The words Frank uses, one by one.

The person on the other side of the conversation. It has a role, a goal, a character and memory of what was said before, and it answers by voice in real time: it can push, leave a silence open, interrupt or end the meeting. During the session it is only that, a counterpart. It gives no advice and shows no metrics.

How the counterpart behaves

The point where a story forks depending on what happened in the previous session. If you held the price, the next session is the technical close; if you gave up margin, it is winning the term back. The counterpart arrives at the branch carrying what was said before, so there is no starting from zero.

The documents and notes an organization hands to Frank so that scenarios speak about its products, its objections and its numbers: a culture handbook, a discount policy, a sales playbook, or five lines written by hand. Frank reads them and proposes scenarios. Someone at the organization approves the ones that go in.

How to prepare your company context

Practicing with a concrete goal, real friction and feedback at the end, and doing it again. It is the method behind Frank: not a talk about how to negotiate, but several hard conversations a week with a counterpart that pushes and a report at the end of each one. Workshops and coaches provide the frame; what is almost never there is volume of practice.

The evidence behind the method

The difference between how a person sounded in a session and how they normally sound, according to their individual baseline. It is Frank’s unit of measure: it does not say whether someone spoke well or badly in absolute terms, it says what changed relative to themselves. A delta can point out that you sped up when price came up, or that you held your tone through the close when you used to lose it.

A standard set of acoustic parameters (the extended Geneva Minimalistic Acoustic Parameter Set) that voice research uses to describe a signal: pitch, energy, voice quality, rhythm. Frank checks its acoustic engine against the reference implementation of eGeMAPS v02 in openSMILE and against Praat, to verify it measures the same thing the literature does. That comparison is engine fidelity, not validation of results: it does not imply these signals predict performance.

The capabilities a conversation puts to the test when something is at stake: holding an objection, delivering bad news, listening without giving in right away, closing. It is what Frank trains; they are also called interpersonal skills or interpersonal intelligence. Frank never says “soft skills”: there is nothing soft about letting someone go, or holding a price in front of a client who threatens to leave.

How a person sounds when they speak under normal conditions: their rhythm, their energy, their usual tone. It is established in the first session and refined over the following ones. It is the reference Frank compares every session against, which is why an accent, a disfluency or an unusual timbre are not noise: they are the starting point.

A point in the session where something happened that is worth listening to again: the counterpart interrupted, there was a long silence, the person sped up when price came up or held the number through the close. Each moment is marked on the report’s timeline with its timestamp. Frank talks about moments because a real conversation is not counted in numbered exchanges: it is remembered by what happened.

The kind of conversation a scenario sets up. It is not a scale: these are different conversations, and a person can hold one well and not the other. A leader who is comfortable with routine feedback can struggle in a conflict, and the other way around.

Routine

A monthly one-on-one, a check-in with a client who is doing fine. The counterpart cooperates, but gives nothing away.

Tense

A client pushing for a discount, a team receiving bad news. The counterpart holds the objection.

Conflict

A cofounder questioning the direction, a dismissal. The counterpart can raise its voice and cut the meeting.

The study of how something is said, not of what is said: the energy of the voice, the rhythm, the pauses, the variability of pitch. Frank runs it on the audio of each session and always compares it against the person’s own baseline, never against an average or an external ideal. It does not infer emotions or mental states and it feeds no decision about anyone: it describes how a conversation sounded.

What Frank measures, and what it doesn’t

The first cycle with an organization: thirty days, a small group, a scenario co-designed with the company and a confidentiality agreement signed before the first session. It starts with an initial measurement, continues with practice on the program’s stories, and ends with a re-assessment and an aggregated report for the organization.

What a pilot looks like from the inside

What a leader sets up for their team for the coming weeks: a set of stories around one goal, such as commercial negotiation, feedback to the team or conflict handling. The same program can hold stories in Spanish and in English. It is the top unit of the three: program, story, session.

How a program is built

The session used to measure progress. It runs on a comparable scenario the person has not seen, never on the one they just practiced, and it is compared against the baseline established at the start. In a pilot it happens in the fourth week.

The concrete situation practiced in a session: who the counterpart is, what they want, what they know and how far they are willing to go. It is written from the company’s context and has a nature: routine, tense or conflict. Frank proposes scenarios; the organization approves the ones that get used.

How to approve and correct the scenarios Frank proposes

One voice conversation, start to finish, between a person and the AI counterpart on a scenario. It runs 5 to 10 minutes, works in the browser with nothing to install, and is available in Spanish and English. While it lasts there is no coach and no metrics on screen; the report arrives at the end.

What happens in a session

What a person receives at the end of a session: a text written in words, with the key moments marked on the timeline and a concrete recommendation of what to practice next time. It is private to the person who practiced; the organization sees aggregated progress, not the session. It does not tell you whether you were right: it tells you how you said it.

How to read a session report

A situation that moves forward chapter by chapter across several sessions, with the same counterpart and the same account, the same team or the same conflict. The counterpart arrives at each session carrying what happened in the last one, and the story can branch depending on how the last conversation ended. Several stories make a program.