EiGENRA Back-Office on Three AI Agents

Departmental ops automation, demonstrated on our own books. Three orchestrated agents — invoicing, client pulse, proposal drafting — run unattended. The same playbook we run for SME services firms, with a public ROI log.

ⓘProject description and metrics shown here are representative — drawn from a real engagement, with the client name anonymized where confidentiality applies.

Most AI consultancies sell transformations they wouldn't bet their own payroll on. We wanted to flip that. Could EiGENRA's own back office — invoicing, client-health monitoring, proposal drafting — be replaced by AI agents we'd be comfortable shipping to a client? Specifically: weekly invoice cycle dragging founder time, no early-warning system for renewal risk, and proposal turnaround that bottlenecked deal flow at the calendar instead of the conversation.

Three orchestrated agents on shared Vercel KV state. Agent 1 (invoicing): a Friday cron pulls active engagements, drafts invoices in your voice via Claude Haiku, sends through Resend, and runs the dunning ladder with warm / firm / final tones — webhook-backed Stripe with the same dedup pattern shipped on Mama Hala. Agent 2 (client pulse): a Monday-morning Sonnet narrative over open tickets, last invoice status, last deploy, last admin login — flags anyone trending the wrong way before they churn. Agent 3 (proposal drafting): a tool-using Sonnet loop that takes a discovery-call transcript plus our scope library and produces a first-draft proposal the founder reviews and signs. Prompt caching on the system message and the scope library keeps model spend under $40 a month. Every agent degrades gracefully — model failure never blocks the underlying flow.

// Proposal-drafting agent — Sonnet tool loop, prompt-cached scope library
const SCOPE_LIBRARY = await loadScopeLibrary(); // cached in KV

const draft = await anthropic.messages.create({
  model: "claude-sonnet-4-6",
  max_tokens: 4000,
  system: [
    { type: "text", text: SYSTEM_PROMPT, cache_control: { type: "ephemeral" } },
    { type: "text", text: SCOPE_LIBRARY,  cache_control: { type: "ephemeral" } },
  ],
  tools: [matchScope, draftMilestones, priceFromRules],
  messages: [{ role: "user", content: transcript }],
});

// Tool loop: max 6 turns, escalate to admin if unresolved
for (let turn = 0; turn < 6 && draft.stop_reason === "tool_use"; turn++) {
  const result = await runTool(draft.content);
  // ... continue
}

// Output: pre-filled proposal, founder reviews and signs.
return formatProposal(draft);

Departmental ops automation running on EiGENRA's own books since launch. Three agents share Vercel KV state: a Friday invoicing cron with a warm / firm / final dunning ladder, a Monday client-pulse narrative over open tickets, invoice status, deploys and admin logins, and a tool-using proposal drafter fed by our own scope library. Model spend held under $40/month thanks to prompt caching. Public build log updated weekly. Every pattern in this case study is reusable in your business — that's the offer.