TL;DR and context
- Adviser training has long been counted in hours and test scores. AI simulations add a score for each rehearsed call, though not yet for the advice a real client receives.
- TODAY’s AI clients have personalities and objections and can end a call, so advisers can rehearse difficult conversations without a senior colleague playing the client.
- Königswege’s published case study reports 80 mentor hours saved a month and a 23.5% rise in simulation scores. These are results reported by the company.
- The case doesn’t establish whether advisers give better advice to real clients. Managers still define the scorecard and decide when trainees are ready.
In 2017, while I was still at college, I took a part time job on a fintech support desk, and training there meant mock calls with a trainer playing the customer. They helped, but the trainer always knew where the call was going. Real callers didn’t, and a few of them had a poor experience while I found my feet.
Two years later at Reuters, I sat through 14 training modules, some of them three hours long, each ending in a multiple choice test, or multiple guess as we called it. My whole MSc in France ran on much the same format. A support desk, a newsroom and a business school are very different places, yet holding my attention was hard in every one of them.
I won’t pretend modules like these are pointless. A data governance course can get the rules across and give an employer a record that staff received them. But regulators have noticed how little that record proves. Britain’s Financial Conduct Authority has told advisers that continuing professional development “is not a collection of attendance certificates”.
Advising a client is a different kind of skill altogether. It means listening, choosing your words and adapting when the other person reacts in a way you didn’t expect. A multiple choice test can show whether you recognise a good answer, not whether you can actually say it when someone is losing patience on the other end of the line. Role play gets closer, but only if someone experienced plays the other person, and for a financial firm hiring hundreds of advisers, that ties training to the people whose time is worth the most.
That is the gap Michael Gackstatter is going after. He is cofounder and chief executive of TODAY, a Copenhagen and Berlin startup that sells software to financial advisers, and he has built its AI training around a client who can take offence and hang up. Each simulated client comes with a personality, objections and limited patience, and a trainee who pushes too far hears the line go dead.
“Our ‘AI customers’ do get offended, raise objections or hang up,” Gackstatter told Industry Contents, “so you can also practise emotionally stressful situations.” His answers were translated from German.
For firms that train new advisers through role play with senior colleagues, the appeal is obvious. The colleague goes back to work while the trainee practises with software, repeating the same awkward conversation as often as needed without asking a manager for another hour. TODAY’s published customer case study reports 80 mentor hours saved a month and a 23.5% rise in simulation scores. Those are exactly the units financial firms have always used to count training, hours sat and scores earned, and that is worth keeping in mind as the story goes on.
A module test records whether an adviser remembers the material. A senior colleague can judge how an adviser handles a difficult conversation, but that takes time away from clients. TODAY gives firms a score for each simulated call against their own criteria. That helps managers track practice, while the effect on advice given to real clients remains unmeasured.
When the trainer has clients of their own
Gackstatter founded TODAY in 2024 with chief technology officer Artem Demchenkov. Both had held senior roles at fintech and insurtech companies, including CLARK, a German digital insurance broker, and Funding Circle, an online lender to small businesses. The company raised €2.8 million in a seed round announced on 8 October 2026, co-led by High-Tech Gründerfonds and Insurtech Gateway, and says more than 3,500 advisers across Germany, Austria and Switzerland now use its platform.
One of its customers is Königswege, a German financial sales organisation that describes itself as one of the fastest growing businesses in its sector, with €40 million in annual revenue and 1,800 sales partners. Its mentors were also its sales managers. They rehearsed prospecting, objection handling and advice conversations one to one with new partners, sometimes at weekends, and every one of those sessions took a manager away from their own clients.
As the organisation grew, that arrangement got harder to manage. Managers couldn’t easily see how often trainees practised, and each mentor brought their own expectations. TODAY’s simulations let trainees practise on their own while managers follow their progress through a dashboard.
“With the top-10 sales organisation Königswege, we got more than 250 advisers fit with simulations, 10 conversations each that would otherwise be held with the manager,” Gackstatter told Industry Contents.
The published case study puts the figure slightly differently, at 235 partners completing 1,690 simulations, or about seven each. For a sales manager who has spent weekends playing an awkward prospect, that is a fairly tangible return, and a growing firm can offer more practice without finding more managers to play the customer. Whether Königswege turned the recovered time into extra revenue or faster onboarding hasn’t been established.
The practice itself is built to feel personal. Each simulated client has a personality, emotions, typical expressions and a backstory, so a trainee might face a sceptical existing customer one day and an acquaintance weighing up financial advice the next.
“Particularly difficult interpersonal situations, such as when you advise old friends or acquaintances, are easy to simulate with TODAY Simulations,” Gackstatter told Industry Contents, “because we can give the AI an arbitrarily complex ‘backstory’.”
Once the call ends, the trainee gets feedback against a scorecard the employer has defined, with measures such as share of talking time and repetition. Marvin Gräber, a director at Königswege, says in the case study that trainees often see AI feedback as more objective and accept it more readily than criticism from a person.
A shared scorecard gives managers a consistent way to review performance. But it is also a decision about what a good conversation looks like, and when an employer defines success, trainees learn to chase whatever it measures.
In financial advice that matters, because a confident conversation or a well handled objection doesn’t necessarily mean the customer got the right recommendation. Königswege’s reported improvement measures performance against its own simulation criteria. The case study stops short of the client on a real call.
Medicine offers a useful comparison. A 2024 meta-analysis in the Journal of Medical Internet Research examined 27 randomised trials involving 1,480 participants in medical and nursing education. It found no statistically significant difference between virtual simulation and training with mannequins or real people on knowledge, procedural skills, clinical reasoning or communication. Put plainly, the software did about as well as a person playing the part.
From training to the working day
The simulations are only one corner of what TODAY sells. The company describes itself as an AI sales operating system for financial and insurance advisers, with agents that work before, during and after a client meeting, transcribing it, handling back office tasks, coaching the adviser and answering incoming calls. A BeBeez report on its 2025 pre-seed round adds that the software writes the call summary, updates the client record and shows talk ratios and customer reactions in a coaching view.
The pitch to advisers is time. “Our agents take over the admin, so advisors can spend their time with clients,” Gackstatter said in the seed announcement. The company reports that advisers save an average of five hours a week on administration and earn up to 10% more revenue. The same measures, such as share of talking time, show up both in simulation feedback and in coaching views of real meetings, so the scorecard follows the adviser out of practice and into the working day.
Regulation, meanwhile, still counts in hours. The IHK, Germany’s chambers of commerce and industry, says that under the country’s implementation of the EU Insurance Distribution Directive, insurance distributors and relevant employees generally must complete 15 hours of continuing professional training a year. Firms remain responsible for deciding whether their training meets the requirements, keeping records and judging when advisers are ready to work with clients.
That is where TODAY could give managers a different job. Instead of playing the customer in every rehearsal, they can look at where a trainee struggled and decide what needs a human coach. The company has evidence of time saved, though the published case study doesn’t measure how that supervisory work changes.
Pilots log hours in simulators so the first emergency isn’t the real one. A simulated customer can hang up ten times without costing a firm a real client, but someone still has to work out what the adviser learned by the eleventh call. I would start by asking the client.
Three questions to ask yourself
- Does our training record hours, scores or what actually happened with clients?
- Who decided what a good client conversation looks like, and when was that last checked against complaints or cancellations?
- If software takes over the rehearsals, what will managers do with the hours they get back?
FAQ
What is TODAY and who is it for?
TODAY is a Copenhagen and Berlin based start-up that sells an AI sales operating system to financial and insurance advisers. It covers meeting transcription, back-office tasks, sales coaching and simulated client conversations for training. The company says more than 3,500 advisers across Germany, Austria and Switzerland use it.
How do TODAY’s AI client simulations work?
Each simulated client has a personality, emotions and typical expressions, and coverage of the seed round says the AI customers are configurable across DISC personality profiles. The trainee speaks to the client, answers objections and can be hung up on. Feedback follows against a scorecard the customer defines, with measures such as talking-time share and repetition.
What results has TODAY reported?
TODAY’s case study with Königswege reports 80 mentor hours saved a month, a 23.5% rise in simulation scores and 235 partners using the tool. These are company-reported figures.
Does AI role-play help advisers prepare for real clients?
That is the aim. The simulations let advisers rehearse objections and difficult moments before a real meeting. A 2024 medical-education meta-analysis found no significant difference between virtual and human-led simulation on communication skills, though financial sales is a newer setting for the technique.
Can simulation training fit German adviser training hours?
The IHK says insurance distributors and relevant employees must generally complete 15 hours of continuing professional training a year. Firms covered by the rule can look at how simulated practice fits into those hours.
Who benefits most from AI adviser training?
Mentors and training managers get hours back. Trainees get repeatable practice and one scorecard. Firms get a practice dashboard, and clients meet advisers who have already rehearsed difficult conversations.
Read next
Compare the costs and tools Use IC Lab to examine pricing models and the practical trade-offs behind AI adoption.