# EPIC Revenue Harness: our approach to operating and learning from revenue

Fabio Munhoz · 2026-11-20

Canonical: https://fabiomunhoz.com/blog/epic-revenue-harness/

Language: en

A proposal for connecting commercial direction, context, execution, and review in a process a company can follow.

## In brief

EPIC Revenue Harness is presented as an approach in development for connecting commercial decisions with execution and learning. It is not an announcement of a finished platform or commercial package. Each process needs reviewable context, accountability, acceptance criteria, and evidence before autonomy expands.

The idea behind this series is straightforward: a revenue operation should be able to explain its decisions, execute within known boundaries, and use outcomes to improve the next decision.

That is the direction I propose for EPIC Revenue Harness. This article presents an approach being developed for EPIC. It does not announce a finished platform, a finalized commercial package, or a set of integrations already available to every company.

The distinction matters to anyone considering a partnership. A clear direction makes it possible to discuss the work required. Each capability still needs to be demonstrated in the process and environment where it will be used.

## Start with the business decision

I would begin with a concrete question from leadership. Should we enter a new segment? Which offer deserves the next investment? Where are we losing opportunities we could realistically serve?

The initial answer may be a hypothesis, provided the company recognizes its evidence and uncertainty. Strategic work defines the buyer, problem, value proposition, and delivery boundaries. It also establishes what would justify a change of direction.

Category Leap enters this proposal as a lens for examining positioning and market alternatives. A Revenue Harness would support the transition from those choices into operations. This relationship describes the proposed approach; it does not claim that a complete commercial method has already been validated.

## Organize context that someone can review

The next step would be to assemble the knowledge needed for the chosen decision. It might include commercial history, the current offer, service constraints, and evidence about the buyer.

The aim would not be to copy everything the company owns. It would be to identify the right source for each question and assign responsibility for keeping it current. Information without a known origin or period of validity should remain visible as a gap, even when an agent can write a convincing answer.

Agent engineering provides a useful reference. Anthropic describes mechanisms for preserving progress and supporting continuity between work sessions. Applying that idea to revenue requires further adaptation because commercial decisions involve people and systems with different responsibilities. [Source](https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents).

## Execute with clear responsibilities

In the proposed approach, each process would begin with a bounded task and acceptance criteria. Researching accounts, preparing a recommendation, and sending a message would be separate steps when their consequences called for different controls.

Agents could support research, preparation, and follow-up. The extent of their autonomy would depend on the demonstrated process, actual permissions, and the accountable person's decision. Leadership would continue to set objectives and decide which commitments the company can make.

Consider an illustrative example: a team wants to test an offer in a new segment. Before sending anything, it would need to confirm who may be contacted, which promises have been approved, and which situations suspend outreach. The process should record completed execution as well as unfinished work.

This example describes an implementation criterion. It is not a client story or a demonstration EPIC has already completed.

## Bring outcomes back to direction

After execution, the question would be: what does this round allow us to conclude? Outcomes should be examined against the initial hypothesis and the conditions under which the action took place.

A reply can indicate interest without confirming purchase intent. A sale may depend on an exceptional condition. A recurring objection may justify revisiting the offer, but it could also reflect a problem with account selection.

My proposal is for each review to end with an explicit decision: retain the hypothesis, narrow its application, test it differently, or abandon it. That decision would update the next round's context, including the supporting evidence and its limitations.

This connects to the Cognitive Revenue Growth hypothesis: investigating whether a company makes better decisions as it preserves and revises what it learns. There is no guarantee that this cycle alone will produce growth.

## What must be demonstrated before expanding

A diagnosis should identify the priority problem and conditions for implementation. A pilot should demonstrate a complete process, including an exception and recovery. Ongoing operation would require owners for maintenance, context review, and failure handling.

For each stage, scope, timing, cost, dependencies, and acceptance criteria need to be defined for the specific engagement. None can be inferred from the Revenue Harness name.

This series already provides a [working definition](https://fabiomunhoz.com/blog/what-is-a-revenue-harness/) and a discussion of [agent coordination](https://fabiomunhoz.com/blog/who-coordinates-ai-agents/). They support an informed conversation; they do not replace a solution demonstration.

**The first conversation I propose concerns a decision your company needs to improve.** From there, we can examine the missing context, the people who should participate, and the test that would help evaluate the next step. That level of clarity is where an approach starts becoming an executable commitment.

## Questions and answers

### Is EPIC Revenue Harness already a finished platform?

This article presents an approach in development, not a finished platform, finalized commercial package, or integrations available to every company. Each capability needs to be demonstrated in its intended context.

### Where would applying the approach begin?

With a business decision that needs improvement. The team would then examine the required context, responsibilities, execution boundaries, and test that could help evaluate the next step.

### What must be demonstrated before expanding operations?

The diagnosis should identify the problem and implementation conditions. The pilot should demonstrate a complete process with an exception and recovery. Ongoing operation needs owners for maintenance, context, and failure handling.
