A founder can leave two sales calls with the same feeling.

The conversation was engaged. The buyer recognized the problem. Several useful questions appeared. A follow-up was agreed. The meeting felt productive.

But the two calls may have performed completely different jobs.

One call may have changed what the founder believes about the market. The other may have tested whether an already chosen sales motion can move a buyer toward a real commitment.

I keep returning to this distinction while working on Sales Black Box. We are developing it as a customer-intelligence layer for phone-heavy teams. Selling it requires learning about call workflows, management ownership, CRM behavior, follow-up, consent, implementation, and the economics of revenue leakage. It also requires testing whether a specific buyer, problem frame, proof path, and pilot ask can work repeatedly.

Those are related activities. They should not be measured in the same way.

I think founder-led sales needs two explicit modes: learning mode and execution mode.

The distinction sounds simple. In practice, it changes how a founder prepares for a call, what they are allowed to change during it, what they record afterward, and what conclusion the company can draw.

One Conversation Cannot Answer Every Question

Founder-led sales is unusually flexible.

A founder can change the product explanation halfway through a meeting. They can propose a custom workflow, narrow the offer, revise the price, remove a weak claim, answer a technical question, and redesign the pilot before the call ends. That flexibility is one of the reasons founders should be close to early customers.

It also creates a measurement problem.

Suppose a buyer becomes interested only after the founder abandons the original message, invents a new use case, offers a custom integration, and changes the commercial ask. The conversation may be valuable. It may even lead to a deal. But it has not shown that the original sales motion works.

It has shown that the founder can rescue a conversation.

The opposite problem also appears. A founder decides that the script must remain stable because the team is “testing the motion.” They continue presenting the chosen problem even when several buyers describe a different workflow and a different source of value. The test looks disciplined, but the company is protecting a weak assumption from new information.

Early sales needs flexibility and comparability. Trying to maximize both inside every call creates ambiguous evidence.

The practical answer is to decide which mode a call is in before it starts.

Learning Mode Changes The Market Model

A learning call exists to reduce a specific uncertainty.

The uncertainty may concern the buyer, the workflow, the trigger, the consequence, the current workaround, the decision owner, the budget path, the implementation constraint, or the proof required for a next step.

The useful output is not a forecast. It is a better model of the market.

For example, a team developing call intelligence may begin with a broad belief that sales leaders want better transcripts and AI summaries. Learning calls may reveal that the transcript is already available elsewhere. The real management problem is that follow-up ownership disappears, CRM notes are inconsistent, and managers cannot identify which calls deserve review.

That learning can change several decisions at once:

• the problem language on the website;
• the qualification questions used in discovery;
• the workflow shown in the demo;
• the product boundary;
• the pilot baseline;
• the buyer who should own implementation;
• the evidence required before a paid continuation.

This is why early selling often feels like product management, product marketing, and commercial design happening in the same room. Pete Kazanjy’s Founding Sales is useful on this point: the first sales motion is discovered and refined before it can be handed to a larger team.

Learning mode gives the founder permission to change important assumptions. The discipline comes from naming which assumption the call is intended to examine.

“Learn more about the company” is too broad. “Determine whether missed follow-up is visible enough for a sales manager to own” is testable. “Understand their needs” is vague. “Find out whether CRM incompleteness affects a downstream decision” gives the conversation a job.

What A Learning Call Should Preserve

The notes from a learning call should preserve more than a polished summary.

I would want to capture:

• the buyer’s language for the problem;
• the actual sequence of the workflow;
• the tools and informal workarounds already in use;
• the event that made the problem relevant now;
• the operational or commercial consequence;
• the person who feels that consequence;
• the objection or hesitation;
• the proof request;
• the contradictions between what the company says and what it does.

The record should also separate three layers:

1. Direct statement: what the buyer actually said.
2. Observation: what happened in the conversation or workflow.
3. Interpretation: what the founder thinks the signal means.

This distinction matters because early companies are excellent at turning an inference into a fact.

“The buyer has urgency” is an interpretation. “The sales manager said a new reporting process must be selected before the next quarter” is stronger evidence. “They liked the pilot” is vague. “They agreed to provide a call sample, name an implementation owner, and review a baseline next Tuesday” describes behavior.

The article CRM Is Becoming The Memory Layer For Revenue Work develops this principle further. A revenue system becomes useful when it preserves changing account state, source, confidence, and next decision instead of only recording that contact occurred.

Execution Mode Tests The Sales Motion

An execution call has a different job.

The company has a working hypothesis about the buyer, workflow, trigger, consequence, proof, and next step. The call tests whether that motion can create comparable buyer movement across similar accounts.

Execution mode does not mean reading a rigid script.

A good seller still responds naturally, asks follow-up questions, and adapts to context. The important point is that the team knows which parts of the motion are being held stable.

For example:

• the target workflow remains the same;
• the qualification logic remains the same;
• the core problem frame remains the same;
• the proof sequence remains the same;
• the commercial ask remains the same;
• the next-step standard remains the same.

Now the company can ask a useful question: does this motion produce similar forms of buyer commitment?

The evidence is not that the call felt positive. It is a change in buyer state.

A buyer may expose the current workflow. They may quantify a consequence. They may introduce the manager who owns the process. They may request security or implementation proof. They may agree to a baseline review, a pilot design, a letter of intent, or a paid diagnostic. They may also provide a clear disqualifier.

Each outcome is more informative than general enthusiasm because it changes what the company can reasonably do next.

This connects to The Buyer Artifact Behind The Deal. In B2B sales, a useful next step often needs to survive beyond the original call. A workflow map, business case, pilot memo, or proof matrix helps the buyer carry the decision internally. In execution mode, the company’s ability to produce and place that artifact is part of the motion being tested.

Buyer Movement Needs A Standard

Execution becomes noisy when every seller uses a different definition of progress.

One seller advances the opportunity because the buyer asked for pricing. Another waits for an economic owner. A founder treats a second call as commitment. A manager wants a defined pilot. The CRM reports stage movement, but the underlying buyer behavior is inconsistent.

A simple commitment ladder can help:

1. The buyer spends attention.
2. The buyer shares workflow detail.
3. The buyer exposes a consequence or trigger.
4. The buyer involves another stakeholder.
5. The buyer provides data, access, or a sample.
6. The buyer agrees to a bounded test and review process.
7. The buyer makes a commercial commitment.

The ladder should not become a universal scoring system. Different products require different behaviors. Its value is that it forces the team to name the evidence behind the stage.

For Sales Black Box, a useful execution signal would be more than interest in AI call analysis. A stronger buyer movement may include identifying one phone-heavy workflow, confirming a manager owner, showing how calls currently enter the CRM, agreeing on consent and data handling, and defining which downstream outcome a pilot should examine.

That does not prove the product will create revenue. It shows that the buyer has moved from category interest toward an operable test.

The Two Modes Contaminate Each Other

The most common failure is not choosing the wrong mode. It is changing modes without noticing.

Learning enters execution when one objection causes the founder to rewrite the pitch, add custom scope, change the price, and offer a different next step. The account may move, but the result is no longer comparable with other execution calls.

Execution enters learning when the founder protects the current story so strongly that contradictory evidence is treated as poor qualification. The company keeps testing a stable motion against the wrong market model.

Both failures are understandable.

Founders want to win the account in front of them. Teams want clean metrics. Investors and managers want a simple answer about whether the motion works. Early markets rarely provide that answer on demand.

The better practice is to make the mode change explicit.

If an execution call reveals a foundational contradiction, mark it. Finish the conversation responsibly, then return the issue to a learning queue. If a learning call unexpectedly produces a strong commitment, preserve it, but do not pretend the full motion has already been tested.

This keeps the company open to surprise without turning every surprise into a new strategy.

Use A Call Mode Card

Before an important founder-led sales call, I would write a small Call Mode Card.

It needs six fields:

Field	Question
Mode	Is this call for learning or execution?
Primary uncertainty	What do we still need to know?
Stable variables	What are we deliberately not changing?
Evidence sought	What behavior, statement, or artifact would matter?
Next decision	What will this evidence allow us to decide?
Owner	Who must update the product, offer, CRM, or sales motion afterward?

In learning mode, the stable variables may be few. The founder may be willing to revise the segment or problem. In execution mode, more of the motion remains fixed so that outcomes can be compared.

The card is useful because it changes preparation. It also improves review. A call can fail to produce a deal and still complete its learning job. Another call can feel insightful while failing its execution job because the buyer did not move.

Review The Modes Separately

At the end of the week, learning calls and execution calls should not be collapsed into one dashboard.

For learning calls, review:

• which assumption changed;
• what evidence caused the change;
• whether the change affects ICP, product, proof, offer, or implementation;
• what remains uncertain;
• which next conversation can test the revised model.

For execution calls, review:

• whether the intended buyer state changed;
• where the motion stalled;
• which objection repeated;
• whether the proof and ask were used consistently;
• whether the account was genuinely comparable;
• what should remain stable for the next test.

The distinction also prevents misleading conversion analysis. A learning call with an intentionally broad exploratory audience should not weaken the same rate used to evaluate a narrow execution motion. An execution call should not be celebrated as discovery merely because the commercial result was weak.

When To Move From Learning To Execution

There is no perfect sample size that marks the transition.

I would look for repeated structure.

The same workflow appears across several relevant companies. Similar triggers create urgency. A recognizable owner feels the consequence. The current workaround follows a pattern. Proof requests begin to repeat. A bounded next step makes sense to multiple buyers.

That is not product-market fit. It is enough coherence to form a testable sales hypothesis.

The company can then define an execution motion and see whether it works without constant reinvention. If it fails, the failure should send the team back to a specific uncertainty rather than into a general rewrite of everything.

This is the bridge between customer discovery and a first repeatable system.

The Practical Shift

Founder-led sales became more useful to me when I stopped asking only whether a call went well.

The stronger question is: what job was this call supposed to perform, and what changed because it happened?

Learning mode should make the market model better. Execution mode should make the sales motion more testable. Both create value, but only when the company knows which kind of value to expect.

For the next ten sales conversations, classify the mode before the meeting. Write down the uncertainty, the evidence sought, and the next decision. Then review the two groups separately.

For a phone-heavy team, Sales Black Box is connected to the evidence layer: preserving calls, buyer signals, follow-up, review, and CRM context. For a founder still designing the first commercial motion, Proof Engine Grow and the First Paying Customers Program address the wider system.

The goal is not to make founder sales less human.

It is to make the learning and the progress easier to trust.

FAQ

What is the difference between a learning call and an execution call?

A learning call exists to reduce one named uncertainty about the buyer, workflow, trigger, owner or proof, and the founder may change important assumptions during it. An execution call holds the workflow, problem frame, proof sequence and commercial ask stable to test whether the motion produces comparable buyer movement across similar accounts.

Why should a founder decide the call mode before the meeting?

Because founder-led calls are flexible enough to blur the evidence. A founder who rewrites the pitch, scope and price mid-call may rescue the account, but the result no longer shows whether the original motion works. Naming the mode first keeps both the learning and the progress measurable.

What counts as progress in an execution call?

A change in buyer state, not a positive feeling. The buyer shares workflow detail, exposes a consequence, involves another stakeholder, provides data or access, agrees to a bounded test, or makes a commercial commitment. A clear disqualifier is useful evidence too.

When should a founder move from learning mode to execution mode?

When structure starts to repeat: the same workflow across relevant companies, similar triggers, a recognizable owner of the consequence, recurring proof requests, and a bounded next step that makes sense to several buyers. That is not product-market fit, but it is enough to form a testable sales hypothesis.

Sources

• Pete Kazanjy: Founding Sales
• Sales Black Box: Customer intelligence for phone-heavy teams
• Proof Engine: Grow and First Paying Customers Program
• Proof Engine blog: The Buyer Artifact Behind The Deal, CRM Is Becoming The Memory Layer For Revenue Work, and What Is Sales Black Box?