What “Forward Deployed Engineer” means here
Industry archetype: an engineer embedded with a customer, working their real process, not building in the abstract back at HQ.
At ClearSky it is a function, not one person: generally 1 business FTE + 1 technical FTE per client, on-site or virtual. They translate the client’s real close (or later P2P / O2C / payroll) into what the platform needs, fast — and they stay close enough to spot the next process to map.
This is not “toss standardized software over the wall.” Every client’s close is their own version of Session 1, written down their own way, or not written down at all. Someone has to turn that into what FlightPlan needs. That someone is the FDE. The intern job is to build the tools those people use.
What the FDE is listening for
If these are known, there is enough to run orchestration. Missing any of them is why discovery exists.
1. Sequence & dependencies
What has to happen before what. Example from the call: you cannot reconcile “widgets ordered vs widgets shipped” until both files exist.
2. Maker-checker
Who does the work, who approves. Especially anything that hits the system of record: the person who drafts cannot be the person who approves (SoD).
3. Completion signal
What “done” looks like. Not binary. High confidence: JE status = complete in the GL. Low confidence: an email thumbs-up (done, or “I will tell you when done”?). Can be partial (4 of 5 bank accounts). Can be multiple sources (file on SharePoint and/or emailed) — product still deciding AND vs OR.
4. Timing anchor
Business day, not calendar date. Also parse Frequency: Monthly / Quarterly / Annual / Adhoc.
5. Risk & controls
Where it can go wrong, and what catches it. Needed both for orchestration and for the client-facing report.
Four names people mix up (from the Q&A)
| Thing | Who makes it | What it is | Goes into FlightPlan? |
|---|---|---|---|
| Artifacts | Client | Anything they give you: checklists (often Excel), SOX/compliance PDFs, SOPs, process writeups. Not “one Excel.” | Raw input |
| Discovery spreadsheet / master task list | ClearSky, then client teams edit | Domain knowledge + artifacts, organized by team. Draft is maybe ~50% right. Teams delete / add / fill. Interviews close remaining gaps. | Yes — this is the input |
| Discovery interviews | FDE + each team | Downstream of the spreadsheet, not parallel. Spreadsheet-alone misses handoffs, deps, risks people only say out loud. Do not walk in with “tell us your process.” | Feeds the spreadsheet |
| Discovery Report | ClearSky → CFO / controller / leadership | ~30-page analysis: task counts, risk, effort, pain points, key-person concentration. Not workflows. Not the product solution. POC value: “I had no idea.” | No. Sibling output of the same spreadsheet. |
Usually: client artifacts → ClearSky draft spreadsheet → one team-edit pass → interviews → maybe one more review before go-live. Count of rounds depends on how messy the inputs are.
Walkthrough: how it was done by hand (first client)
This is the current manual path. It does not scale. That is why the tools exist. Example client in the room: BlackBerry, anonymized in later files as Veltara.
- Raw accounting-ops checklist. Few columns. Tasks grouped by category (Bank/Cash, Prepaid, Capital Assets, Leases), not by owning team. “Responsible” is staff initials (MM ≠ a role). Timing is day-of-month offset vs period end. Dozens of near-duplicate rows = one task × many legal entities / bank accounts. Maker-checker is a “review, sign and date” row with no structural mark. This is what extraction tooling has to survive.
- Draft discovery spreadsheet. Do not take the raw sheet to interviews. Add a “gold” view of a normal close. Prompt for: (a) upstream tasks the checklist implies but does not list, (b) tasks a normal close usually has that are missing. Organize by team. Push to all 10 teams. Blank “list everything you do” under-reports; a 20-row strawman gets fast deletes and adds. Tax example: ~75-row checklist, most of it parallel work, not close.
- Discovery interviews after the spreadsheet pass. Walk the team’s section: why, how, who, who reviews, where it goes. Interviews exist to close gaps steps 1–2 surfaced.
- Discovery Report for leadership (next section).
- Same 776 tasks → FlightPlan fields: Task, Owner, Checker, Completion signal, Dependency, Timing anchor. Every field should trace to an artifact, a team update, or an interview moment. That provenance is what tooling must keep, not just the final cell.
Discovery Report numbers (BlackBerry / Veltara, as on the slides)
Ten finance teams, quarterly close. Numbers will drift a few tasks as the spreadsheet is cleaned — treat as order-of-magnitude, not a live system of record.
close tasks mapped
teams interviewed
high-risk (18%)
fully manual
high effort (15%)
of volume in Accounting Ops (438)
Pain the report is for (CFO audience)
- Manual consolidation — NetSuite, Adaptive, Workday, ADP… dumped into Excel before close work starts.
- Key-person concentration — one AccOps analyst owns 243 tasks; a departed AR manager took the team’s how-to with her.
- Cascade risk — currency reval waits on bank clearing + AP close; tax provision waits on accounting.
- Highest high-risk counts — Tax (38) and Revenue (19), both tied to statement accuracy.
Effort / risk table by team (slide 16)
| Team | Tasks | Eff. High | Eff. Med | Risk High | Risk Med |
|---|---|---|---|---|---|
| Accounting Operations | 438 | 41 | 353 | 44 | 366 |
| FP&A | 113 | 13 | 84 | 10 | 89 |
| Tax | 103 | 39 | 41 | 38 | 36 |
| Treasury | 36 | 2 | 11 | 15 | 11 |
| Revenue | 27 | 13 | 12 | 19 | 7 |
| Accounts Payable | 16 | — | — | 1 | 14 |
| External Reporting | 15 | 1 | 8 | 3 | 7 |
| Payroll & Benefits | 12 | 9 | 3 | 12 | — |
| Accounts Receivable | 9 | — | — | — | 7 |
| Revenue – QNX | 7 | — | 4 | — | 4 |
| TOTAL | 776 | 118 | 516 | 142 | 541 |
Why you’re building FDE tools
Metric: time-to-live. BlackBerry discovery ran ~two months (first time + calendar of other people’s time). Target: compress toward two weeks (you still cannot make the client turn the sheet around in a day). Every week shaved is a week closer to them running on ClearSky.
Automate (this is the job)
- Artifact intake and draft spreadsheet.
- Consistency / gap / contradiction checks (e.g. X depends on Y but X is BD2 and Y is BD3).
- Provenance: this task came from the checklist / this interview line / we proposed it as a gap.
- Discovery Report generation. Later: in-product, not a PDF email; snapshot after each close; flux vs 3/6 months ago.
Do not automate yet
- Discovery interviews — keep human for the next ~3 clients until the pattern is clearer. Semi-auto / optional later is an open question.
- Exact FlightPlan schema, field names, platform mechanics — explicit non-goal of this session. Next deck.
FDE stays in the loop for judgment. Tools reduce variance and wait time; they do not skip the human on calls that still need a table.
Prior session: Finance and the close. SOX originals: JE · policies · TFN.
Originals
Fireflies: The role of the FDE at ClearSky · 19 Aug 2026 · 55 min · id 01M086GPD6WQ19PPWNHVATVHTV · open in Fireflies. Follow-ups from the call (not on this page): anonymized Veltara task list + discovery report.
Every short form
Hover any dotted term. Or search here. One line each.