Tips for Building Transparent AI Models

Data privacy and ethics must be a part of data strategies to set up for AI. Alignment and transparency are the most effective solutions. Both must be part of product design from day 1. Myths: Customers won’t share data if we’re transparent about how we gather it, and aligning with customer intent means less revenue. Instacart customers search for milk and see an ad for milk. Ads are more effective when they are closer to a customer’s intent to buy. Instacart charges more, so the app isn’t flooded with ads. SAP added a data gathering opt-in clause to its contracts. Over 25,000 customers opted in. The anonymized data trained models that improved the platform’s features. Customers benefit, and SAP attracts new customers with AI-supported features. I’ve seen the benefits first-hand working on data and AI products. I use a recruiting app project as an example in my courses. We gathered data about the resumes recruiters selected for phone interviews and those they rejected. Rerunning the matching after 5 select/reject examples made immediate improvements to the candidate ranking results. They asked for more transparency into the terms used for matching, and we showed them everything. We introduced the ability to reject terms or add their own. The 2nd pass matches improved dramatically. We got training data to make the models better out of the box, and they were able to find high-quality candidates faster. Alignment and transparency are core tenets of data strategy and are the foundations of an ethical AI strategy. #DataStrategy #AIStrategy #DataScience #Ethics #DataEngineering

Medical AI can't earn clinicians' trust if we can't see how it works - this review shows where transparency is breaking down and how to fix it. 1️⃣ Most medical AI systems are "black boxes", trained on private datasets with little visibility into how they work or why they fail. 2️⃣ Transparency spans three stages: data (how it's collected, labeled, and shared), model (how predictions are made), and deployment (how performance is monitored). 3️⃣ Data transparency is hampered by missing demographic details, labeling inconsistencies, and lack of access - limiting reproducibility and fairness. 4️⃣ Explainable AI (XAI) tools like SHAP, LIME, and Grad-CAM can show which features models rely on, but still demand technical skill and may not match clinical reasoning. 5️⃣ Concept-based methods (like TCAV or ProtoPNet) aim to explain predictions in terms clinicians understand - e.g., redness or asymmetry in skin lesions. 6️⃣ Counterfactual tools flip model decisions to show what would need to change, revealing hidden biases like reliance on background skin texture. 7️⃣ Continuous performance monitoring post-deployment is rare but essential - only 2% of FDA-cleared tools showed evidence of it. 8️⃣ Regulatory frameworks (e.g., FDA's Total Product Lifecycle, GMLP) now demand explainability, user-centered design, and ongoing updates. 9️⃣ LLMs (like ChatGPT) add transparency challenges; techniques like retrieval-augmented generation help, but explanations may still lack faithfulness. 🔟 Integrating explainability into EHRs, minimizing cognitive load, and training clinicians on AI's limits are key to real-world adoption. ✍🏻 Chanwoo Kim, Soham U. Gadgil, Su-In Lee. Transparency of medical artificial intelligence systems. Nature Reviews Bioengineering. 2025. DOI: 10.1038/s44222-025-00363-w (behind paywall)

As AI becomes integral to our daily lives, many still ask: can we trust its output? That trust gap can slow progress, preventing us from seeing AI as a tool. Transparency is the first step. When an AI system suggests an action, showing the key factors behind that suggestion helps users understand the “why” rather than the “what”. By revealing that a recommendation that comes from a spike in usage data or an emerging seasonal trend, you give users an intuitive way to gauge how the model makes its call. That clarity ultimately bolsters confidence and yields better outcomes. Keeping a human in the loop is equally important. Algorithms are great at sifting through massive datasets and highlighting patterns that would take a human weeks to spot, but only humans can apply nuance, ethical judgment, and real-world experience. Allowing users to review and adjust AI recommendations ensures that edge cases don’t fall through the cracks. Over time, confidence also grows through iterative feedback. Every time a user tweaks a suggested output, those human decisions retrain the model. As the AI learns from real-world edits, it aligns more closely with the user’s expectations and goals, gradually bolstering trust through repeated collaboration. Finally, well-defined guardrails help AI models stay focused on the user’s core priorities. A personal finance app might require extra user confirmation if an AI suggests transferring funds above a certain threshold, for example. Guardrails are about ensuring AI-driven insights remain tethered to real objectives and values. By combining transparent insights, human oversight, continuous feedback, and well-defined guardrails, we can transform AI from a black box into a trusted collaborator. As we move through 2025, the teams that master this balance won’t just see higher adoption: they’ll unlock new realms of efficiency and creativity. How are you building trust in your AI systems? I’d love to hear your experiences. #ArtificialIntelligence #RetailAI

Demandbase has used AI to score 38B accounts, predict 4M opportunities, and launch 20k outcome-based advertising campaigns. Here are 3 best practices for using AI in your account-based GTM: 1. Start with data AI strategy should start with data cleansing and enrichment. Not all data is equal, it’s important to understand what signal matters most and to focus on quality over quantity – you don’t need 150M contacts weighing down CRM, you need 100k highly accurate contacts from your ICP. 2. Build healthy models There are three best practices here too: (i) Know what the strongest signals are. For example, for tech companies generally technographics, industry, and revenue ranges are strong signals for ICP models, while campaign responses, sales activities, website engagement, and intent are strong signals for pipeline prediction models. (ii) Build specialized models for different products, regions, and aspects of your GTM. For example, models focused on acquisitions of new logos, models focused on customer retention, and models focused on gross retention. (iii) Models need to be re-trained frequently to avoid following behind your GTM evolution. 3. Avoid black boxes AI models have to be transparent. Without transparency you can’t tell if the AI model is making a recommendation that you know for obvious reasons is flawed. Transparency enables Marketing and Sales to improve their messaging and activation by learning directly from model recommendations. And transparency is critical for data science teams at your company driving AI strategy across the enterprise. There’s a lot of hype and promise in AI. What’s working best for account-based GTM’s is focusing on the strongest signal, prioritizing quality of data over quantity, using specialized models, re-training models frequently, and making sure AI is transparent.

AI products like Cursor, Bolt and Replit are shattering growth records not because they're "AI agents". Or because they've got impossibly small teams (although that's cool to see 👀). It's because they've mastered the user experience around AI, somehow balancing pro-like capabilities with B2C-like UI. This is product-led growth on steroids. Yaakov Carno tried the most viral AI products he could get his hands on. Here are the surprising patterns he found: (Don't miss the full breakdown in today's bonus Growth Unhinged: https://lnkd.in/ehk3rUTa) 1. Their AI doesn't feel like a black box. Pro-tips from the best: - Show step-by-step visibility into AI processes - Let users ask, “Why did AI do that?” - Use visual explanations to build trust. 2. Users don’t need better AI—they need better ways to talk to it. Pro-tips from the best: - Offer pre-built prompt templates to guide users. - Provide multiple interaction modes (guided, manual, hybrid). - Let AI suggest better inputs ("enhance prompt") before executing an action. 3. The AI works with you, not just for you. Pro-tips from the best: - Design AI tools to be interactive, not just output-driven. - Provide different modes for different types of collaboration. - Let users refine and iterate on AI results easily. 4. Let users see (& edit) the outcome before it's irreversible. Pro-tips from the best: - Allow users to test AI features before full commitment (many let you use it without even creating an account). - Provide preview or undo options before executing AI changes. - Offer exploratory onboarding experiences to build trust. 5. The AI weaves into your workflow, it doesn't interrupt it. Pro-tips from the best: - Provide simple accept/reject mechanisms for AI suggestions. - Design seamless transitions between AI interactions. - Prioritize the user’s context to avoid workflow disruptions. -- The TL;DR: Having "AI" isn’t the differentiator anymore—great UX is. Pardon the Sunday interruption & hope you enjoyed this post as much as I did 🙏 #ai #genai #ux #plg

AI success isn’t just about innovation - it’s about governance, trust, and accountability. I've seen too many promising AI projects stall because these foundational policies were an afterthought, not a priority. Learn from those mistakes. Here are the 16 foundational AI policies that every enterprise should implement: ➞ 1. Data Privacy: Prevent sensitive data from leaking into prompts or models. Classify data (Public, Internal, Confidential) before AI usage. ➞ 2. Access Control: Stop unauthorized access to AI systems. Use role-based access and least-privilege principles for all AI tools. ➞ 3. Model Usage: Ensure teams use only approved AI models. Maintain an internal “model catalog” with ownership and review logs. ➞ 4. Prompt Handling: Block confidential information from leaking through prompts. Use redaction and filters to sanitize inputs automatically. ➞ 5. Data Retention: Keep your AI logs compliant and secure. Define deletion timelines for logs, outputs, and prompts. ➞ 6. AI Security: Prevent prompt injection and jailbreaks. Run adversarial testing before deploying AI systems. ➞ 7. Human-in-the-Loop: Add human oversight to avoid irreversible AI errors. Set approval steps for critical or sensitive AI actions. ➞ 8. Explainability: Justify AI-driven decisions transparently. Require “why this output” traceability for regulated workflows. ➞ 9. Audit Logging: Without logs, you can’t debug or prove compliance. Log every prompt, model, output, and decision event. ➞ 10. Bias & Fairness: Avoid biased AI outputs that harm users or breach laws. Run fairness testing across diverse user groups and use cases. ➞ 11. Model Evaluation: Don’t let “good-looking” models fail in production. Use pre-defined benchmarks before deployment. ➞ 12. Monitoring & Drift: Models degrade silently over time. Track performance drift metrics weekly to maintain reliability. ➞ 13. Vendor Governance: External AI providers can introduce hidden risks. Perform security and privacy reviews before onboarding vendors. ➞ 14. IP Protection: Protect internal IP from external model exposure. Define what data cannot be shared with third-party AI tools. ➞ 15. Incident Response: Every AI failure needs a containment plan. Create a “kill switch” and escalation playbook for quick action. ➞ 16. Responsible AI: Ensure AI is built and used ethically. Publish internal AI principles and enforce them in reviews. AI without policy is chaos. Strong governance isn’t bureaucracy - it’s your competitive edge in the AI era. 🔁 Repost if you're building for the real world, not just connected demos. ➕ Follow Nick Tudor for more insights on AI + IoT that actually ship.

Stop Showing Me Agentic AI. Show Me Your Data Model. Another week, another enterprise conference and endless demos. Agents talking to agents. Avatars smiling. Everyone in the room nodding when they really are doubting the pitch. We’re in a phase where AI marketing has outpaced AI maturity. Every platform claims “agentic capabilities.” But very few can trace their own data lineage or explain how their models actually make decisions. Here’s the truth. Applied AI is infrastructure. And the infrastructure that separates hype from reality comes down to three things: 1. The Glass Model — Visibility If I can’t see why the model made a decision, I can’t trust it. A glass model means full observability: inference logs, bias checks, attribution maps, and performance drift reports. Show me explainability dashboards, not animated demos. 2. Data Labeling — Integrity Every insight is only as good as the data that trained it. Who labeled your data? How accurate is it? How do you measure drift and quality over time? If you can’t answer that, you don’t have intelligence, you have noise disguised as insight. 3. Data Cataloging — Governance If you don’t know where your data lives, who owns it, or how it connects, your “AI” is built on sand. Real applied AI runs on structured, versioned, lineage-tracked data. That’s how you prevent bias, ensure compliance, and scale with confidence. Too many vendors are hiding behind buzzwords. They call it “agentic.” They say it “learns on the fly.” Translation: they haven’t built the plumbing yet. When done right, it’s not flashy, it’s harmonized. The ontology connects people, jobs, skills, and content. Labels evolve under defined governance. Models are transparent, auditable, and explainable in real time. That’s applied AI, transparent, traceable, and trusted. So here’s my advice to every enterprise buyer: Stop being impressed by animated demos and agent chatter. Ask tougher questions. Show me your ontology Show me your labeling standards and quality metrics Show me your data catalog: lineage, access, and ownership Show me your bias and drift dashboards Show me how you roll back a model that fails If a vendor can’t open the hood, they don’t have applied AI, they have a slick marketing strategy. It’s the framework that makes features possible. 1. Glass Model 2. Labeling 3. Cataloging Everything else is just an advertisement for your shortcomings. — Cliff Jurkiewicz, Chief Strategist, Phenom #AI #AppliedAI #DataStrategy #EnterpriseTech #TrustInAI

👋 Hey Data Pills Community, New Data Pills Newsletter Alert! From Black Box to Glass Box: Demystifying AI Through Explainable AI (XAI) Artificial Intelligence (AI) is transforming our world, from revolutionizing healthcare and finance to powering autonomous systems. But as AI grows more sophisticated, it also becomes more complex—often functioning as a "black box" where decisions are made without clear explanations. ❓ How can we trust AI if we don’t understand how it reaches conclusions? ❓ What happens when AI makes a mistake—in a medical diagnosis, a loan approval, or a self-driving car? ❓ Can we ensure fair, ethical, and bias-free AI if we can't interpret its reasoning? 💡 This is where Explainable AI (XAI) comes in! In our latest Data Pills newsletter, we break down why XAI is crucial, how it helps build trust, accountability, and compliance, and explore leading frameworks that bring transparency to AI decision-making. What’s Inside This Edition? - Why AI Explainability Matters – Understanding how AI models make decisions and why interpretability is critical. - XAI Techniques – From visual explanations like LIME, SHAP, and Grad-CAM to textual and mathematical methods. - AI in High-Stakes Industries – Real-world XAI applications in healthcare, finance, and autonomous driving. - Choosing the Right XAI Tool – A deep dive into LIME vs. SHAP to help you select the best method for model interpretability. - Top Open-Source XAI Frameworks – A curated list of GitHub repositories to start making your AI models more transparent. 🎯 Whether you're an AI researcher, a data scientist, or just an AI enthusiast, this newsletter provides actionable insights on making AI more explainable and ethical. 📩 Read the full newsletter here: 💬 Join the conversation! How important do you think AI transparency is? Have you used any XAI tools in your projects? Drop your thoughts in the comments! ⬇️ #AI #ExplainableAI #XAI #MachineLearning #DataScience #BlackBoxAI #ResponsibleAI #TrustworthyAI