Amir Elion

#AI is a challenge to on our road to #sustainability. At the same time, it will be part of the solution. Our very own Dr. Helge Maas just co-authored 'A Roadmap to Sustainable AI in the Public Sector'. It highlights how AI – when applied responsibly – unlocks major gains for governments and citizens alike. Here are some of the examples we share in the report: ➡️ AI, including Gen AI, can boost productivity by 15.5%, helping civil servants deliver more with less – reducing delays, paperwork, and energy use. ➡️ Our smart energy command centers powered by AI and IoT can cut building energy consumption by 30%. ➡️ AI-driven analytics can help governments anticipate demand, allocate resources more effectively, and reduce waste across sectors. The key? Developing and implementing a sustainable AI roadmap. Explore how to lead with purpose and performance: https://lnkd.in/eDjHUdny Download the full report here 👉 https://bit.ly/4fBoRWu Tom van den Nieuwenhuijzen | Felix Caspari | Marc Reinhardt Frank Jacobsen | Oliver Stuke #PublicSector #Sustainability

34

AI is only as good as the data you feed it. But what truly makes it powerful is the context we keep enriching it with and the purpose we choose to build it on. That’s why I found this perspective on Systemic AI so powerful. It moves beyond experiments and pilots to a framework that helps enterprises truly lead with AI and drive impact at scale. MathCo leaders have this unique framework of Systemic AI and as a Board Member of the company I have seen the traction it is getting with global enterprises and it’s being translated into real value for these businesses. Give this a read - “Systemic AI” : https://lnkd.in/e7fsYJAG #SystemicAI #OwnYourIntelligence #EnterpriseAI #AgenticAI Sayandeb Banerjee Aditya Kumbakonam Anuj Krishna Chandni Bose

156

1 kommentar

Over the past few years, we’ve been helping some of the world’s largest public companies experiment with AI. But one pattern kept repeating: no one had a clear vision for what a truly AI-enabled organization looks like. So we built one. In our new report, The Self-Driving Organization, we lay out five levels of AI maturity, from tool-assisted teams to fully adaptive, learning enterprises. It’s not a tech roadmap, it’s an organizational one. The goal isn’t to automate everything; it’s to create systems that learn, coordinate, and act with integrity. If your company is serious about AI, set your compass: “Let’s become a Level 4 self-driving organization by 2027.”

127

18 kommentarer

Scaling AI in healthcare? Don’t skip the stress test. If you’re building an AI assistant or agent to support patients, health coaches, or clinicians - testing it with a few example user messages isn’t enough. At Sacher AI, we’ve worked on both sides of the challenge: 🟠 Building AI assistants that augment health coaches 🟠 And building the tools to monitor and evaluate those assistants, both before and during real world deployment with patients That includes evaluating both: 🟠 What the AI says (e.g. hallucinations, motivational tone, inappropriate empathy) 🟠 What the user says (e.g. confusion, emotional risk, escalation cues) We’ve learned the hard way: factual accuracy isn’t enough. Behavioural effects, like how language shapes trust, how tone influences motivation, or how a model’s “reassurance” might simulate empathy — matter just as much. That’s why we’re developing PromptSafe and a broader suite of behavioural and clinical safety tools, including: 🟠 Synthetic persona testing - simulated conversations scored by LLM evaluators, with actionable feedback to guide improvements (available now via PromptSafe) 🟠 Predictive classifiers and hybrid neuro-symbolic models 🟠 Deterministic logic to assess tone, safety, and fidelity to clinical and behaviour change frameworks like FAST Not all of this is available today, but we’re building toward it. PromptSafe is actively used during development to simulate real conversations, identify potential behavioural or clinical risks, and provide actionable feedback, helping teams improve their assistants and agents before they ever reach real users. Because when AI systems influence how people think, feel, and behave, performance benchmarks aren’t enough. They need to be tested in ways that reflect real world use and account for the behavioural complexity of human–AI interaction. If you’re developing an AI assistant or agent, how are you stress-testing it before scale? And what are you doing to monitor behavioural and clinical safety once it’s live?

22

1 kommentar