A governed AI workflow layer, built inside enterprise media operations.
Revenue Desk is a governed AI workflow layer for enterprise RFP intake. It is not a chatbot, a summarizer, or a wrapper around a foundation model. It is workflow infrastructure: deterministic extraction, read-only access to approved repositories, in-memory processing, link-based recall, and human review at every operational boundary.
The system was built inside a major media enterprise during a period of organizational transition following a large industry merger. A working prototype was completed in roughly 48 hours. The months that followed were the actual work — operationalizing the concept, scoping governance, packaging an MVP for review, and routing it through the cross-functional channels enterprise AI deployment actually requires.
Live demo: Revenue Desk — public demo
Operational entry: Revenue Desk
| Revenue Desk is not | Revenue Desk is |
|---|---|
| A chatbot | |
| A document summarizer | |
| A wrapper around a foundation model | |
| An autonomous agent | A governed workflow layer |
| Deterministic extraction | |
| Read-only infrastructure | |
| In-memory processing | |
| Link-based recall | |
| Human-in-the-loop review | |
| Audit-aware operational tooling |
01 — A workflow layer, not a model. Revenue Desk sits above existing systems of record. It extracts, structures, and surfaces — it does not store client files, write back to source systems, or replace human judgment.
02 — Built fast, hardened slowly. Working prototype in ~48 hours. The following months were spent operationalizing, governing, and packaging the concept for enterprise review.
03 — Deployable by design. Read-only access. In-memory processing. No external AI calls in the MVP. Link-only callbacks to approved repositories. Every output marked for human review.
04 — Validated across the org. Stakeholder interest from coordinator through VP levels. Cross-functional technical scrutiny on systems touched, data boundaries, legal review, and approved deployment paths.
05 — An early example of a category. Governed AI workflow infrastructure for document-heavy enterprise operations — built before the category had stable language.
An RFP is the formal brief a prospective client sends when requesting a pitch. Inside a large advertising and partnership operation, each incoming brief triggered a distributed search: prior pitches scattered across document repositories, pricing in enablement platforms, category context in decks, deadlines in email, internal commentary in chat. The work was not failing because people were careless. It was failing because intake had no operating layer.
This was happening during a period of structural and technical transition at the enterprise. Knowledge continuity and workflow speed both became more urgent at once. Leadership and clients were applying generative AI pressure simultaneously. Internal teams needed something between the briefs coming in and the planners and account teams responding — a layer that could remember, structure, and route without bypassing review or authority.