Forward Deployed Engineering — Quick Reference

Last Updated: 2026-06-30 Full Chapters: docs/08-Forward-Deployed-Engineering/

FDE vs. Adjacent Roles

Role Primary Accountability Writes Production Code? Post-Sale? Metric
FDE Client value realization Yes Full lifecycle NRR, adoption rate
Sales Engineer Deal closure Rarely Pre-sale only Pipeline, bookings
Professional Services Project delivery Yes Contract term Milestones, delivery
Solutions Architect Architecture design Rarely Design phase Architecture sign-off

FDE Engagement Lifecycle

text
Discovery → Assessment → POC Design → POC Execution
    → Architecture Review → Production Planning → Launch → Expansion

Definition of done for each phase:

Phase Exit Criteria
Discovery Current-state map complete; all key stakeholders interviewed; open questions documented
Assessment Three-dimension readiness scored; remediation plan for gaps; POC start date confirmed
POC Design Hypothesis explicit; success criteria signed off in writing; production gap mapped
POC Execution All encounters evaluated; go/no-go decision made
Architecture Review Risk register complete; blocking issues documented with owners
Production Planning Migration plan complete; App Orchard submitted; governance approvals tracked
Launch Shadow mode complete; canary expanded; baseline KPIs established
Expansion Use case 2 scoped; value report delivered

Discovery — Key Principles

  • Spend 1–2 days on pre-discovery research before the first meeting
  • Listen:talk ratio in discovery sessions: 30:70 (FDE:client)
  • Three layers to populate: technical environment, pain points, organizational dynamics
  • Discovery complete = all key stakeholders interviewed, all open questions documented, compliance included
  • Never accept self-reported data quality — run sample queries

AI Readiness Assessment Scorecard

Dimension Score 4: Ready Score 3: Near-Ready Score 2: Developing Score 1: Blocking
Data Maturity All required data > threshold; accessible via approved path Minor gaps; resolvable in 2–4 weeks Significant gap in quality or access PHI policy blocks; data not available
Infrastructure Cloud PHI approved; LLM access confirmed; gateway deployed PHI approval pending; gateway planned Cloud policy unclear; connectivity untested On-prem only; no connectivity path
Organization Exec sponsor; champion identified; AI governance in place Most elements; 1–2 gaps resolvable Champion or tech lead absent No sponsor; no governance; no capacity

Any dimension at Score 1 = blocking regardless of overall score

POC Design Checklist

  • [ ] POC hypothesis written (capability + data source + success criteria + time constraint)
  • [ ] Success criteria signed off in writing by client before execution begins
  • [ ] Production gap analysis mapped (8 dimensions: data, concurrency, governance, prompt mgmt, error handling, observability, integration, security)
  • [ ] Go/No-Go decision framework defined (GO / Conditional GO / Extend / No-Go / Redesign)
  • [ ] Client engineering team included in execution from Day 1
  • [ ] Healthcare: clinical evaluation rubric designed; minimum 2 physician evaluators; safety checks defined
  • [ ] App Orchard submission timeline factored in (8–12 weeks)

Architecture Review — 8 AI-Specific Risk Patterns

Pattern Detection Question Primary Risk
PHI in observability traces Where do LLM payloads get logged? HIPAA violation
No AI gateway Where do API keys live? No audit log, cost control
CDS Hook without circuit breaker What if LLM is slow / unavailable? Clinical workflow blocked
No model version pinning What model version is specified? Unexpected behavior change
AI draft auto-files to EHR What approval step exists? Unapproved AI content in record
No prompt version management How are production prompts changed? Quality regression, no rollback
FHIR over-broad scopes System/ or patient/ scopes? Minimum Necessary violation
No fallback workflow What do clinicians do if AI is down? Workflow disruption

Value Engineering — Healthcare Quick Reference

Use Case Primary Benefit Measure KPI Baseline Source
Discharge Summary AI Physician time reduction Min per discharge (before/after) EHR audit log: note open → note signed
Prior Auth AI Denial rate reduction Denial rate by payer/service line RCM system
Medical Coding AI CC/MCC capture improvement CC/MCC rate change Coding department reports
Patient Engagement AI 30-day readmission reduction Readmission rate change EHR + CMS data

All financial figures are illustrative — validate against client-specific data.

ROI model formula:

text
Net 3-Year Value = (Annual Benefit × 3 years × adoption ramp) − (One-time cost + Annual ops × 3 years)
Payback Months = One-time cost ÷ (Monthly benefit − Monthly ops cost)

Communication Formats

Situation Format Length Lead With
Executive update One-pager 3-min read Bottom Line (BLUF)
Technical status Dual-section status report 1–2 pages Overall status indicator
Blocking issue Escalation Half page Situation + Impact + Decision required
AI risk Risk brief Half page Clinical consequence language

BLUF principle: The most important message goes in the first sentence. Never bury the conclusion.

Common Objections — Quick Responses

Objection Root Cause to Diagnose Key Response Element
"Our data is too sensitive" PHI-in-cloud policy, BAA, past incident Map the exact data path with specific controls at each step
"ROI isn't clear" Missing model, wrong benefit category Build model with client data; conservative adoption rate
"We'll build it ourselves" Cost, control, prior failure Honest build-vs-buy: opportunity cost of engineering time
"AI makes mistakes" Liability, safety, skepticism Physician-in-loop design; evaluation results; safety checks
"Security review takes too long" CISO not aligned, docs missing Proactive documentation package + CISO briefing session
"Too disruptive" No champion, past failure Champion-led evaluation; additive not replacement workflow

Healthcare FDE — Critical Stakeholders

Stakeholder Priority Can't Proceed Without
CMIO Critical Clinical AI governance; physician champion access
Privacy Officer / Compliance Critical PHI-in-cloud approval; BAA process; HIPAA sign-off
CIO Critical IT architecture approval; security scope
Physician Champion Critical Clinical adoption; evaluation credibility
IT Director (Clinical Systems) Important FHIR access; App Orchard navigation

Healthcare FDE — Critical Timelines

Item Typical Timeline FDE Action
Epic App Orchard review 8–14 weeks Submit during POC design, not after POC success
Anthropic BAA negotiation 2–4 weeks Initiate in Assessment phase
PHI-in-cloud policy approval 4–8 weeks (if not pre-approved) Surface in Discovery
IT Security review 6–8 weeks (if no prior AI vendor) Surface in Assessment; provide docs proactively
Physician champion training 1–2 weeks Schedule during POC Execution

Interview Quick Reference

FDE vs. SE: SE = pre-sale, deal metric; FDE = full lifecycle, client success metric; FDE writes production code

Objection framework: Acknowledge → Diagnose → Respond (technical specifics) → Test (did it resolve?)

POC failure vs. production failure: POC failure costs 4 weeks; production failure costs 6 months and trust

Healthcare regulatory: Non-device CDS = clinician can independently verify basis; SaMD = clinician cannot

App Orchard: Epic's app review — 8–14 weeks; submit in parallel with POC, not after

Physician-in-the-loop: AI produces draft; physician reviews, edits, approves; physician signature = attestation

Value engineering: Time savings + quality improvement + revenue impact + cost avoidance; label all figures illustrative

See Also