In a 2021 Rise of AI Virtual Chat held with AWS, Fabian Westerheide interviews Michael Brehm, CEO of i2x, about AI-driven customer experience and the wider European deep tech ecosystem. Brehm describes i2x's real-time voice analytics product, built on the company's own speech-to-text engine, which gives call center agents live coaching during customer calls in under half a second, and argues full automation of human-to-human phone conversations remains more than a decade away. He argues Europe and Germany produce strong research and patents but fail to convert them into scaled applications and companies, contrasting a roughly 150 billion euro valuation for the whole European deep tech startup sector with TikTok's similar valuation as a single AI-driven product. He covers i2x's consent-based approach to handling call data, including anonymization and access controls, and closes with broader reflections on AI's effect on productivity, employment and wealth concentration, and on the fragmentation of the internet into US, Chinese, Russian and other regulatory blocs, plus practical details on i2x's use of AWS cloud infrastructure.
Key takeaways
- i2x built its own speech-to-text engine to give call center agents real-time coaching during calls in under half a second.
- Brehm estimates full automation of complex human-to-human phone conversations is still more than ten years away.
- He values the entire European deep tech startup sector at about 150 billion euros, comparable to TikTok's valuation as a single AI-driven product.
- i2x only processes call data with consent from both agents and customers and anonymizes names, addresses and card numbers.
- Brehm expects the internet to keep fragmenting into separate US, Chinese, Russian and other regulatory ecosystems rather than converging.
Detailed notes
i2x: real-time voice coaching for call center agents
Brehm explains that i2x grew out of his experience running a large call center with 2,500 telesales agents at a previous company, and describes the goal of feeding agents real-time coaching, defined as a full feedback loop in under half a second, during live calls. i2x built its own speech-to-text engine to control data quality, speed and integration, then layered real-time tips on top, such as prompts to speak faster or avoid certain phrasing. He frames the product as an augmentation layer for human-to-human conversation rather than a step toward full automation, arguing that automating a call with ten to fifteen questions would require near-perfect accuracy at every step, which current technology cannot reliably deliver, so he estimates full automation of such interactions is more than ten years away.
Can Germany and Europe do deep tech?
Brehm says Europe is strong in R&D, producing comparable numbers of patents, including in quantum technology, to the US and China, but weak at turning that research into scaled companies. He cites a recent study valuing the entire European deep tech startup sector at about 150 billion euros, comparable to TikTok's own valuation as a single AI-driven product. He points to poor university licensing practices in Europe, where universities may demand royalty fees that can stall funding rounds, as one concrete obstacle, and argues the core problem is a weak culture of building use cases and businesses around research rather than a lack of technology itself.
Handling sensitive call data at i2x
Asked about privacy, Brehm says i2x only processes call data with the consent of both agents and customers, and recommends giving consent because it improves service quality. He describes layered access controls limiting how many staff can view raw data, contractual restrictions on data use, anonymization tools that strip names, addresses and credit card numbers, and automatic labeling of most data, with human review limited to short clips of ten to thirty seconds.
AI, productivity and wealth concentration
Responding to a question about AI reducing total work hours needed for growth, Brehm argues the fear of technology destroying the total amount of available work has recurred since early industrial weaving machinery and has historically proven wrong, noting Germany combines a high density of robots per capita with one of the world's lowest unemployment rates. He expects rising demand for new kinds of work even as routine tasks automate, but raises concern about wealth concentration, citing Thomas Piketty's argument that returns to capital outpace wages, and says he personally favors Germany's social market economy and redistribution through taxation over what he sees as a starker US model.
A fragmenting internet: US, China, Europe and beyond
Discussing global technology competition, Brehm expects the internet to keep fragmenting into separate regulatory ecosystems, citing China's firewall, Russia building its own internet infrastructure, and India blocking Chinese apps, while Western markets remain comparatively open to each other despite competition and platform monopolies. He says he hopes for closer collaboration between the US and Europe, referencing a speech by President Biden calling for stronger Western cooperation, and credits decades of European integration with an unusually long period of peace and prosperity.
Working with AWS
In closing, Brehm contrasts i2x's use of AWS cloud infrastructure for storage, compute and model iteration with his earlier experience at a company that needed staff to physically install servers in racks. He says i2x does not offer on-premise deployments, instead tapping into clients' telephony systems in real time through software, whether the client's own system is cloud-based or on-premise infrastructure.
Quotes
Technology is only as good as the use case that's behind it.
Michael Brehm
Our vision is to build a software layer on top of the human brain to help them to interact better with other humans.
Michael Brehm
It's not the end user's fault, it's our fault if we don't build this better UI, better applications.
Michael Brehm
Questions answered in this talk
What does i2x's software do for call center agents?
According to Michael Brehm at this Rise of AI Virtual Chat, i2x analyzes sales calls in real time using its own speech-to-text engine and gives agents live coaching, such as suggested phrasing or pacing, with a feedback loop under half a second.
Why has Europe struggled to build large deep tech companies despite strong research?
According to Michael Brehm, Europe produces patent output comparable to the US and China, including in quantum technology, but lacks a culture of turning research into applications and scaled startups, valuing its entire deep tech startup sector at about 150 billion euros, similar to TikTok alone.
How does i2x protect customer privacy when analyzing call recordings?
According to Michael Brehm, i2x only processes calls with consent from both agents and customers, restricts data access through layered permissions and contracts, anonymizes personal details such as names and card numbers, and limits human review to short clips.