Consultancy Scoping & Proposals — Cutting Admin Without Cutting Judgment
Billable consultancies lose days to scoping docs and proposal assembly. LLM tool paths from discovery notes to CRM — human-approved, margin-positive.
Partners sell judgment. Associates spend evenings formatting Word docs. That is not billable. That is leak.
The scoping leak pattern
- Discovery call notes in Otter/Zoom/transcript
- Partner mentally scopes — nothing structured in CRM
- Associate rebuilds scope from memory
- Proposal version 4 emailed at 11pm
- Won deal — handover missing half the context
We quantify days from verbal yes to signed SOW and rewrite cycles per proposal.
What LLM integration fixes (era three only)
| Step | Old way | Integrated way |
|---|---|---|
| Discovery | Notes scattered | Transcript → structured scope fields in CRM |
| Scope | Partner reinvents | Draft SOW from past similar engagements (cited) |
| Pricing | Excel archaeology | Pull rate card + effort model; partner adjusts |
| Handover | Email thread | Delivery brief auto-drafted from signed scope |
Every write path requires named approver before client sees it.
Why we refuse "auto-send proposals"
Legal, pricing authority, and relationship nuance stay human. AI that sends proposals without partner review is a liability — and a fast way to lose margin on underscoped work.
KPIs we hold ourselves to
- Hours from discovery → draft proposal
- Scope change rate post-signature (should fall)
- Associate rework hours per won deal
What to do next
Start with the audit — we map scoping leak in RM before recommending build.