Who Was Allowed to Promise That?

The line in the notes that nobody can honor

Three weeks ago there was a customer call. Somewhere in the middle of it, an engineer answered a direct question and said the export API would be ready by the end of the quarter. The customer wrote that down. The account manager wrote it down. The note from the meeting says the export API ships by end of quarter, and it does not say who said it, or on what basis, or whether anyone else in the company had heard about it.

The engineer was estimating. They had looked at the work once, thought it was a few weeks, and answered honestly instead of deflecting. They do not set the roadmap and would never claim to. But nothing about the sentence, as it landed in the room or as it landed in the notes, carries that distinction.

By the time this surfaces, the customer has built a launch plan on it. Somebody now has to explain that the company never agreed to that date. The explanation is completely true and it does not help at all, because from where the customer is sitting, the difference between what the company decided and what the company's engineer said out loud on a call is not a difference they can see.

Nothing in the room tells you what a sentence is worth

Meetings run on statements that carry wildly different weight and sound almost identical. "I think that's a couple of weeks of work." "We should be able to get that to you by the end of the month." "Yes, we can do that." The first is a guess, the second is an intention, the third is a commitment, and in the room they are separated by nothing more than tone and how confident the speaker happened to sound.

The speaker usually knows which one they meant, at least roughly. The listener has to infer it from seniority, delivery, and their own hopes. Then the note gets written and the last of the signal disappears, because a guess and a commitment both become one bullet under next steps, phrased in the same neutral voice.

The word worth borrowing here is standing. Standing is whether a statement was a guess, an intention, or a commitment, and whether the person was speaking for themselves or for a group. Standing is the single most useful property of anything said in a meeting, and it is the property that every summary destroys first.

This is not a problem of people overclaiming. Most people are careful. The care lives in how they say it, and how they say it is exactly what does not survive into the record.

The listener's belief is what does the binding

Contract law has spent a very long time on this exact asymmetry and landed somewhere most operators find uncomfortable. Under the Restatement (Third) of Agency, published in 2006, apparent authority is "the power held by an agent or other actor to affect a principal's legal relations with third parties when a third party reasonably believes the actor has authority to act on behalf of the principal and that belief is traceable to the principal's manifestations."

Read that carefully. The test is not what the person was actually allowed to do. It is what the other side reasonably believed, and where that belief came from. The comment to the same section says the definition applies to actors who appear to be agents and are not, "as well as to agents who act beyond the scope of their actual authority." Acting past your remit is the case the doctrine was built for.

There is a real limit, and it matters. Apparent authority traces to the principal's manifestations, not the agent's own claims, so nobody can talk themselves into authority they were never given. But putting a person on the call, with a title, fielding questions about the product, is a manifestation. The company put them there.

None of this means a customer conversation is a contract, and nobody should read a blog post as legal advice about their own situation. The point is that the doctrine exists because the underlying asymmetry is real and old. The person hearing the promise cannot see the org chart. They can only see who the company sent.

The company never says who speaks for it

Organizational research reaches the same place from the opposite direction. In their review of the psychological contract literature, Coyle-Shapiro and Parzefall note that the employer side of any working relationship is usually represented by multiple agents, and that organizations recruit, socialize and offer inducements "without specifying who personifies the organization in these activities." There is no moment where a company announces who is allowed to speak as the company.

It gets stranger. They cite work by Guest and Conway arguing that a manager only counts as a legitimate representative if the manager perceives themselves that way, and separately that employees often do not treat their line manager as speaking for the organization unless that manager sits high enough in the hierarchy. So both people in a conversation can be wrong, in opposite directions, about whether the company just said something.

That is worth sitting with, because it means the confusion is structural rather than a lapse by anyone involved. Nobody is being sloppy. The information required to answer "was that the company talking?" is not present in the room, and it is definitely not present in the notes.

Inside a company it is worse than outside. An external party at least knows they are dealing with an organization and might ask who signs off. Internally, everyone assumes shared context, so a commitment from another team gets absorbed into a plan without anyone thinking to ask whether the person who said it can spend the resources it will take.

A promise is not an expectation, and the record keeps neither

The same literature draws a distinction that turns out to be the whole problem in miniature. An expectation is a general belief that something will or should happen. A promise is a specific belief that something will happen, formed because of communication or behavior indicating an intention to do so. Only obligations that come from an explicit or implicit promise create the felt contract; a general expectation does not.

So the difference between a promise and an expectation lives entirely in how a statement was made, not in what it was about. Which is to say it lives in the exact material that a recap discards. "Export API by end of quarter" is a sentence that could have come from either, and once it is written that way the evidence needed to tell them apart is gone.

This is also why having the transcript does not fix it. A transcript preserves the words, and the words were never the ambiguous part. Standing is about who was speaking, what they were answering, and what they believed they were signing up for, which is context wrapped around the words rather than the words themselves.

Silence is how it becomes the company's position

Agency law has one more move that maps onto office life uncomfortably well. Even where an agent acted with no authority at all, the principal becomes bound if it ratifies the act. Express ratification is a statement. Implied ratification, which is the common one, happens when the principal knows about the unauthorized transaction and accepts the benefits of it. And ratification takes effect as soon as the principal objectively manifests acceptance, even if that acceptance is never communicated to anyone.

The internal version of that is almost completely silent. The date gets mentioned in a standup. It appears in a status update because it was in the notes. Customer success builds an onboarding timeline around it. Marketing pencils in a moment for it. Nobody ever said yes. Nobody said no either, because saying no requires first noticing that a decision was made, and nothing about the way it entered the system looked like a decision.

That reframes where the failure actually is. It is not the sentence in the meeting. Someone will always answer a direct question with more confidence than the situation supports, and that is not a fixable property of humans in rooms. The failure is the three weeks afterwards, where the only thing standing between an unauthorized commitment and a real one is whether somebody with the standing to decline it happened to see it go by.

Every company already has this problem. What varies is how long the gap is between the sentence and the first person who could have said something.

What a record has to hold

Most meeting notes capture the content of a commitment and nothing about its provenance. Six fields fix that, and none of them require anyone to behave differently in the meeting itself.

• The commitment in the words the person actually used, not a tidied version. "Should be a couple of weeks" and "we will deliver in two weeks" are different facts and only one of them was said.
• Who said it, and what their role is relative to the thing being promised. An engineer promising a date, a support lead promising a policy change, and a founder promising pricing are three different situations.
• Whether they were answering for themselves or for a team, if that was stated. If it was not stated, record that it was not.
• The standing they gave it: guess, intention, or commitment. When the meeting never established this, the record should say so, because nobody asking is itself the finding.
• What has to be true for it to hold, and who else would have to agree. Conditions are the first thing a summary throws away and the first thing anyone needs back.
• A date by which it gets confirmed or withdrawn.

The last field is the one doing the work

Five of those six are description. The sixth is a mechanism. A confirm-or-withdraw date forces the commitment in front of someone who can accept it or kill it, on a schedule, rather than leaving it to the chance that the right person reads the right note.

The obvious alternative is an approval gate: nobody commits anything externally without sign-off. It sounds responsible and it does not survive contact with a real meeting. The customer asks the engineer, because the engineer is the one who knows. A rule that the engineer must not answer produces "I'll have to check on that" as the universal response, including for the many questions where a straight answer would have been fine and useful.

There is a second cost to the gate that is easy to miss. The person who overcommits on a customer call is very often the most engaged person on the call. They cared enough to answer. If the visible consequence of answering is being singled out, the behavior you select for is silence, and silence in customer calls is expensive in ways that never show up as an incident.

So the design goal is cheap correction rather than prevention. Keep the standing attached to the statement, route it to the person who can rule on it, and let them ratify or replace it within days. Either it becomes a real commitment, which is fine, or it gets swapped for an honest "we looked at it properly and here is the actual date", which is a completely survivable conversation in week one.

What it costs to take something back, over time

In week one, withdrawing a commitment is a correction. Nothing has been built on top of it and the other side is mildly annoyed at worst.

In week four, it is a renegotiation. Plans exist. Someone has told their own boss. The conversation is now about what you will do instead.

In month three, it is a broken promise, and it does not matter at all what the org chart says about who was empowered to make it. The other side heard the company. As far as the relationship is concerned, the company said it.

The words in the retraction never change. What changes is how much has been stacked on top of the thing being retracted, and that stacking happens quietly, in status updates and plans, during the exact period when nobody is looking at the original sentence.

Teams tend to file this under discipline, or communication, or hiring more careful people. It is a memory problem. Everything needed to catch it existed at the moment the sentence was spoken and got thrown away in the first summary written afterwards. A record that keeps who said it, on what basis, and what still had to be true buys a company a few days of cheap options that it currently does not have. That is what work memory is for: Driffle keeps the meeting, the screen context, and the follow-up attached to each other, so a commitment and its standing stay together instead of arriving in someone's plan as a bare date with no history.