Perspectives
Realism is a red herring
How human the AI customer sounds is not the curriculum. What moves ability is how you respond, how we analyse it, and whether practice is designed so you sell better in the real world.
The claim
How chatty the AI customer sounds is not what makes you better at sales.
A slightly more colloquial buyer line is fun in a demo. It barely changes how you answer - and your answers are the whole point of practice. What you said, what you missed, and whether the session was designed to fix that - that decides if you leave stronger.
You already know AI still sounds like AI sometimes
If you have used LLMs for a few years, you have heard it: great at many jobs, and still often recognisably not fully human. That gap is real. It is also not where sales training should put its ego.
Wrong centre of gravity
Make the pitch “we sound more human”
Some products treat closing that last gap in phrasing - more colloquial, less “AI” - as the main story. Impressive in a demo. Weak as a theory of how people get better at selling.
Right question
Did the learner’s ability move?
Past a solid bar for believable conversation, making the customer’s wording sound a bit more human does not change how you respond - so it does not change sales-learning outcomes. The work that does is analysis, critique, and pedagogy.
Flashy proof points are easy to sell
Nice voices. Beautiful UI. Fine-tuned models so the buyer sounds like your sector. Impressive demos. Often the wrong centre of gravity.
Easy to advertise
Theatre of realism
If the pitch leads with how human the AI customer sounds, you are being sold cosplay - not a curriculum.
Harder to advertise
Whether ability moved
Scoring that is not arbitrary. Feedback that names the skill. Modes that match load. Practice that returns so memory sticks.
Poor to excellent is about the seller
What separates weak from strong salespeople is how they respond: what they say, what they extract, how they recover. Not how perfectly human the buyer’s small talk sounds.
More like real life
“Hmm, yep this is Tom - who’s this?”
More formal / “AI-ish”
“Hello, this is Tom. Thanks for calling. Who is this?”
The second line is exaggerated on purpose - real systems are better than that. Still: does either line change how you, the seller, should open, qualify, or handle what comes next? Almost never. Past a good-enough bar for believable conversation, extra polish on wording does not add learning. The underlying deal does: objections, withheld facts, timing, complications.
Objections on time
Pressure arrives when it should - so you practise handling it, not scoring the accent.
Facts you must earn
Important detail withheld until great questioning finds it. That is the drill.
Your line under load
The flow has to feel like a real deal. The grade is still on what you did.
Behavioural fidelity moves the needle. Obsessing over whether every phrase sounds perfectly colloquial - past “good enough” - is a flashy priority, not a learning one.
Good enough still has to be excellent
None of this means a weak, unbelievable buyer is fine. Past the bar, more colloquial polish stops buying learning. Below the bar, practice collapses. VendAce aims well above that bar - voices, UI, and conversation design included - and still refuses to treat that polish as the product.
Must clear
Believable deal behaviour
Objections, withheld facts, timing, nuance - a customer that acts like a real conversation, not a script reader.
Where the weight goes
Your responses and the teaching system
How we analyse what you said, what we recommend, how we critique you, and whether the pedagogy makes you better next time - in the real world.
Where Infinite Arc puts the weight
If a team leads with fine-tuning for sound-alike buyers and under-invests in pedagogy, scoring, and mastery - they are proving they do not understand the job.
What you did wrong
Specific enough to change the next attempt - not a vague vibe summary.
How to improve
Modes and coaching that teach the move - not only replaying the call louder.
Memory and mastery
Spacing, return, difficulty - so ability sticks instead of peaking once.
Analysis that holds up
Consistent enough for a person - and for a team. Not arbitrary scores.
Related: Outcomes over applause · How VendAce evolved · The science behind VendAce · DIY chat is not training
Judge what the product prioritises.
Try the modes and the feedback. Conversation quality should clear a high bar - and still not be the headline over learning.