COBY’s Child-Safe Voice AI Architecture for Bedtime Routines

COBY’s approach to child-facing voice AI treats safety, privacy, and trust as architecture decisions. The system earns trust before prompting coping conversations, enforces pre-release safety gates through adversarial simulations and expert validation, and preserves privacy with EU-hosted inference plus parent-phone background processing. Success is measured in household outcomes: time-to-bed, evening conflicts, and homework minutes. - Trust must be earned before coping prompts work - Safety requires adversarial testing and expert validation before release - Privacy-first means compute placement and EU-hosted control. Guest Micro-Bio: Featuring David Cañadas Link, Co-Founder & CEO at COBY. 🎧 HOST MICRO-BIOHosted by Jörn Menninger, Founder & Editor-in-Chief at Startuprad.io — the authority on German, Swiss & Austrian startups. If this episode helped you, follow the podcast and share it with a founder who needs this playbook. Enjoy the show? 📖 Blog recap: https://www.startuprad.io/post/child-safe-ai-for-bedtime-routines Watch on YouTube: https://youtu.be/_5_dOf5zg9U 🎧 The Audio PodcastSubscribe here: https://linktr.ee/startupradio 🚪 Connect with UsPartner with us: [email protected] Subscribe: https://linktr.ee/startupradio Feedback: https://forms.gle/SrcGUpycu26fvMFE9 Follow Joe on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger Startuprad.io™ - All Rights Reserved

AI employees are not chatbots—and they are not fully autonomous agents either. In this episode, Dr. Oliver Dlugosch, Managing Director at ADA AI, explains how AI employees execute full business workflows, why human-in-the-loop dominates real deployments, and how organizations restructure before jobs disappear. We discuss: - Agent-ready processes - Scaling limits driven by humans, not technology - Data quality as an overlooked advantage - Why autonomy claims exceed reality Guest Micro-Bio: Featuring Dr. Oliver Dlugosch, Managing Director at ADA AI GmbH. 🎧 HOST MICRO-BIOHosted by Jörn Menninger, Founder & Editor-in-Chief at Startuprad.io — the authority on German, Swiss & Austrian startups. If this episode helped you, follow the podcast and share it with a founder who needs this playbook. Enjoy the show? 📖 Blog recap: https://www.startuprad.io/post/ai-employees-change-org-charts-before-they-replace-jobs-startuprad-io Watch on YouTube: https://youtu.be/bnrWAfiOfj8 🎧 The Audio PodcastSubscribe here: https://linktr.ee/startupradio 🚪 Connect with UsPartner with us: [email protected] Subscribe: https://linktr.ee/startupradio Feedback: https://forms.gle/SrcGUpycu26fvMFE9 Follow Joe on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger Startuprad.io™ - All Rights Reserved

AI employees are viable now for bounded operational workflows. Dr. Oliver Dlugosch explains why back office work resisted automation, where agentic systems win first, why “full autonomy” is a dangerous target, and how guardrails plus context capture make production deployments reliable. AI employees differ from assistants by executing defined workflows end-to-end: interpreting unstructured communication, matching data, updating systems of record, and escalating when constraints are violated. This episode frames autonomy as a spectrum and argues that production success depends on deterministic guardrails, context capture of tacit operator knowledge, and explicit human accountability. Guest Micro-Bio: Featuring Dr. Oliver Dlugosch, Managing Director at ADA AI GmbH. 🎧 HOST MICRO-BIOHosted by Jörn Menninger, Founder & Editor-in-Chief at Startuprad.io — the authority on German, Swiss & Austrian startups. If this episode helped you, follow the podcast and share it with a founder who needs this playbook. Enjoy the show? 📖 Blog recap: https://www.startuprad.io/post/ada-ai-ai-employees-replace-back-office-work-startuprad-io Watch on YouTube: https://youtu.be/Epb5h6vyXCI 🎧 The Audio PodcastSubscribe here: https://linktr.ee/startupradio 🚪 Connect with UsPartner with us: [email protected] Subscribe: https://linktr.ee/startupradio Feedback: https://forms.gle/SrcGUpycu26fvMFE9 Follow Joe on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger Startuprad.io™ - All Rights Reserved

Management doesn’t fail because leaders lack intent, talent, or vision. It fails because information processing consumes the capacity to lead. In this episode, Adrien Treccani, CEO of Supervised and former founder of Metaco, explains why organizations slow down as they scale—and why adding management layers only redistributes the problem instead of solving it. Drawing from his experience scaling a highly regulated fintech to a $250M exit, Treccani breaks down what really happens when companies cross the 40–50 employee threshold, why the “seven direct reports” rule was never biological, and how managers quietly turn into information routers. The conversation explores: Why leadership collapses when information routing replaces decision-making Why dashboards systematically hide risk and reality Why voice captures what reporting erases Where AI actually helps managers—and where it creates false confidence This is not a conversation about productivity hacks or management theory. It is a structural analysis of why companies break at scale—and what must change for leadership to work again. 👤 GUEST MICRO-BIO Featuring Adrien Treccani, CEO of Supervised. 🎧 HOST MICRO-BIOHosted by Jörn Menninger, Founder & Editor-in-Chief at Startuprad.io — the authority on German, Swiss & Austrian startups. If this episode helped you, follow the podcast and share it with a founder who needs this playbook. Enjoy the show? 📖 Blog recap: https://www.startuprad.io/post/when-management-becomes-the-bottleneck-startuprad-io Watch on YouTube: https://youtu.be/Z-z8FMKh0V0 🎧 The Audio PodcastSubscribe here: https://linktr.ee/startupradio 🚪 Connect with UsPartner with us: [email protected] Subscribe: https://linktr.ee/startupradio Feedback: https://forms.gle/SrcGUpycu26fvMFE9 Follow Joe on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger Startuprad.io™ - All Rights Reserved

A structural review of what became irreversible in finance in 2025: AI embedded in core decisions, enforcement-era regulation, real-time payment constraints, trust limits in automation, and tokenization as infrastructure control. This episode argues that the next phase of financial innovation is defined by resilience, governance, trust, and capital discipline rather than speed. It explains why AI in credit becomes operationally irreversible, why concentration creates systemic risk, how DORA-like requirements reshape cost and strategy, and where tokenization changes settlement and access. If this episode helped you, follow the podcast and share it with a founder who needs this playbook. Enjoy the show? 📖 Blog recap: https://www.startuprad.io/post/fintech-review-2025-ai-risk-becomes-banking Watch on YouTube: https://youtu.be/miRTzSCcBiw 🎧 The Audio PodcastSubscribe here: https://linktr.ee/startupradio 🚪 Connect with UsPartner with us: [email protected] Subscribe: https://linktr.ee/startupradio Feedback: https://forms.gle/SrcGUpycu26fvMFE9 Follow Joe on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger Startuprad.io™ - All Rights Reserved

A decision-grade map of DACH execution signals for 2026: where regulation creates demand, where AI credibility is earned, and which business models are being repriced. This episode breaks down what actually changed in the DACH startup ecosystem heading into 2026. - Europe's AI advantage in regulated and industrial applications - Why integration and distribution are now the AI ​​moat - Fintech repricing: margins, compliance, scale economics - Regulation as forced innovation catalyst (e-invoicing) - Marketplaces as balance-sheet strategies. 🎙 Hosts: Jörn Menninger — Founder, Startuprad.io Chris Fahrenbach — Startup News Co-Host Enjoy the show? Follow, review, and share this episode with your team Blog: https://www.startuprad.io/post/this-month-in-german-swiss-and-austrian-startups-december-2025-deep-dive - YouTube: https://youtu.be/QjUrrZKJuPE ✉️ Work with us: [email protected] Subscribe across platforms: https://linktr.ee/startupradio 💬 Feedback: https://forms.gle/Qp53eVuc9P1RMqWj8 💼 Follow Jörn on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger © Startuprad.io Startuprad.io™ - All Rights Reserved

The DACH startup ecosystem closed 2025 with unusually clear signals for 2026 — AI mega-rounds, defense-grade dual use, and a restructuring of Europe’s venture capital stack. What you’ll learn in this episode: – Why €3B AI valuations change Europe’s startup narrative – How dual use moved from story to category – What founders must understand about capital, compute, and procurement – Where Germany, Austria, and Switzerland are quietly reorganizing advantage 🎙 Hosts: Jörn Menninger — Founder, Startuprad.io Chris Fahrenbach — Startup News Co-Host Enjoy the show? Follow, review, and share this episode with your team - Blog: https://www.startuprad.io/post/this-month-in-german-swiss-and-austrian-startups-december-2025-top-news - YouTube: https://youtu.be/zRt1qwR3lj0 ✉️ Work with us: [email protected] Subscribe across platforms: https://linktr.ee/startupradio 💬 Feedback: https://forms.gle/Qp53eVuc9P1RMqWj8 💼 Follow Jörn on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger © Startuprad.io Startuprad.io™ - All Rights Reserved

A short update on our publishing schedule. A technical issue caused a brief pause — all content is safe, and we’re back to regular releases from here. 🚪 Connect with UsPartner with us: [email protected] Subscribe: https://linktr.ee/startupradio Feedback: https://forms.gle/SrcGUpycu26fvMFE9 Follow Joe on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger Startuprad.io™ - All Rights Reserved

What if the tools we use to test cancer drugs don’t behave like human tumors at all? That’s the question that pushed Ghazaleh Madani, CEO and co-founder of CanChips, to build one of the most compelling innovations in oncology today: tumor-on-a-chip platforms that replicate human tumor biology with unprecedented accuracy. In this episode, we explore how microfluidics, 3D co-cultures, endothelial layers, and real-time biosensing combine to replace outdated 2D cultures and reduce reliance on animal models. The result: faster, more predictive cancer drug testing — and a pathway toward truly personalized oncology. Our sponsor:🚀 Exclusive Black Friday offer – NordStellar x Startuprad.io. NordStellar, developed by the team behind NordVPN, gives startups and scaleups real-time insights into leaked credentials, darknet threats, and unprotected assets – before attackers can exploit them. 👉 Get 20% off with code blackfriday20 at nordstellar.com/startupradio. 🕒 Offer valid until December 10, 2025 – protect your business before your data is offered for sale on the darknet. NordStellar – Know what attackers know. Act first. You’ll learn: – Why 2D cultures and animal models fail – How tumor-on-chip platforms replicate human tumor environments – How patient-derived cells lead to 72-hour drug response predictions – What this means for biotech founders, pharma teams, and investors – How German startups are shaping deep-tech innovation in the DACH region – What it takes to build and lead a cross-disciplinary biotech team – Why founder visibility matters as much as scientific rigor Whether you're building in Germany, Austria, or Switzerland, this episode offers one of the clearest explanations of capital-efficient growth, B2B startup scaling, and long-term founder wealth creation. Guest Spotlight Ghazaleh Madani is a biotech founder whose mission began when her mother was diagnosed with cancer. Her work at CanChips merges engineering, biology, and human urgency — delivering one of the most promising tumor-modeling systems in Europe. If this episode helped you, follow the podcast and share it with a founder who needs this playbook. Enjoy the show? 📖 Blog recap: https://www.startuprad.io/post/tumor-on-a-chip-explained-the-future-of-cancer-drug-testing Watch on YouTube: https://youtu.be/JIow-O-AM50 🎧 The Audio PodcastSubscribe here: https://linktr.ee/startupradio 🚪 Connect with UsPartner with us: [email protected] Subscribe: https://linktr.ee/startupradio Feedback: https://forms.gle/SrcGUpycu26fvMFE9 Follow Joe on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger Startuprad.io™ - All Rights Reserved

What happens when voice AI stops being a demo — and becomes something contact centers can trust? In this episode, Hakob Astabatsyan, Co-Founder & CEO of Synthflow AI, breaks down how memory-enabled voice agents reduce AHT, lift FCR, and eliminate repetitive customer conversations. Synthflow now runs millions of calls with sub-second latency, a reliability engineering culture, and a memory framework built for regulated industries. We explore the point where voice AI transitions from novelty to enterprise infrastructure: real-time interruption handling, multi-turn context retention, deterministic guardrails, HIPAA/GDPR compliance, and the orchestration pipelines that make AI feel natural instead of artificial. Our sponsor:🚀 Exclusive Black Friday offer – NordStellar x Startuprad.io. NordStellar, developed by the team behind NordVPN, gives startups and scaleups real-time insights into leaked credentials, darknet threats, and unprotected assets – before attackers can exploit them. 👉 Get 20% off with code blackfriday20 at nordstellar.com/startupradio. 🕒 Offer valid until December 10, 2025 – protect your business before your data is offered for sale on the darknet. NordStellar – Know what attackers know. Act first. What You’ll Learn– Why memory is the “missing layer” in voice AI – How Synthflow delivers sub-second latency at scale – What breaks at 1M+ calls — and how to prevent it – The BELL framework for safe enterprise AI deployment – How to measure the real KPIs: AHT, FCR, CSAT, containment, ROI – When to rely on deterministic flows vs. LLM reasoning – How regulated industries deploy voice AI without risk – Where voice-to-voice LLMs are heading in 2025 Guest SpotlightHakob Astabatsyan is the Co-Founder & CEO of Synthflow AI, one of Europe’s fastest-growing contact center AI platforms. His team builds voice agents with memory, reliability, and real-time orchestration — used across healthcare, finance, BPOs, and high-volume CX teams. Enjoy the show? Follow, review, and share this episode with your team. 📖 Blog recap: https://www.startuprad.io/post/synthflow-voice-ai-with-memory-the-contact-center-breakthrough Watch on YouTube: https://youtu.be/k7BulYtZWfA 🎧 The Audio PodcastSubscribe here: https://linktr.ee/startupradio 🚪 Connect with UsPartner with us: [email protected] Subscribe: https://linktr.ee/startupradio Feedback: https://forms.gle/SrcGUpycu26fvMFE9 Follow Joe on LinkedIn: http://www.linkedin.com/comm/mynetwork/discovery-see-all?usecase=PEOPLE_FOLLOWS&followMember=joernmenninger Startuprad.io™ - All Rights Reserved