Operating Evidence 01 · Intake Architecture

Compressing Time-to-Value Through Intake Architecture

The intake layer captured client intent. It did not capture what the advisor needed to act. Every new engagement opened in deficit, and the deficit compounded before the relationship had a chance to produce value.

Environment
Fora Travel · Remote / NYC
Role signal
Implementation · Onboarding · Product Ops
Mechanism
Advisor-readiness intake + human review gates
4–5 → 1–2
Clarification loops

Reduced in better-qualified intakes after the intake architecture rebuild.

~20 min
Advisor time recovered

Per case through AI-assisted synthesis with human review gates.

100%
Human review retained

For ambiguous or high-exposure routing decisions.

6 → 14
Fields rebuilt

Expanded around readiness, not form length for its own sake.

The failure was clarification debt.

The client had submitted intent, but the advisor did not yet have action-ready information. The first call risked becoming a reconstruction exercise instead of a value-producing interaction.

The redesign started from the advisor's first productive action and worked backward: what does the advisor need to know before call one to act without asking a single clarification question?

Intent capture became advisor-readiness.

Before

  • Destination and dates captured.
  • Budget and flexibility reconstructed manually.
  • Advisor opened in clarification mode.
  • First call delayed productive action.

After

  • Routing signals captured early.
  • Missing-information checks surfaced before call one.
  • AI synthesis compressed preparation time.
  • Human review protected judgment.

Commitment is blocked until risk is checked.

Supplier confirmation

No client-facing promise leaves the operation before availability is verified.

Alternate routing

Disruption paths require a viable fallback before recommendation.

Cost exposure

Financial impact is made explicit before approval.

Human review

Ambiguous or high-stakes cases stay human-in-the-loop.

From first intent to first visible value.

SubmitClient intent captured with readiness cues.
DetectMissing information flagged before advisor work begins.
SynthesizeAI creates advisor-ready context.
ReviewHuman confirms ambiguous or high-exposure cases.
ConfirmFirst call becomes confirmation, not reconstruction.
Product-facing translation

The current synthesis layer works when inputs are text-based and structurally complete. It degrades when context arrives through voice notes, social messages, referral chains, or fragmented channels. That becomes a classification model and integration problem, not a generic summarization problem.