Docs / 08-transformation/03-discovery/pilot-order-desk-discovery-plan
Pilot discovery plan — email-order desk
How TPL gathers the information needed to scope, design, and baseline the REQ-014 pilot — CEO clarification call, department deep dive, systems/access verification, sample-data collection, and go/no-go scoping.
Pilot discovery plan — email-order desk
Information-gathering plan for the REQ-014 pilot: the department that reads email orders, creates SoftOne ERP orders, matches products on the supplier B2B system, and follows up with the customer until completion. Goal: within ~2 weeks of the CEO call, know enough to write the pilot scope, success criteria, and technical design — or say clearly what blocks it.
What we must learn (information targets)
| # | Target | Why it matters | Source |
|---|---|---|---|
| 1 | Which department owns the flow — name, headcount, manager, a champion | Pilot owner + interviewees | CEO call |
| 2 | Volume & shape of order intake — orders/day, emails per order, formats (free text, PDF/PO attachments, Excel, images), languages, customer segments | Sizes the extraction problem and the KPI baseline | CEO call → dept interviews → mailbox sample |
| 3 | The as-is process end-to-end — every step, handoff, exception path from email arrival to “completed” | Process map (SIPOC + swimlane) is the design substrate | Dept interviews + desk-side observation |
| 4 | SoftOne specifics — version/edition, Web Services/API availability & licensing, order-entry screens/fields used, customer & product master data quality | Gates the automated order-creation leg (also HANDOFF pending: verify SoftOne Web Services) | IT interview + SoftOne partner if needed |
| 5 | Supplier B2B system(s) — which vendor(s), portal vs API/EDI, product-code mapping today (supplier code ↔ SoftOne code), account/credential model | Gates the product-matching leg; may be the hardest integration | CEO call (identify) → IT/dept (verify) |
| 6 | Follow-up practice — what “follow up until completed” means concretely: statuses tracked, who chases whom, channels, SLAs, what customers expect | Defines the agent’s autonomy boundaries | Dept interviews |
| 7 | Baseline KPIs — time per order, orders/person/day, error/rework rate, follow-up latency, order-to-completion lead time | Without a baseline the pilot can’t prove anything | 1-week measurement + mailbox/SoftOne history |
| 8 | Sample data — 20–50 representative order emails (with attachments), matching SoftOne orders, B2B match examples, incl. ugly/exception cases | Feeds extraction prototyping before any integration exists | Dept, after NDA/data-handling agreement |
| 9 | Constraints — data handling (customer PII), mailbox access approval, pilot users’ availability, no-go areas, budget/timeline expectations | Shapes what we may build and by when | CEO call |
| 10 | Success definition — what result makes the CEO extend beyond the pilot | The actual acceptance criterion | CEO call |
Phases
Phase 0 — CEO clarification call — ✅ DONE (2026-07-06)
- Instrument: questionnaire-pilot-order-desk.md — administered as a guided conversation, recorded with Fathom; transcript in raw/, distilled into the interview note per ADR-003/ADR-012.
- Outputs achieved: pilot department identified (cylinders team, 3 people, champion Fotini); volumes and the three sub-processes captured; supplier-B2B assumption corrected — suppliers are email-only, ~3 customer platforms instead; constraints (no budget figure, urgent, no stated red lines) and success definition (one platform, morning worklist, time-to-close) captured; IT contact identified (Evangelia — intro pending); sample-data set (20–50 order emails + SoftOne orders) requested.
- REQ-014 moved
captured→validatedin the same change. Phase 1 starts when the team-details email + Evangelia intro arrive.
Phase 1 — Department deep dive (week 1 after the call)
- Interviews: department manager (60–90 min, questionnaire-department-heads.md as base) + 2–3 order-desk operators (questionnaire-process-owners.md per key process).
- Desk-side observation: sit with an operator for half a day; time 5–10 real orders end-to-end. Observation beats self-reporting for baselines.
- Map the process: SIPOC + swimlane (Mermaid), one file, destined for
05-Processes/(first real content — create the folder then). - Collect the sample set (target #8) under the agreed data-handling terms.
Phase 2 — Systems & access verification (parallel, week 1–2)
- IT interview (questionnaire-technical-it.md §SoftOne + mailbox sections): SoftOne Web Services confirmed working (test call), M365 mailbox access model (shared mailbox? Graph API consent?), supplier B2B access demonstrated live.
- Deliverable: a technical feasibility note per leg (email intake / SoftOne write / B2B match / outbound follow-up) with a red–amber–green rating and what’s missing for each.
Phase 3 — Synthesis & pilot scoping (week 2)
- Baseline KPI sheet from Phase 1 measurements.
- Pilot scope proposal: which legs are automated in the pilot vs deferred, autonomy boundaries (draft-vs-send), success thresholds, timeline, pilot team.
- Scored entry in the AI use-case pipeline; go/no-go review with the CEO.
Risks / open questions
- SoftOne Web Services: stated available on the call — still verify with a live test call (licensing/version unclear); program-level gate (see HANDOFF).
Supplier B2B may be portal-only— resolved (call, 2026-07-06): suppliers have no systems at all, email-only RFQs. The matching leg becomes email drafting/parsing; the customer-platform leg (~3 platforms, login-based) is the new integration unknown — names/access to capture in Phase 2.- The pilot predates the on-prem platform — decide pilot infrastructure explicitly (cloud interim vs waiting for hardware); needs an ADR when decided.
Reconfirm the client email domain— resolved (call, 2026-07-06): Prime Products attendees joined on@primeltd.com(Fathom labels); the Cloudflare Access allow policy already matches. Website remains primeproducts.gr.
Related
- REQ-014 — the pilot requirement
- questionnaire-pilot-order-desk.md — the CEO call instrument
- requirement-analysis-framework.md — general discovery method this plan specializes
- customer-service department plan — likely owning department (to confirm)