# Is your AI executing a strategy that still makes sense?

Fabio Munhoz · 2026-11-20

Canonical: https://fabiomunhoz.com/blog/ai-commercial-strategy/

Language: en

Automating a commercial decision also requires preserving the conditions for reconsidering it.

## In brief

Checking an agent’s execution and reviewing a commercial strategy are different tasks. Record the assumptions behind customer and offer choices, the signals that call for review, and who can change direction. When the decision changes, update the context agents use.

Imagine a company that chose to focus prospecting on a specific segment. At the time, interviews suggested an urgent need, the product appeared to fit, and the team could reach buyers.

Months later, agents are still researching companies in that segment, preparing messages, and organizing replies. Execution has improved. The choice of market has stayed the same.

This is an illustrative scenario, but the question applies broadly: when did someone last examine the assumption directing the work?

In the [article on coordinating agents](https://fabiomunhoz.com/blog/who-coordinates-ai-agents/), the challenge was helping decisions across teams respect the same context. Here, the question sits one level above that: whether the shared direction remains appropriate.

## Efficiency needs a commercial reference point

A task can be completed correctly while contributing little to the business. Researching contacts in the selected segment fulfills the request. It does not demonstrate that the segment remains the right choice.

I would separate the review into two parts. The first checks execution: did the agent use permitted sources, follow the criteria, and deliver something usable? The second examines the business decision: do those criteria still describe customers we want and are able to serve?

This separation protects the analysis. Without it, a poor market choice can be blamed on the model, while a sound choice can conceal execution problems.

It also clarifies responsibility. The person approving the strategy needs to know when its assumptions are being challenged. The team operating agents needs a way to present that signal without independently changing the commercial direction.

## Make strategic assumptions open to review

An ideal customer profile often describes industry, company size, and organizational characteristics. To support learning, I would add the reasoning behind that selection.

Which problem matters to the buyer? What would make them act now? Why would our offer be a suitable alternative? What evidence could show that we are wrong?

Instead of recording only “prioritize this industry,” the team might write: “we believe companies in this situation struggle to integrate their operations and are willing to invest in a solution.” That statement still needs evidence. Its benefit is that the team can discuss what it is testing.

Contextual assessment also appears in the NIST AI Risk Management Framework: decisions to use AI should consider purpose, risks, impacts, costs, and benefits. This does not determine a commercial strategy, but it reinforces the need to evaluate an application in its actual context. [Source: AI Risks and Trustworthiness, section 3](https://airc.nist.gov/airmf-resources/airmf/3-sec-characteristics/).

## Look for evidence against the current choice

Return to the example. Replies to outreach increase, but conversations reveal a change: buyers need to solve a problem the offer does not yet address.

One possible response is to adjust the writing to secure more meetings. Another is to examine whether the promise attracted a need different from the one the company intends to serve.

The review should recover what buyers said, under which conditions, and what happened afterward. An isolated objection does not represent the entire market. Several similar objections may also come from the same contact source or the same sales script.

Preserving competing explanations matters. The segment may be wrong. The offer may be unclear. The person contacted may not participate in that decision. Each explanation calls for a different investigation.

## Decide when to optimize and when to revisit direction

I would document the boundaries before expanding automation. Formatting adjustments can remain within the approved process. Changes to the audience, promise, or commercial terms need to return to the person accountable for strategy.

A simple record can include the current hypothesis, supporting evidence, signals that trigger a review, and the decision owner. It should also include a review date. Waiting for a major failure is an expensive way to discover that nobody was monitoring an assumption.

When direction changes, update the context agents use and identify the decision version. If older documents remain available without a validity indicator, the next execution may retrieve guidance that leadership has already abandoned.

A review also needs an outcome that people can act on. Keeping the same direction is a valid decision when the evidence supports it. Asking for another test is valid when uncertainty remains. In either case, record the reason so the next review does not restart the same debate.

## A strategy review with an operational consequence

Suppose the team finds interest in the selected segment, but only when the offer includes an integration that is not yet available. The record should not end with “we need better messaging.” The constraint concerns the promised delivery, and the next decision needs to account for it.

Leadership could consider developing the integration, pursuing a different buyer profile, or adjusting the offer's scope. Each option affects delivery capacity, cost, and timing. An agent can organize evidence and prepare alternatives. Choosing among them requires commercial authority and participation from the people who would deliver the work.

After the decision, check its application in a new task. Has the agent stopped selecting accounts that depend on the integration? Does the sales material describe the correct scope? Can the account owner explain the change?

This application check matters because updating a document does not guarantee an operational change. A review reaches execution when the criteria, sources, and responsibilities used in daily work reflect the current decision. Preserve the evidence behind the change as well, so future recommendations can be assessed against the same history.

## A question for the next leadership meeting

In this series, the proposed role of a Revenue Harness includes making the relationship between hypothesis, execution, and outcome visible. That allows leadership to reconsider both how work gets done and whether it should continue.

This does not require changing strategy after every disappointing result. It requires recognizing what the evidence supports and which test could reduce the remaining uncertainty.

**Which assumption about your customer, offer, or market must remain true to justify what your agents are doing today?**

If your team can answer and show how it monitors that assumption, automation serves a direction that can be examined.

## Questions and answers

### Why does good AI execution not prove that a strategy is sound?

An agent can correctly follow its instructions without demonstrating that those criteria still describe the right customer or market. The commercial choice needs its own review.

### Which assumptions should a team monitor?

Record the problem that matters to the buyer, the reason to act now, the fit of the offer, and evidence that could challenge the choice.

### When should a change return to leadership?

Changes to audience, promise, or commercial terms should follow the owners and boundaries established by the company. Operational adjustments may remain within the authorized process.
