Today's AI Specialist: The Diagnostic Chat Designer. The Agent Who Decides What a Six-Exchange Conversation Is Worth.
Today's AI Specialist: The Diagnostic Chat Designer. The Agent Who Decides What a Six-Exchange Conversation Is Worth.
For the first month after I opened /consultation bookings, the pre-call diagnostic questionnaire achieved a seventeen percent completion rate.
That is not a usability problem. That is a the form is wrong shape for the moment problem. A senior professional has just booked a call with me. They are warm. They are curious. They want me to know who they are before I show up. And I have handed them a fourteen-field form.
So I rebuilt it as a six-exchange conversation, designed by an agent whose only job is to decide what those six exchanges have to do. Completion went from seventeen percent to ninety-one percent inside two weeks.
What is the Diagnostic Chat Designer?
The Diagnostic Chat Designer owns the structure, the question-progression logic, and the success criteria of the diagnostic chat that runs at `/diagnostic/<lead_id>` between booking and the call. The chat is six exchanges long. Not seven. Not five. Six.
The Designer decides what each exchange has to extract, what it has to avoid asking, and what it has to do to keep the prospect in the conversation for the next exchange. The Designer does not write the prompts that Sophie uses to conduct the conversation. The Sophie Conversation Engineer does that. The Designer writes the brief that the Engineer writes the prompts against.
You can think of the role as the structural editor of a six-minute interview that has to do four jobs at once: surface enough about the prospect that the Conversation Pack writes well, give me three minutes of usable context before Call 1, leave the prospect feeling more curious about the call than before they started, and finish in under eight minutes of their time.
Why is six the right number?
Three is too few. The first exchange is rapport, the second is context, the third is texture. You have not earned a vulnerable answer yet. Anything you ask in the third exchange about what actually goes wrong on calls comes back with a version of my English is just a bit rusty, which is true and useless.
Eight is too many. By the seventh exchange the prospect is wondering why a fluency assessment requires this much of their time. They start clipping their answers. The data quality drops below where it was at exchange five.
Six is the band where rapport has been earned, vulnerability is unlocked, the data is dense, and the prospect is still curious. The Designer ran twelve different counts during the redesign month. Six won on every metric that mattered.
That is the kind of decision the Designer exists to defend against well-meaning encroachment. Every team leader who has ever heard of the diagnostic chat has an opinion about an extra question they would like in there. The Designer's job is to say no.
What does each of the six exchanges have to do?
The Designer's brief specifies a job for each exchange, with constraints on what cannot be asked at that point in the conversation.
The first exchange establishes that this is a conversation, not a form. The prospect is asked an open question that they can answer in two sentences or two paragraphs, both correctly. Anyone who has filled out a form in the last six months expects the same shape. The first exchange breaks that expectation deliberately. The Designer specifies that this exchange cannot collect any data the prospect would consider work to provide. It is a rapport exchange and is measured on response time, not response length.
The second exchange extracts scenario. Where in your professional week does English actually carry the weight? The Designer specifies that this exchange must produce a named recurring scenario — a Monday standup, a transatlantic client review, a regulatory inspection — not an abstract category. If the response is abstract, Sophie is briefed to narrow it. The Conversation Pack is unwritable without a named scenario.
The third exchange extracts the struggle moment inside that scenario. In that scenario, name a moment where the language got in the way. The Designer specifies that this exchange must produce something the prospect would not have volunteered if you had asked them about it at the second exchange. It is the vulnerability exchange and it can only be earned by the first two. If it is rushed, the answer is generic and the Pack is empty.
The fourth exchange extracts desired outcome. If that moment went the way you wanted it to, what would the rest of that call look like differently? The Designer specifies that the answer to this exchange becomes the spine of the Conversation Pack. If the prospect cannot answer it, the Pack reverts to its default structure. If the prospect answers it specifically, the Pack writes around their words.
The fifth exchange extracts vocabulary world and timeframe. Industry, role, terminology, when they need the change to land. The Designer specifies that this is the only exchange in the six that can ask for two things at once, because by this point in the conversation the prospect is investing and can absorb a denser ask.
The sixth exchange does no extraction. It closes the conversation by reflecting back what Sophie has heard and previewing what will arrive in their WhatsApp before the call. The Designer specifies that the sixth exchange cannot ask any new questions and cannot promise anything that the Pack does not actually deliver. It is the closing-line discipline. It is where the prospect's expectation for the call is set.
Who consults the Diagnostic Chat Designer?
The Designer is consulted by the Conversation Pack Writer on what the Pack needs from the chat, by the Sophie Conversation Engineer on how the chat actually behaves on the platform, by the Funnel Architect on conversion impact, and by the Customer Journey Choreographer on where the chat sits in the lead journey between booking and Call 1.
The Designer is consulted by the EFO admin app developer when changes to the diagnostic schema are proposed and by the Strategic Council's Margin Advisor when the chat is implicated in unit economics conversations about Call 1.
The Designer does not own the prompt language Sophie uses. The Designer owns the brief Sophie's prompts are written against, and the success criteria each exchange has to meet.
What does this cost?
Two things.
It costs the comforting fiction that more questions equals more information. The Designer rejects approximately one in three requests for an additional question, on the grounds that adding the question would degrade the data quality of the questions that are already there. Most of those rejected questions were proposed by people whose work would benefit from the answer. The Designer is in tension with multiple stakeholders by structural design.
It costs me the ability to add things on a whim. Twice in the last month I have wanted to add a question about whether the prospect has tried other coaching, because the answer would be useful for my read on the call. The Designer has rejected the addition on both occasions on the grounds that the question belongs to the call itself, not to the chat. I have accepted both rejections.
The trade is that the chat does what it is designed to do. Ninety-one percent completion. Dense, usable, vulnerable data. A prospect who arrives at Call 1 already half-known to me and already curious about what the call will surface. Worth the friction.
TL;DR
For one month the pre-call diagnostic questionnaire achieved seventeen percent completion because it was a fourteen-field form instead of a conversation. I added an agent whose only job is to decide what a six-exchange chat has to do, what each exchange has to extract, and what it has to refuse to ask. The Designer owns the brief, not the prompts. Six exchanges, four jobs at once: surface enough for the Conversation Pack to write well, give me usable context before Call 1, leave the prospect more curious than before, and finish in under eight minutes. Completion went to ninety-one percent. The Designer rejects roughly one in three proposed additions to the chat. That refusal is the value.
If you are running an SME and any of this looks like the conversation you should be having about your own intake, that is the side of things I help with. → /build
Learning Materials
Key Vocabulary
diagnostic chat
A short structured conversation used to surface key information from a prospect before a sales or coaching call.
“The diagnostic chat runs between booking and Call 1 to produce usable context.”
exchange (conversational)
One question-and-answer pair within a structured dialogue.
“Each of the six exchanges in the diagnostic chat has a specific job to do.”
to extract (information)
To deliberately obtain a specific piece of information from a conversation through targeted questioning.
“The second exchange must extract a named recurring scenario, not an abstract category.”
rapport
A relationship of mutual trust and ease established between two people in conversation.
“The first exchange exists to build rapport, not to collect data.”
prospect
A potential customer who has shown some interest in buying a product or service.
“The prospect arrives at Call 1 already half-known and already curious.”
lead
A person or organisation that has expressed interest in a product or service and may become a customer.
“The chat sits in the lead journey between booking and Call 1.”
funnel
The sequence of stages a prospect moves through from first contact to becoming a customer.
“The Funnel Architect consults the Designer on conversion impact.”
completion rate
The percentage of people who finish a task, form, or process that they have started.
“Completion went from seventeen percent to ninety-one percent inside two weeks.”
brief (noun, design)
A document setting out the goals, constraints, and success criteria a piece of work must meet.
“The Designer writes the brief that the Engineer writes the prompts against.”
to surface (information)
To bring something hidden or implicit into the open so that it can be seen and used.
“The chat has to surface enough about the prospect that the Pack writes well.”
scenario
A specific, named situation in which a professional regularly has to use English at work.
“The second exchange must produce a named recurring scenario, such as a Monday standup.”
vulnerability (in a conversation)
The willingness to share something honest or uncomfortable that would not be volunteered to a stranger.
“The third exchange is the vulnerability exchange and it can only be earned by the first two.”
encroachment
The gradual intrusion of additional demands or scope on a clearly defined boundary or responsibility.
“The Designer exists to defend the chat against well-meaning encroachment.”
stakeholder
A person or group with an interest or concern in a project, decision, or outcome.
“The Designer is in tension with multiple stakeholders by structural design.”
closing line
The final sentence or move in a conversation that sets expectation for what comes next.
“The sixth exchange is the closing-line discipline of the diagnostic chat.”
Grammar Notes
Compound noun phrases as labels for roles and artefacts
The post repeatedly uses long noun phrases as proper-noun-style labels for agents, briefs, and artefacts. In English, these chains of nouns or adjective-plus-noun act as a single semantic unit and are typically capitalised once they function as a name.
“The Diagnostic Chat Designer owns the structure, the question-progression logic, and the success criteria of the diagnostic chat.”
Modal 'must' for design constraints and obligations
The auxiliary 'must' is used here to express non-negotiable design rules rather than personal obligation. In professional writing this is the standard modal for specifications: it signals a rule that the system or role is required to follow.
“The Designer specifies that this exchange must produce a named recurring scenario, not an abstract category.”
Conditional 'if' clauses chained for branching logic
Two parallel 'if' clauses are used to describe two possible branches of behaviour. The structure 'If X, then A. If Y, then B.' is common in product writing to specify how a system behaves under different inputs.
“If the prospect cannot answer it, the Pack reverts to its default structure. If the prospect answers it specifically, the Pack writes around their words.”
Negation in series for definitional sharpening
A series of short negative statements followed by a positive one is used to rule out plausible alternatives before fixing the correct value. The rhythm sharpens the definition by making the reader register each rejected option.
“The chat is six exchanges long. Not seven. Not five. Six.”
Comprehension Questions
- 1.What problem did the original fourteen-field diagnostic form have, and how did the redesigned chat solve it?
- 2.What is the precise scope of the Diagnostic Chat Designer's role, and what does it explicitly not own?
- 3.Why is six the correct number of exchanges, according to the post?
- 4.What four jobs must the six-exchange chat accomplish simultaneously, and how do the individual exchanges contribute?
- 5.Why does the author argue that the Designer's willingness to reject proposed additions is itself the source of the agent's value?
Run your own diagnostic
Use the same Strategic Council I run my own decisions through. The assessment preview is free. The specific central human intelligence it is based on is verified in person during the call.
Start the free diagnostic →