Building trust in cleantech and diagnostics

Key principles such as #patientprivacy, transparency, #accessibility, and human oversight are fundamental for building trust in AI-driven systems. For instance, ensuring compliance with data protection laws like HIPAA and GDPR is critical, as patient privacy remains a top concern, with 75% of consumers expressing worry about how their health data is used. Another overlooked aspect is the lack of diverse datasets. Bias prevention in AI algorithms is equally important; studies have shown that training data imbalances can exacerbate health disparities, disproportionately affecting underrepresented groups. While AI systems can reduce diagnostic errors by up to 30%, these gains depend on rigorous testing for clinical accuracy and ongoing algorithm validation with #realworld data. Continuous monitoring is essential for addressing risks as #healthtech evolves rapidly, requiring ethical deployment to prioritize patient outcomes over financial incentives (a tough pill to swallow for many). #ai should complement—not replace—human decision-making. The path forward demands interdisciplinary collaboration to navigate ethical challenges, creating #ai solutions that are both equitable and reliable for all populations.

🚨 The Trust Gap: Why AI Adoption in Medicine Remains Painfully Slow Despite the extraordinary capabilities of today's AI systems, healthcare adoption remains frustratingly limited. The evidence is rapidly accumulating—from Google's AMIE outperforming physicians in diagnostic accuracy to transformative AI-driven imaging systems demonstrating real-world clinical benefits in large-scale RCTs—yet implementation lags far behind potential. Why? In my opinion, it all comes down to trust. Trust isn't optional in medicine—it's fundamental. Healthcare professionals rightly demand extraordinary standards of reliability before changing practice. And while tech enthusiasts often dismiss this caution as resistance to change, they're missing a critical point: trust in healthcare must be earned, not assumed. Consider what builds trust in other contexts: 🔬 Real-world validation: I wrote before about how the Swedish MASAI mammography trial with 105,000+ participants showed AI could increase cancer detection by 29% while reducing radiologist workload by 44%. This level of rigorous, large-scale validation—not purely academic benchmarks involving multiple choice questions or company press releases—is what drives adoption. 📱 Experiential discovery: Physicians must directly experience AI's capabilities in low-stakes environments before trusting it with patient care. Abstractions and algorithm descriptions aren't enough—hands-on experience with realistic cases demonstrates both power and limitations. 🔄 Transparency over opacity: The trend of AI companies keeping model weights and code proprietary (as Google did with AMIE) fundamentally undermines trust. In medicine, "trust me" never suffices—"verify for yourself" must be the standard. Open models and open weights enable crucial independent validation. 📊 Calibrated communication: The hype cycle damages credibility. When AI companies make grandiose claims followed by disappointing reality, they erode the trust capital needed for meaningful adoption. Rigorous communication must replace marketing hyperbole. 🤝 Collaborative design: Systems developed with—not for—clinicians gain faster adoption. Physicians rightfully reject technologies that fail to account for workflow realities or that prioritize technical elegance over clinical utility. The path forward is clear: we need AI systems with rigorously demonstrated real-world benefits, developed transparently with clinician partnership, and communicated honestly without exaggeration. This isn't just ethical—it's practical. The fastest route to widespread AI adoption in healthcare isn't through marketing or mandates. It's through painstakingly earned trust.

when everything was working… but nothing made sense while working on utility-scale solar analytics with mahindra’s cleantech arm, we had just rolled out performance monitoring across 5 sites 30+ inverters generation, PR, CUF, weather data - all coming in everything looked clean data pipeline was solid charts were updating in near real-time. and yet - every site head we spoke to had a different interpretation of the same numbers “irradiance is off” “generation looks low...but it is accurate” “your PR calc is wrong — we benchmark differently” “our SCADA says something else” same system same formula 5 different truths the problem was not data it was interpretation drift every site had different: irradiance sensor placements reference module assumptions panel degradation benchmarks maintenance cycles even different expectations of what “good PR” means some sites cleaned modules weekly others skipped for months one site manually adjusted CUF based on cloud cover another included BESS downtime in the denominator and somehow — the product was expected to “make it all consistent” so what did we do? we tried normalising everything but that just made everyone equally unhappy finally, we paused and reframed the problem: > the goal is not to show one “true” metric the goal is to show context so that disagreement becomes insight - not confusion what we built (and what changed): 1/ confidence badges → every metric showed what it was based on (sensor health, last calibration, fallback logic, etc) → people started trusting the number because we showed its limitations 2/ override interface for ops teams → allowed site heads to annotate data: “PR low due to heavy rain” or “DC isolation ongoing” → dashboard stopped being a passive viewer. became a dialogue. 3/ “why is this low?” traces → our system explained, step by step: PR low → due to low irradiance → matched with satellite → no cleaning logs for 14 days → confidence: medium what i actually learned: 1/ the product was not broken. but it wasn’t useful either without context, even the most accurate metric is just a number 2/ truth in cleantech is layered the system can not declare it it has to build for dialogue 3/ product managers in cleantech are translators not between user and dev but between physical behavior and digital expectation the dashboard worked but users did not trust it - because they couldn’t see what it was seeing so we stopped trying to be correct and started trying to be transparent that is how trust gets built not from precision from participation

Brilliant, disruptive deep tech solutions will save our climate. But without establishing trust early on, they might never see the light of day. Here’s 6 steps deep tech climate solutions need to take: 1.Prove your scientific and technical credibility Publish peer-reviewed research or invite third-party studies on your technology. If feasible, share your data openly. Green tech startup H2Pro invented a two-stage process that produces hydrogen using less energy. A Nature Energy study showed their method had a lab efficiency of 98.7% compared with the standard 68%. This gave a huge boost of confidence to potential investors and partners, leading to over $100 million in funding since their inception. 2.Show it works in the real world. Sublime Systems is pioneering low-cost, low-carbon cement. They started with a pilot plant that produced only 100 tons of cement per year. For reference, average cement plants produce a million tons a year. But this pilot project was necessary to win the industry and customers' trust. As a result, they received funding for future plants. Notably, the US government selected them among 6 producers for a $1.6 billion grant. 3.Build a reputable network. Being associated with industry leaders, renowned scientists, and institutions rubs off on you. If they believe in you, why shouldn’t others? Build out a robust advisory board with such experts. This isn’t just about their approval — the guidance will actually build a much better product. Sustainability solutions ranging from Watershed to Tetra Pak have all gained credibility from their excellent advisory panels. 4.Leverage grants and partnerships. For example, Federally Funded Research and Development Centers (FFRDCs) spent over $29 billion in 2023. Partnering with FFRDCs, such as the National Renewable Energy Laboratory, help climate solutions become market-ready technologies. Knowledge partnerships, such as the Network for Business Sustainability (NBS), are valuable, too. They connect you with communities of relevant professionals — scholars, business leaders, policymakers, and more. 5.Regularly update stakeholders Throughout your journey, communicate honestly with stakeholders about progress, challenges, and milestones. By keeping an open dialogue going, your potential customers remain invested in your mission. It shows your commitment and reliability. Deep tech climate ventures need to focus on their groundbreaking solutions — but they also need to win trust along the way. Without it, you might get stuck in the lab forever. We crafts brand messaging that wins over your stakeholders and establishes trust. Reach out to learn how!