All workflows
Sales

Sales proposal and RFP response generation

Draft a branded, on-message commercial proposal from Salesforce opportunity data and a Box library of past decks, with citation-backed sections.

Sales proposal generation: $0.77/run, 50% invisible in Box re-embedding and revision-draft re-injection. Firedog joins the Salesforce, Box, and LLM lines into one per-proposal P&L.

Model 38%Prompt 12%Orchestration 10%Retrieval 18%History 22%
ComponentWhy it is unattributedcalls$ / run
Model · 38%Proposal synthesis (Claude Sonnet 4)The model ingests a stitched context of Salesforce opportunity fields, product/pricing data, and retrieved past-proposal passages, then writes a 3,000-5,000 token structured proposal with section-level citations.tokens in: 70,000 · tokens out: 5,500Visible on the invoice.4$0.29
Prompt · 12%Proposal template, brand voice, and pricing schemaAn 8,000-token prefix encoding the proposal structure, tone-of-voice rules, and pricing/discount logic is prepended to every call without prompt caching.tokens in: 32,000 · tokens out: 0Visible on the invoice.4$0.096
Orchestration · 10%Salesforce and Box ingestion, section routing, assemblyAn ingestion agent reads the Salesforce opportunity and pulls matching case studies from Box; a routing agent dispatches each proposal section to a sub-call; a merge agent assembles and formats the final document. Each is a billed model call.tokens in: 16,000 · tokens out: 2,000Salesforce and Box ingestion agents are logged as 'CRM sync' and 'content pipeline' steps by the integration layer, never attributed to the proposal workflow's cost center.8$0.078
Retrieval · 18%Past-proposal and case-study vector lookupsEach section call retrieves the most relevant chunks from a Box-backed index of prior proposals and win stories. Embeddings are regenerated on every run rather than incrementally updated.tokens in: 20,000 · tokens out: 0Re-embedding the Box content library on each run lands on the embedding-provider invoice, separate from the LLM spend the sales-ops lead reviews.12$0.14
History · 22%Prior-draft re-injection across revision cyclesEach revision run prepends the full prior proposal draft so the model can edit against its own output. A 5,000-token draft compounds to 20,000+ tokens re-billed across four revision passes.tokens in: 55,000 · tokens out: 0Revision context is framed as 'draft history' by the editor app; it appears nowhere in the per-run cost the sales team sees on the model dashboard.4$0.17
A tailored proposal costs ~$0.77 to draft; across a 25-rep team it is ~$14.5k/yr, and 50% is Box re-embedding, CRM ingestion hops, and prior-draft re-injection on every revision that no dashboard joins.
$0.77
cost per run
$15k
per year
50%
unattributed today
$2.32
per user, per day
Scenario: 20-25 rep SaaS sales team, weekly proposal cadence. 25 active users, 3 runs/day, 250 working days. 18 documents/run. Avg doc: 6,000 tokens.

Default model: claude-sonnet-4

Trigger: New opportunity reaches the proposal stage in Salesforce, plus inbound RFPs. Typically 2-4 proposals per rep per week with 2-5 revision runs each.

Who runs it: B2B sales and pre-sales teams at growth-stage SaaS companies. A rep pulls together a tailored proposal or RFP response per live opportunity; revision cycles with sales engineering add several more runs before it ships.

What Firedog shows

  • Per-run cost split across model, Salesforce/Box orchestration, retrieval, and revision history with exact token counts.
  • How much of each proposal's spend is re-embedding an unchanged Box library versus new synthesis.
  • Projected annual saving from caching the proposal template prefix and switching Box to incremental embedding updates.

Decisions this informs

  • Cache the 8,000-token proposal template and brand-voice prefix via Anthropic prompt caching; at 4 calls/run this drops that input cost by roughly 10x.
  • Switch the Box library to incremental embedding so only newly added proposals are re-embedded, not the whole corpus each run.
  • Store prior drafts as structured section diffs instead of re-injecting full text on every revision pass.

Understand every AI dollar.

Book a demo to see how Firedog attributes every AI request across your organization, and what your leadership team can do with the data.