea Room gets pricing screens only; Executive gets read-only access everywhere. This is advice, not something we need before building: every role is yours to set up and change at any time, so an answer here only tunes the starting set. (Our ref: OQ-53 — closed internally as not required; optional advice)
Q2. How does a new user get added, and what happens when someone leaves? — decided 24 September 2026
Three parts:
- When an Administrator adds someone, should the person get an invitation email, or simply be told they can now sign in? (Updated 4 Sep 2026: sign-in is with each person’s company Microsoft account — there is no CRM password to set, so the earlier invite-to-set-a-password / temporary-password wording no longer applies.) Decided 24 September 2026: no invitation email. Your Administrator adds the person in the CRM and tells them; they then sign in with their company Microsoft account. Nobody can sign in unless an Administrator has added them.
- Can one person hold more than one role? Decided 11 September 2026: no — each person holds one role, and someone doing two jobs (say, Tea Room and Sales Manager) gets a role built to hold both sets of permissions.
- When someone leaves, what should happen to their accounts and deals — pass to their manager, reassign manually, or leave them in place under the closed account? Decided 24 September 2026: reassigned first. The system will not deactivate a person while they still own leads, accounts or deals, or while anyone reports to them; your Administrator reassigns those, and a reassignment screen moves them in bulk. Deactivating a person ends their sessions at once. People are never deleted, only deactivated (and reactivated if they return), so everything they did stays attributed to them. Nobody can deactivate themselves.
Why we are asking: it changes the sign-up screens and how we handle a departing user’s records. Decided (24 September 2026): no invitation email; one role per person; a leaver’s records and reports are reassigned before they are deactivated. Previously: we proposed an invitation email, and that a leaver’s records would stay in place until an Administrator reassigned them. Tell us if you disagree. The live system starts with one account in the IT Admin role; which email address it uses is a new item in What the Project Needs from Empire Teas. (Our ref: OQ-51)
Q3. Who signs off a quote that is below the minimum margin? — please confirm our plan
You told us a quote below the gross-profit floor needs the Managing Director’s approval. Here is how we intend to build exactly that, and we would like a yes or a correction rather than a decision from scratch.
Our plan: the Managing Director has a dedicated account profile. It sees everything an Executive sees — every deal, quote (with its margin), account, published price and cost, read-only — plus the one right that, by default, nobody else has: approving a below-floor quote. By default Administrators and Sales Managers cannot approve, and the Tea Room sets the floor, so in the recommended set-up no single person can lower the floor and then approve a quote against the lower figure. Every approval records the name, email and time of the person who gave it.
Updated 24 September 2026: this is delivered as the recommended set-up rather than as a rule the system enforces. Whoever administers users and roles could give the approval right to another role — their own included — and the system records every such change, with who made it and when. Because who holds the administrator role is therefore the real control, name those people with care; we also recommend you keep at least two, so access can always be managed when one of them is away. Previously: we said Administrators cannot approve.
Two things to settle with it:
- Cover for absence. While the Managing Director is away, below-floor quotes wait. If you want a deputy who can approve during that time, name them — an Administrator can hand the right to a named person for a period, and the change itself is recorded.
- Does the Managing Director also work in the system day to day — for example as a Sales Manager or an Administrator? If so, tell us which role; it decides whether their one role also carries those permissions (one role per person — see Q2).
Why we are asking: it decides who sees the approval button. If we do not hear: the Managing Director alone approves, with no deputy named — and your administrator can change this at any time. (Our ref: OQ-53 — closed internally as not required; optional advice)
Certifications
Q4. Where are the certifications your buying standards hold currently kept? ⚠️
We know that buyer certification requirements are tracked in a spreadsheet. What we still need is the other half: which of your buying standards hold which certifications (Rainforest Alliance, organic, Fair Trade, HACCP and so on) — and where that information lives today. A spreadsheet? The QC team’s own records? Paper certificates in a file?
We also need a sample of the actual data, in whatever form it exists.
Why we are asking: the system is meant to warn a rep when a buyer needs a certification the selected standard does not hold. Without knowing what your standards hold, that check cannot be built. If we do not hear: this feature cannot be built at all — it is the one item on this list with no safe fallback. (Our ref: OQ-9)
Standards and pricing
Q5. Where does your full list of buying standards live today, and can we have a copy?
The complete list of standard numbers with their client, grade and specification — a spreadsheet, your ERP, or somewhere else? (The weekly price workbook carries only the standards priced that week.) And separately — we would like the actual weekly bulk price list in whatever format the Tea Room produces it today.
Why we are asking: there is no separate product catalogue in the new system — every item you sell, bulk or private label, is a buying standard — so this list is what we load on day one, and the real price list is the only way to confirm our pricing design matches how you actually work. If we do not hear: we build against our design and validate later, which risks rework in the pricing area. (Our ref: OQ-27)
Q13. A few details of the weekly price workbook — new, 27 September 2026; narrowed 6 October 2026
You confirmed on 13 September how a weekly price is worked out — the rupee price, converted to dollars at a rate, plus a margin; on 17 September we added an overhead in rupees before the conversion — and that the Tea Room uploads the workbook into the system and can then correct any single price by hand.
On 6 October we settled two of the points we had asked about, so you do not need to answer them:
- The exchange rate and the overhead are stored with each price, so the system works whether the Tea Room uses one rate for the whole week or different rates for different standards.
- The layout stays as it is: the upload accepts the workbook as it looks today, and lists any row it cannot read so the Tea Room can check it. Nobody has to change how the sheet is made.
Two small points are still open, and the Tea Room is probably the right people to ask:
- The unit — are the numbers in the workbook rupees per kilogram?
- Who produces the sheet each week?
Why we are asking: the unit is shown on the upload and price screens, so it must be right; the name tells us who to work with when we test the upload. If we do not hear: we build on rupees per kilogram, show that unit on screen so a mistake is easy to spot, and ask the Tea Room when we test the upload. (Our ref: OQ-59)
Q14. Do private-label standards get a weekly price? — new, 27 September 2026
Every item you sell, bulk or private label, is now a buying standard. Bulk standards get a new price every week. We understand private-label work is priced for each deal on the quote instead. Is that right — no weekly price for a private-label standard? And does a private-label standard carry a buying cost, so that the system can check the quote’s margin against the minimum?
Why we are asking: the system stops a rep quoting a bulk standard that has no price this week. That check must not block a private-label quote. If we do not hear: private-label standards are priced on the quote, with no weekly price, and the check does not apply to them. We will ask again about the buying cost before the margin check is built. (Our ref: OQ-61)
Email and inquiries
Q6. How should the shared inbox work? — decided 24 September 2026
Decided 24 September 2026: info@ stays outside the CRM in this version. Whoever reads info@
today forwards an inquiry to the rep who covers it — in line with what you told us on 9 September about
inquiries going to whoever looks after that region — and the rep’s own connected mailbox brings it into
the CRM, where the rep creates the lead. The three points below therefore no longer need an answer.
If you would rather have info@ inside the CRM, it can be added as an ordinary user, provided it is a
licensed mailbox that can sign in to Microsoft itself — a Microsoft 365 shared mailbox, which has no
sign-in of its own, cannot be added this way — and anything done while signed in as info@ is
recorded as info@, not as the person using it. Tell us if you want that.
The original question, kept for reference. Inquiries arriving at a shared address such as info@ need
a home in the system. Three things to settle:
- Should shared mail appear in the same inbox as a rep’s own email (with a label showing where it came from), or on a separate screen?
- Who can see it — everyone, or only certain roles?
- When someone picks up a shared inquiry, does it get assigned to them automatically, or does a manager assign it?
Why we are asking: it changes the inbox screen design.
Previously, if we did not hear: same inbox with a source label, visible to Sales Reps and above, and
whoever replies first picks it up — replaced by the decision above.
What we heard on 9 September: an inquiry to the shared address that concerns a region goes to
whoever looks after that region, and the same first email often reaches both the shared address and a
person’s own inbox — it must become one lead, not two. Whether the shared address is a mailbox or a
distribution list now matters only if you choose to add info@ as a user.
(Our ref: OQ-52)
Q7. When should the system flag a deal as going quiet?
We will work out typical deal durations from your Pipedrive history. But the point at which you want a warning is a business judgement, not a statistic — and it differs between bulk and private-label.
Roughly: after how many days of no contact should a bulk deal be flagged? And a private-label deal?
Why we are asking: it sets when reminder alerts fire. If we do not hear: we set defaults from your historical averages. These stay adjustable in settings afterwards, so this one is low-risk. What we heard on 9 September: the real problem is leads and deals sitting idle with nothing prompting a follow-up — you want the flag and a follow-up action (a reminder, or a drafted email for a person to approve). The day thresholds still come from your Pipedrive history. (Our ref: OQ-12, OQ-13)
Your ERP
Q8. Confirming IFS, and what needs to pass to it. — updated 12 November 2026
At the progress meeting of 10 November you told us IFS is being replaced by another ERP, not yet chosen. Until you choose it and give us its details, this version keeps to the handover document, and no automated link is planned. One thing is still open:
- When a deal is won, what information does the ERP need — customer details, product, quantity, price, delivery terms, anything else? And should anything come back the other way, such as shipment or invoice status?
To be clear on scope: this version produces a handover document that someone enters into your ERP. A direct automated link between the two systems is a separate piece of work, priced separately. This question is about getting the document right.
Why we are asking: it defines what goes on the handover document. If we do not hear: we use the standard set — customer, product, quantity, price, delivery terms. (Our ref: OQ-2, OQ-19)
The trade-show capture app
Q9. The prototype does more than we have scoped — and less. ⚠️
The working prototype you have been shown has grown beyond what was agreed for this version, and is also missing two things that were agreed. We would rather raise this now than have it surface at testing.
In the prototype, but not in the agreed scope for this version:
- Campaign-based scoring of leads
- Splitting a pile of business cards captured in one photo
- Importing cards from the camera roll
- Loading an exhibitor list before a show to rank who to target
- Duplicate detection
Agreed for this version, but not yet in the prototype:
- Scanning badge QR codes
- Working properly with no signal — right now a capture taken without connectivity is lost, which matters on a trade-show floor. Updated 11 October 2026: the app opens with no signal, and a capture that cannot be sent is kept on the phone, marked “not sent”, and sent as soon as the app is open with a connection, so no capture is lost. It does not send in the background while the app is closed.
Our position is that the five extras stay out of this version and the two gaps get built. But you have seen the extras working, so we want your agreement rather than assuming.
Why we are asking: it settles what the capture app contains at handover. If we do not hear: we build to the agreed scope — the two gaps in, the five extras out. What we heard on 9 September: badge QR scanning is needed (please send us a photo of a recent badge so we can see what it encodes); the existing-client check should happen during the scan itself, showing whether the company or person is already a customer and how often you deal with them; and you would like a generic same-day email to people met at the stand (not built in this version — see Q15, withdrawn 11 October 2026). Campaign planning is already in the prototype. The other extras stay out, as proposed. Updated 12 November 2026: loading an exhibitor list before a show to rank who to target is now in this version, in Phase 1, following the progress meeting of 10 November: the app plans each show, with every exhibitor researched and ranked, and visit notes for each stand (Product Requirements, Section 3.5). Cards are captured one at a time, so splitting a pile of cards in one photo stays out, as does importing from the camera roll. What is still open here is the badge: until we have a photo of a recent badge, the app reads the common QR formats, and we confirm the real one when the photo arrives. (Our ref: OQ-55)
Q15. The same-day email to people met at the stand — new, 27 September 2026 — withdrawn 11 October 2026; no answer is needed. The capture app will not send this email in this version. The same message can go to the people met at a show as one bulk email to a saved list of contacts (Phase 2), or a rep can email any lead from the CRM. The question is kept below as it was asked.
You asked on 9 September for a generic email to go the same day to everyone met at the stand. We propose sending it from the rep’s own mailbox: a Microsoft 365 shared mailbox cannot send on the system’s behalf in this version. One email goes to each person, not one to a list. Please confirm the sender, and settle two more things:
- Should the rep send each email right after the capture, or review and send the day’s list together in the evening?
- Should the rep approve each email before it goes, or is approving the template once enough?
Why we are asking: it decides how the follow-up screen in the capture app works. If we do not hear: we build our proposal — sent from the rep’s own mailbox, the rep approves each email before it goes, and it can go at any time that day. (Our ref: OQ-60)
Bulk email
Q10. How should newsletters and other bulk emails work?
Bulk email is part of Phase 2 (agreed 9 October 2026): the week’s prices to every customer at once, and one message to a saved list of contacts — newsletters, product promotions, trade-fair invitations — with opens and clicks shown for each send.
The price side needs nothing from you: each customer still receives only their own standards, so it is the same email we already send per account, just sent for the whole round in one action. Updated 18 November 2026: the price side is built and now arrives in Phase 1. The newsletter side stays in Phase 2.
The newsletter side is one message going to many people at once, and that brings obligations a one-to-one email does not. Updated 19 November 2026: we have started building the newsletter side early, and we chose the four points below ourselves, picking the simplest option for your staff. Please confirm them, or tell us what to change.
- Who receives it. Everyone on the chosen list receives it, unless they have unsubscribed. Your contacts are business contacts, so we do not ask for a separate “agreed to receive” tick on each one.
- Unsubscribing. Every newsletter has a link at the bottom, with your postal address. The link opens a page with one Unsubscribe button, and needs no sign-in. Pressing the button stops that person receiving newsletters. (The link alone does not unsubscribe, because email security systems open every link in an email before the reader does.) The CRM then leaves them out of every later newsletter. Someone who can edit the contact can turn newsletters back on; they should do so only when the contact asks. Unsubscribing stops newsletters only: that person still receives their weekly prices and ordinary emails from your staff.
- Which address it comes from. The newsletter comes from the mailbox of the person who sends it,
so replies reach that person. It goes out at up to 20 emails a minute (about 15 minutes for 300
contacts), which keeps within Microsoft’s limits. One person can send to at most 5,000 contacts in
any 24 hours, counting their weekly price emails, which keeps within Microsoft’s daily limit. Bounced emails come back to that mailbox, and the
CRM reads them, so it can show what was not delivered and leave dead addresses out of later sends.
A shared address (for example
news@empireteas.com) is not built. It would need a DNS change by your IT provider and a mailbox for the replies. - Who it goes to. A saved list of contacts, filtered by market, customer type, tag or any field you add, as you asked on 9 October.
Why we are asking: you may need something different. For example, you may need a tick for contacts who have agreed to receive newsletters, if you send to people in Europe. Or you may want a shared sending address. If we do not hear: the newsletter side works as described above. Any change you ask for at UAT is made as ordinary work. (Our ref: OQ-62)
Accounts
Q11. Are the lists on the account screen right?
To start building, we wrote the choices shown on each account ourselves, from the prototype: customer types (Importer, Distributor, Supermarket, Foodservice, Wellness, Private-label), regions, packaging formats, the Incoterms you trade on (EXW, FOB, CFR, CIF), and the list of certifications. Please check them and tell us what to add, remove or rename.
One point in particular: on an account, the certification field records what that buyer requires. Is that the same list as the certifications you hold, or different?
Why we are asking: these choices appear on every account, and your migrated Pipedrive data will be matched against them. If we do not hear: our lists stand, and we correct them when the Pipedrive export shows your real values. In this version, changing a list is a small update from us rather than a setting you edit. (Our ref: OQ-66)
Q12. After how many days without contact should an account count as “dormant”?
The accounts list marks an account Dormant when nobody has been in touch for a while — no email, call, sample, quote or note. We need the number of days. Should it be the same for bulk and private-label buyers?
Why we are asking: the Dormant badge tells reps which customers are going quiet. If we do not hear: we use 90 days for every account. It is a setting, so changing it later is quick. (Our ref: OQ-67)
Q16. Several parties on one lead — new, 27 September 2026
You told us on 9 September that an existing customer sometimes starts a new project and brings in agencies, so that several companies and people are on one email thread about one lead. You wanted them grouped under that lead, and said the idea still needed thinking about. Our proposal: each lead and deal keeps one main customer, plus a list of other people involved, each marked with their part — buyer, agency or agent. Does that fit how these projects work?
Why we are asking: it decides how the lead and account screens show who is involved, and which emails are linked to the lead. If we do not hear: we build the proposal above. (Our ref: OQ-57)
Q17. Past orders on the account screen — new, 27 September 2026
The requirements show a customer’s past orders on the account screen, read from your ERP. In this version the CRM hands a won deal to your ERP as a handover document that someone enters there, and nothing comes back from the ERP, so no order history reaches the CRM. We see two options:
- (a) a simple “last order” entry on the account — date, standards and value — that a rep or the Tea Room types in, as a stop-gap; or
- (b) nothing for now; the past-orders panel arrives with the follow-on work for the ERP that replaces IFS, which is priced separately.
Why we are asking: option (a) is small, but it is typed in by hand and will go out of date unless someone keeps it current. If we do not hear: (b) — nothing is built until the ERP connection exists. (Our ref: OQ-63)
The AI
Q18. Does it matter to you where the AI processes your data? — new, 30 September 2026 — answered and closed 4 November 2026; no reply is needed unless you disagree.
Our answer: the AI may process a request in any of Amazon’s commercial regions.
Why this is safe, whichever region does the work (Amazon’s own Bedrock documentation):
- every request starts in your own AWS account;
- the request travels only on Amazon’s own network, encrypted;
- Anthropic does not see it, and neither Amazon nor Anthropic uses it to train a model.
Why we chose it: the worldwide route costs about 10% less than US-only. It also has a request limit of its own, so a batch of new leads does not slow the answers people are waiting for, and Amazon releases new models on it first.
If your company ever needs processing kept in the US, that is a setting, not new work: just tell us. The question as we asked it on 30 September is kept below, with the 4 November change marked.
The AI features send buyer inquiries, synced emails and account history to Anthropic’s Claude models through Amazon Bedrock, from inside your own AWS account. Our proposal said this would happen “inside Empire Teas’ own AWS account and region”. As Amazon offers the newest models today, that is not quite how it works:
- when someone is waiting for the answer — reading an inquiry, drafting a reply, preparing an account briefing — the smaller model runs in one of Amazon’s US data centres;
- the work the smaller model does on its own for each new lead — its summary, the spam check and the research — uses Amazon’s worldwide routing, so a request may be processed in any of Amazon’s regions. Updated 4 November 2026: we moved this work there because Amazon gives the worldwide route its own limit on requests a minute, so a batch of new leads no longer slows the answers people are waiting for, and it costs about 10% less;
- the newest larger model, which we have chosen because it is better and costs the same, is offered for now only through Amazon’s worldwide routing. Every request still starts in your account.
Amazon already offers a US-only option for the earlier models, and now for the newer model too. It costs about 10% more, and moving to it is a setting, not new work.
Why we are asking: some of your buyers are in the EU, and some companies have rules about where customer data may be processed. If you have none, the worldwide option is cheaper and gets new models first. If we do not hear: we use the worldwide option for the larger model while it is the only one, and for the smaller model’s background work on new leads; we move either to US-only only if you ask. (Our ref: OQ-15)
Part 2 — What we will find out ourselves
Updated 26 October 2026: since 6 October, no Pipedrive access or review session is needed. You shared exports of your leads (13 September) and of your organisations and people (28 September), and we have answered what we can from them, including the record counts. For the second phase we will ask for an export of your deals, which answers the questions about pipelines, stages, deal lengths, grades and pack formats. The rest — active users, templates and sequences, forecasting settings, inquiry languages and the features the team would miss most — we will ask you directly. The list below is kept as it was first written.
| What we will look at | Why |
|---|---|
| How many pipelines exist and what the stages are called | Your current stages are the starting point for ours — designing them from scratch produces something unfamiliar |
| Leads versus deals — how the team uses each | Sets where a new inquiry lands |
| How many private-label clients versus bulk buyers | Confirms the two-pipeline design |
| Total record counts to migrate | Sizes the data migration |
| Which tea grades appear in real deals | Fills the grade dropdown |
| Typical order volumes and pack formats | Fills the volume and packaging dropdowns |
| How long deals actually take | Feeds the “going quiet” thresholds in Q7 |
| Active user count | Sizes rollout and training |
| Email templates and sequences in use | Sizes what needs moving across |
| Whether weighted forecasting and idle-deal alerts are used, and at what settings | Sets sensible defaults |
| What languages inquiries arrive in | Confirms whether translation is needed |
| Which features the team would most miss | Backs the “nothing lost in the move” commitment — this one we will ask in conversation on the day |
Updated 26 October 2026: no session is needed. Your Pipedrive exports of 13 and 28 September answer what they can; for the second phase we will ask for an export of your deals, and we will ask you the rest directly (What the Project Needs from Empire Teas, Section 1.3).
What we would like back, and by when
| What | Ideal timing | |
|---|---|---|
| ⚠️ | Q1 — what each of the six roles should hold on day one, and who reports to whom | Any time; built on the stated assumption, which you can correct at UAT |
| ⚠️ | Q4 — standard certification data | Any time; built on the stated assumption, which you can correct at UAT |
| ⚠️ | Q9 — a photo of a recent visitor badge (the rest of Q9 settled 12 November 2026) | As soon as you can; the app reads the common QR formats until then |
| Pipedrive (Part 2) — exports received 13 and 28 September 2026; no session needed (6 October 2026). For the second phase, an export of your deals | Before Phase 2 starts | |
| Q3, Q5, Q7, Q8 | Over the first month, as each area comes up | |
| Q2 and Q6 — decided 24 September 2026; no reply needed unless you disagree | — | |
| Q18 — where the AI may process your data: answered and closed 4 November 2026 (any of Amazon’s commercial regions); no reply needed unless you disagree | — | |
| The live system’s administrator email (new, 24 September 2026) — the address, confirmed by your IT, of the account that starts in the IT Admin role (listed in What the Project Needs from Empire Teas) | Before the live system is first started — it cannot start without it | |
| Q10 — how newsletters and other bulk emails work (bulk email added to Phase 2, 9 October 2026; our choices made 19 November 2026) | At the Phase 1 hand-over, to confirm our choices; the newsletter side is built on them meanwhile | |
| Q11, Q12 — the account lists and the dormant threshold | Before your Pipedrive data is migrated; each has a safe default meanwhile | |
| Q13, Q14 — the weekly price workbook details and private-label pricing (new, 27 September 2026) | Before the Phase 1 acceptance round — the price upload and review screens are built on the answer given under If we do not hear | |
| Q16, Q17 — several parties on one lead, and past orders (new, 27 September 2026) | Before Phase 2 starts — the lead and account screens are built on the answer given under If we do not hear |
Nothing here is a criticism of the information already provided — it has been more than enough to design and price the work. These are the last details that only come into focus once building starts.
Happy to walk through any of these on a call rather than in writing, if that is easier.
Change history
Older changes to this document, newest first.
- 4 November 2026: Q18 answered and closed — we have settled where the AI may process your data: in any of Amazon’s commercial regions, every request starting in your own AWS account. The AI’s background work on each new lead (its summary, the spam check and the research) now uses Amazon’s worldwide routing, as the larger model already does.
- 26 October 2026: Part 2 — no Pipedrive access or review session is needed (your exports answer what they can, and we ask you the rest); the timing for Q13, Q14, Q16 and Q17 updated, since the screens they concern are built.
- 11 October 2026: Q15 withdrawn — the same-day email from the capture app is not built in this version; Q9 gains a note on captures taken with no signal.
- 6 October 2026: Q13 narrowed to two points.
- 30 September 2026: every lead has an owner; who can hand a record over. Later the same day, Q18 added: where the AI may process your data.
- 27 September 2026: five questions added (Q13–Q17); Q1’s reporting-line answers restated for the “all records” permission.