← All projects

Workforce Command

A lead-generation workspace for agencies. It researches each company with sources you can check, drafts proposals from work you have already won and tells you who to follow up with today — and never sends a message on its own.

Role
Founder · Product, design & engineering
Engagement
Personal Project
Category
Product
Industry
Lead Generation
Year
2026
Live site
Preview
6AI modules
4AI providers behind one gateway
9test suites, including tenant isolation

The client

Agencies usually find leads in one tool, research them in browser tabs, write pitches in a doc and track follow-ups in a spreadsheet. Workforce Command puts that in one workspace built for relationship-first outreach: every company fact carries its source and date, proposals are grounded in the agency’s real case studies, and a person copies and sends every message.

What we did

  • Product strategy
  • Brand identity
  • UX & UI design
  • Multi-tenant API (Node.js)
  • AI gateway & Outreach Studio
  • Marketing site

Stack

ReactTypeScriptNode.jsMongoDBLLM APIsOllamaVercelFigma
Inside the product

Key features

Lead workspace

Connections filtered by age, role, industry and priority, each with the best-matching portfolio proof and a status from new to won.

Company intelligence

A directory de-duplicated by domain, with enriched facts shown alongside their source, confidence and as-of date, plus website audits for SEO, UX and speed.

Copy-ready outreach

Open a lead to get a tailored message built from matched work, then log the touchpoint and set the next follow-up date in the same drawer.

Follow-up worklist

Due today, overdue and upcoming queues driven by a configurable cadence, with manual overrides that stick.

Outreach Studio

AI drafts proposals that cite their sources, builds brand boards from a niche, palette and typeface, and recommends type pairings for the client’s industry.

Isolated workspaces

Each agency’s data stays in its own organisation, with roles, an audit log and CSV or JSON export.

Under the hood

How it's built

A lean API

The API runs on Node’s built-in HTTP server and the MongoDB driver with almost no other dependencies, which keeps it small and every route explicit.

Each agency walled off

Every request resolves to a user’s membership in one organisation, data is indexed per organisation and companies are de-duplicated by domain within it. Dedicated tests check one agency can never read another’s data.

Hardened sign-in

Cookie sessions with CSRF checks, rate-limited sign-in and rotating refresh tokens that shut down the whole session if an old token is ever reused.

One gateway, any model

Six AI modules, from proposals and brand boards to website audits and portfolio matching, go through one gateway that can switch between hosted providers and a local model.

Checked output, honest fallback

Model output is parsed and checked for its required fields. If a provider fails, a rule-based generator answers instead and the result is labelled as a demo, never passed off as AI.

Colour maths in code

Brand boards calculate WCAG contrast ratios from the palette in code instead of asking the model, so the accessibility notes are always accurate.