01
What it is
A phone practice system. Your colleague calls an AI customer who behaves like a real one: asks back, objects, and hangs up if the call is going nowhere. Afterward they get a scored review, quoting their own sentences.
Training systems
Practice with an AI partner, not on real customers
The same voice technology, turned inward: your colleague practices with an AI counterpart and gets scored feedback afterward, backed by quotes. Two of the three systems are already in live use.
The whole thing in five points
Enough to decide whether to keep reading.
02
What it is for
Onboarding: the new hire does not learn on your customers. Measurement: you find out who says what today, and how. Product launches: within days you can see where the new material landed.
03
What the manager sees
A team overview and individual progress across five dimensions, with full transcripts. A threshold score you can set: until someone clears it, they do not go to a customer alone. The score is for development — not an appraisal, and not a basis for commission.
04
What we need from you
What you already have: training material, call outlines, typical objections, product details — and whatever makes you call a conversation good. Nothing has to be rewritten, and there is no IT integration.
05
What we suggest
One team, a few weeks, with the metrics agreed up front. Then you decide about the rest based on numbers.
The problem
Four costs nobody measures today
1 · The new hire
Their first fifty calls are spent on real customers
There is nowhere to practice. Real prospects pay the tuition: a botched contact, a bad first impression — and one more person who will not pick up next time.
2 · The mentor
Practice comes out of your most expensive colleague’s day
The person who can teach is also the person who sells. Every practice conversation comes out of their time, so everyone gets a little of it — and most people face their first customer too early.
3 · The existing team
After the training, nothing is measured
The course happened. How each person opens, handles an objection, or what they leave out day to day — of that there is an impression at best, and no number.
4 · A new product or new terms
After the briefing, nobody checks
The feedback arrives when the colleague is already stuck in front of a customer. Until then, nothing measures where the material landed and where it did not.
What we base this on
~15%
The share of financial advisors still with the company that hired them four years on. Skills training and mentoring are among the main retention factors.
LIMRA research
50–70%
The share of new insurance agents who drop out within the first twelve months.
industry analyses
~$165
What a single hour of corporate training costs on average. Role-play is the most expensive part of it, because one trainer can only work with one or two people at a time.
ATD 2024
The call
How a practice call goes
Your colleague hits the call button. It rings, and the other side picks up: “Hello, Bálint Takács speaking.” From there it is a conversation, with no keyboard. Bálint is 44, a regional director, paying off a mortgage — and he was not expecting the call. This is a cold call, the hardest situation to practice; a referral-based call can be set up as its own scenario.
01
It rings
A few seconds to get ready, exactly like a real call.
02
The conversation
He asks back, objects, drifts off topic. He does not help, and he is not enthusiastic.
03
He decides how it ends
If his patience runs out, he hangs up. If you convince him, you have an agreement.
04
The review
Within a minute: scores, quotes, and one concrete piece of advice.
05
The trend
Every call feeds into the progress curve.
An excerpt from a practice call
Bálint
Hello, Bálint Takács speaking.
Advisor
Good afternoon, my name is Péter Kovács from…
Bálint
Sorry — where did you get my number?
Advisor
I got your number from a business database…
Bálint
I see. Look, I have two minutes, then I am in a meeting. Tell me quickly what this is about.
…from here on, how long the conversation lasts is up to the advisor.
No specifics, and he hangs up
If all he gets is platitudes, or the caller is not listening to him, he loses patience and ends the call. There is no way to dodge the hang-up.
Available any time
He picks up at eleven at night. He does not take offense at the twentieth attempt, and he does not tell anyone how badly it went.
No two calls are the same
The same mood plays out differently across two calls — this is not a recording being replayed. Which is why running it again is worth something.
What your colleague gets after the call
Quotes from their own sentences
“No number followed the sentence ‘this really is a good offer’. That is exactly what Bálint asked about twice, and both times he got a generality back.”
Why he hung up
Every call shows the real reason for the hang-up: no specifics, patience ran out, the conversation drifted — or they actually reached an agreement.
One concrete step
A one-sentence tip on the weakest dimension for the next call. Not course material — one thing to do differently. The full transcript is there to read back as well.
Manager view
Team and individual views
Team overview
Who practiced how much, what the average scores are, which way the trend is going — broken down by dimension.
Allowances, per person
You allocate the practice allowance. A new joiner gets more of it during their onboarding month.
Assignments
“Two calls with the difficult customer type this week, everyone.” You set it, and you see who did it and how it went.
A threshold for working solo
A score threshold before someone goes to a customer on their own. The same works for periodic re-checks.
Weekly summary
An automatic email with the week’s results — and a flag if someone has not practiced for two weeks or their trend is sliding.
Executive roll-up
The same picture aggregated at office, region, or whole-organization level; where requirements differ, you can set a separate bar.
Three trainers, one technology
Pick the one your team needs.
All three run on the same engine — the script and the scoring criteria are what differ.
Sales trainer
The rep calls an AI customer who asks back, objects, and hangs up if the conversation is going nowhere. Afterward they are scored across five dimensions, with quotes from their own sentences.
cold calling · objection handling · closing
Customer service trainer
The same system with a different script: your colleague practices with an impatient, complaining, or hard-to-follow customer. We build in your own complaint types and the tone you expect.
complaint handling · defusing anger · onboarding
Language trainer
Speaking practice in a foreign language: the system is the conversation partner, tracks progress, and builds exercises around what is not working yet. Its focus is spoken fluency.
speaking · level progress · daily practice
The sales trainer is in use in the financial sector — among others we built it for OVB Hungary, with 1337 Partners.
Your content
The demo is a generic sample. The live system is built from your material.
You do not get an off-the-shelf product, you get one that measures your practice. Four things are tailored to you:
Who it should call
Your typical customers: the young couple taking out a first mortgage, the business owner putting money aside for retirement, the client whose savings plan is maturing. Gender, age, life situation, and voice are all configurable.
Which objections it should raise
The ones you hear most: “I already have an advisor.” “I do not believe in these.” “Send me an email.” From your real list, not a template.
What it should know about the product
Your products, your terms, your arguments. The AI customer asks about exactly the details people ask about in your market.
What it should score against
This is the important one: it evaluates by your methodology. If a call at your company follows a defined shape, it measures the conversation against that — not against a general textbook.
Four steps, roughly a month
Week 1
Handing over the material
One call, and you hand over what you already have.
Your side: about 2 hoursWeeks 2–3
Your version
We build the customer characters, the situations, and the scoring criteria to your methodology.
Your side: 1–2 calls with whoever owns trainingWeek 4
Launch
Accounts, teams, allowances. People log in with an email and a code, and start practicing.
Your side: a name list and a short briefing for the teamWeeks 4–8
Trial and measurement
Practice runs, numbers accumulate. At the end we review together against the metrics agreed up front.
Your side: practice plus one closing callNo IT integration
A standalone practice interface, not connected to your internal systems. What has to be set up: accounts, teams, allowances, and a name list.
Roll it out in stages
You do not have to start with the whole organization. One office, a few weeks — then there is something to decide on.
Extend it later
A new product, a new objection, a new situation can be added at any time, without rebuilding anything.
Legal and data handling
What gets created, where it is stored, and who sees it
We settle this up front — without it, people will not practice honestly either. Most of it is your call: what follows is the default.
What is created
A text transcript of the conversation, the scores, and a record of who practiced how much and when. Audio recordings only if you ask for them. All of it in an EU data center (Frankfurt).
What is not created
No real customer data is involved: practice runs against invented customers. We get no access to your internal systems, and we do not observe live calls.
Who sees what
The colleague sees their own calls, their direct manager sees the team’s. Above that, only as much as you allow: aggregate numbers by default, not individual transcripts.
Not an HR instrument
The score is for development, and we put that in the contract: it cannot be grounds for dismissal or any sanction. That keeps the system outside the stricter EU rules on AI evaluation of employees.
What it is not good for — We would rather say this up front than have it surface in month three.
A machine customer is not reality
The AI customer is more predictable than a person. Practice prepares you, but the first real conversation always brings something new.
It does not replace the mentor
It takes the hundredth role-play off the mentor’s shoulders, not the professional guidance. What it does add to a coaching conversation is something that was not there before: a written transcript and numbers.
It does not measure live calls
It measures practice, not real customer conversations. Someone who does well here has better odds in the field — but this is readiness measurement, not quality assurance.
It is not an HR tool
It was built for development: the score exists so the colleague knows what to work on. Worth saying out loud at rollout too — otherwise nobody practices honestly.
Common questions
What people ask before they start
Should I make it mandatory for the team?
Only for new joiners — there the threshold score is part of getting ready. With the existing team it works when it arrives as a tool, not as surveillance. The best pattern: the manager runs a few calls themselves and shares the result.
Does the score affect commission or appraisals?
No — and we put that in the contract. The score is for development: it does not enter performance reviews and it is not the basis for any decision. That single sentence decides whether the team practices honestly.
Which languages does it work in?
It was built for Hungarian, with its own pronunciation rules and its own scoring engine. The language trainer can also be used for speaking practice in other languages.
How long until it runs on our own material?
Roughly a month for one team: handover, building your version, launch — then four weeks of trial against metrics agreed in advance.
Try it
Call the AI customer — right now.
It takes two minutes: it runs in the browser, with nothing to install. Pick a mood, call the customer — and after the call you see the scored review.
The system is already running
The call, the review, and the progress tracking are live — you can try it on the demo any time. What is missing is not the technology, it is your content.
Built for Hungarian
Comparable international systems were built for English. This one was built for Hungarian speech and Hungarian conversational dynamics — with its own pronunciation rules and its own scoring engine.
Tested on live calls
Behind the demo are weeks of testing with real salespeople. That is where the details came from that keep it from sounding like a robot: when it is patient, when it interrupts, when it hangs up.