Davinder Kohli’s Post

Imagine a world where AI and digitalization are not just tools for profit, but powerful enablers of global prosperity and well-being. Discover how the principles of 'Sarbat da Bhalla' and 'Seva' can guide us in leveraging technology to drive inclusive growth and uplift underserved communities. Sara Agarwal Sonny Kakar Edna Diez Gurminder Singh Ahuja, MD, FAAOS Fernando Carranza #AI #GenerativeAI #Digitalization #Prosperity

Super stoked about having a virtual fireside chat with Professor Stefano Puntoni at The Wharton School. During our conversation, we will explore the transformative capabilities of AI in enhancing decision-making, examine its potential human impact, and address common fears associated with AI integration. We will also delve into the practical applications of AI, its integration into business processes, and the role of synthetic data in AI. Please join us on August 20th, 2024 at 12p ET. Register: https://lnkd.in/e9DDCAHH Hosted by:  Wharton Exec Ed Alumni Club (WEEAC) Deepak Rana Adriana Muñoz Virginia Sayer

Ready or Not, Here AI Come! - Part 2: Responsibility 🛡️ With power comes responsibility, and accountability must follow. GenAI is becoming increasingly ubiquitous in our digital world—whether it’s CoPilot drafting emails, creating RFP responses, or developing fully functioning code.   Increased automation, especially using GenAI, can have unintended consequences. For example, I recently received a strong (rightfully so) response from a customer whose name was misspelled by an API that extracted their first name from LinkedIn. While this was a rare occurrence—just 0.01%—such issues can’t be entirely ruled out.   But what we can do is minimize the chances of these occurrences and be better prepared to handle the fallout. Here are some strategies to ensure our AI applications are responsible and our organizations are accountable.   🔍 Accept the Possibility of Unintended Consequences - Think through all potential negative scenarios that can occur. Could they harm your company's brand, affect customers, or result in financial loss? Acknowledging these risks is the first step to mitigating them.   📊 Prepare Your Response - Assess your organization’s readiness to handle fallout. Establish clear protocols for responding to incidents quickly and effectively.   🛑 Check for Toxicity in Prompts - Even if a prompt isn’t engineered to extract sensitive information, it could still provoke a response that harms your brand. Implement checks using tools such as Guardrails to identify and mitigate such risks.   🤖 Monitor for Inappropriate Responses - Consider guardrails to detect “hallucinated” responses—incorrect or nonsensical outputs generated by the LLM. Regular audits can help identify and correct these issues.   ⚖️ Address Data Biases—While we can't control the data used to train LLMs, we can control the data used for augmentation (RAG) and fine-tuning. Ensure your data represents all customer segments that represent your target audience.   👩💻 Incorporate Human in the Loop - Implement a human oversight layer in critical AI processes to ensure that decisions made by AI align with ethical standards and business values.   🔄 Implement Continuous Audits - Regularly monitor and audit your GenAI systems to detect and mitigate biases, errors, and other issues over time.   By implementing these strategies, we can create AI systems that are not only powerful but also responsible. I’d love to hear how you’re approaching Responsible AI. #GenerativeAI

Ready or Not, Here AI Come! - Part 1 🔐 Security   Remember the thrill of playing hide-and-seek  🎮, counting down, and finding our friends before they could hide? We felt a similar excitement exploring the niche capabilities of different LLMs, building RAG prototypes, and fine-tuning models. We impressed our business partners with Generative AI's potential to lower operational costs, increase productivity, and boost customer engagement without worrying about production readiness.   Now that we have piqued our business's interest, it’s no longer “ready or not, here AI come” – it’s time to ensure our GenAI apps are enterprise-ready!   🚀 What Makes a GenAI App Enterprise-Ready? An enterprise-ready GenAI application is more than just a functional prototype. It needs to be robust, scalable, secure, and compliant. While many principles of enterprise readiness apply to both AI and non-AI applications, Generative AI poses unique challenges. In this series, I'll focus on three key areas that require special attention: Security, Responsibility, and Data Management.   Part 1: Security Large Language Models (LLMs) are vulnerable to prompt injection attacks—where malicious inputs can manipulate a model’s behavior, potentially revealing proprietary information or generating harmful content.   Here are some considerations for securing GenAI-based applications:   🔒 API-Based Approach: Secure all interactions with LLMs behind robust APIs. This control mechanism allows for data flow regulation and additional security checks.   🔑 Token Authentication: Use technologies like OAuth to ensure only users with valid tokens can access the application. This prevents unauthorized access.   🛡️ Role-Based Access Control (RBAC): Assign roles and permissions based on user groups, ensuring only authorized personnel can access specific functionalities.   🚨 Detect Improper Prompts: Incorporate tools like Guardrails to detect and mitigate improper prompts, preventing unintended outputs that could be damaging.   📂 Secure Retrieval from Vector Stores: Ensure data retrieved from vector stores during RAG implementations is validated and sanitized, preventing unauthorized access or manipulation.   🔐 Protect Sensitive Data in Augmentation: Use encryption and anonymization when augmenting models with proprietary data. Consider policy-based partitioning of data to further secure it.   🔍 Filter Output for Sensitive Data: Employ Guardrails to scan AI-generated outputs for sensitive information before they're displayed or stored.   📝 Audit Prompts and Outputs: Regularly audit log files to detect sensitive information, prompt injections, and unwanted revelations in LLM responses.   Once you have the right security measures in place, your GenAI applications will be better equipped to handle the challenges of the real world. What techniques are you using to secure your GenAI-based applications? #GenerativeAI #Security Creospan Inc. Dhairya B. Kevin E Dooley Michael Lyons

Building the Right Team for Business Success 👥 Whether you're a startup founder or a business leader in a Fortune 100 company, the right team makes all the difference. VCs and executives alike invest in teams, not just ideas or the promise of 10x returns. So, what defines the "right" team?   It's a team with diverse skills, mutual trust, empathy, and a shared goal. As a leader, it's your responsibility to assemble this team. Here's what I've learned from my experience in my current company and during my startup:   🧐 Clarify the Problem - Gather evidence supporting the problem, define scenarios, and understand its magnitude. This helps in selecting the team members with the necessary skills and emotional intelligence.   💡 State the WHY - Clearly articulate why solving this problem matters. A compelling WHY motivates the team to stay committed and driven.   🎩 Define Your Role - Identify the unique tasks only you can handle, ensuring you focus on areas where you add the most value.   🤝 Collaborate with Peers - Discuss potential solutions with peers from technology, operations, sales, marketing, and legal to gain diverse perspectives. This will help identify the team members required to pull off the initiative.   🔍 Leverage Diverse Perspectives - Build a team with diverse backgrounds and experiences to foster innovation and comprehensive problem-solving.   🖼️ Keep a Big Picture View - Use visual aids to represent problem scenarios and solutions. This will make discussions with your team more grounded and focused.   📜 Establish Guiding Principles - Set principles that align the team and keep everyone focused on the problem at hand.   💸 Monitor Financials - Plan finances meticulously. Determine the seed investment, resources needed, and long-term financial requirements for sustaining core and supplemental teams.   With the right team in place, leverage digital collaboration tools to share ideas and maintain cohesion in the work environment. What strategies do you use to assemble the right team? #Leadership #OrganizationalBehavior 

Getting Business Leaders to Listen to Your GenAI Solutions 🚀 Technologists have moved from building proof-of-concepts to scalable, enterprise-ready Gen AI architectures. Yet, many struggle to capture the attention of their business leaders. Why?   Technologists often fall into the trap of discussing Gen AI features, advanced scalable architectures like RAG, fine-tuning, and techniques for generating, extracting, and summarizing multimodal content. While impressive in peer-to-peer technical discussions, business leaders want solutions that drive revenue growth, lower customer acquisition costs, and boost workforce productivity.   If you want to engage your business leaders on Generative AI, consider these strategies:   🧠 Know Your Business - Understand your company’s business model and the processes that support revenue generation and critical operations.   🔍 Identify Candidate Business Processes - Pinpoint one or two processes ripe for optimization, reducing customer friction, or adding value.   📊 Establish a Baseline with Numbers - For those processes, determine current metrics such as NPS for customer engagement, time to purchase, cost to serve, and customer volume served.   💡 Focus on Value Creation - Explain how Generative AI can benefit customers in these processes. Highlight what becomes possible with AI that isn't possible today. For retail, this might mean providing a highly personalized experience based on customer intent.   💰 Determine Value Capture - Show how the value created by Generative AI translates into profit or cost reduction. For retail, this might mean upselling or cross-selling; for healthcare, it could be reducing hospital length of stay.   ⚖️ Solution Implications - Evaluate the social, legal, and ethical implications of the proposed solution. For healthcare, consider how an AI solution might be biased towards a particular group of patients or may inadvertently disclose sensitive information.   🚀 Propose a Pilot - Outline a pilot for your Gen AI initiative, define success metrics, define MVP features, identify team members, estimate the investment needed, and the ROI.   Adopting these strategies may help you build stronger relationships with your business teams. How do you engage with your business leaders? Creospan Inc. #GenerativeAI #DigitalStrategy

The Art of Creating a Transformational Vision 🎯 Transformational visions aren't just about cutting-edge tech—they're about inspiring your entire organization to see beyond the horizon and work together to create something extraordinary. 🌟   A digital leader's success is measured by their ability to grow revenue, optimize operational efficiency, retain and develop talent retention, and enhance brand reputation. Here are some pointers to help digital leaders craft and execute a compelling vision:    👂 Listen to Customers - Understand what’s working, identify friction points, and recognize their evolving needs.   📊 Understand Customer Segments - Segment customers based on CLV, needs, and demographics to determine the total addressable market for each segment.   🎯 Align with Company Strategy - Ensure your vision supports the company’s broader goals, such as entering new markets or driving growth in specific areas.   💰 Revenue Stream Analysis - Assess your business portfolio’s revenue streams, identify underperforming areas, and highlight high operational expenses.   🤖 Explore Transformative Technologies - Identify automation and customer experience enhancement opportunities across the entire business flow. Technologies like Generative AI can modernize workflows, reduce friction, and streamline digital journeys.    💸 Financial Investments - Develop financial models detailing required investments, operational expenses, break-even points, projected revenue, and cost savings for implementing new digital solutions.   🤝 Forge Strategic Partnerships - Focus on core competencies and identify necessary partnerships to augment capabilities across the value chain and scale capacity.    📅 Organizational Readiness - Evaluate the organization’s readiness in terms of talent, technical debt, technology integration, and timelines.    🗣️ Stakeholder Engagement - Regularly communicate with all stakeholders to align expectations and gather valuable feedback.   🌟 Cultivate a Culture of Innovation - Foster a culture that encourages creativity, agility, and continuous learning to support the transformational vision. #GenerativeAI #Leadership #DigitalStrategy #Innovation What strategies have you found effective in crafting a compelling vision? ✨

AI or Not?: Key Considerations for Your Business 🤔 Every organization aims to integrate Generative AI into its operations, and for good reason—AI offers the potential for significant growth 🚀 and efficiency. However, not every problem requires an AI solution. Here are some scenarios where using AI might not be the best approach: 📝 Unclear Problem Statement: If the problem isn't well-defined, developing an AI solution can lead to confusion and inefficiency. 🔄 Rules-Based Logic: When the solution can be easily coded with a set of rules, a traditional approach is often more effective and straightforward. ⏳ One-off Scenarios: Investing in AI might be overkill for one-off problems. Simpler solutions can be more cost-effective. 💸 Cost vs. Benefit: If the cost (in terms of time and money) to develop an AI solution outweighs the potential benefits, it's better to seek alternative methods. 👤 Absence of Human in the Loop: AI systems often require human oversight to ensure accuracy and accountability. Without this, the system can produce unreliable results. 🔧 Need for Determinism: When you need a 100% deterministic system, traditional methods are more suitable than probabilistic AI models.  📉Lack of Quality Data: AI thrives on high-quality data. AI will struggle to produce meaningful results if your data is incomplete or unreliable.  🔐 Ethical and Privacy Concerns: In situations involving sensitive information or high ethical considerations, the risks might outweigh the benefits of using AI. By carefully evaluating these factors, you can ensure that AI is used effectively and responsibly within your organization. Remember, not every problem is a nail that needs an AI hammer! 🔨✨ Have you encountered situations where AI wasn't the right fit? #GenerativeAI #DigitalStrategy #BusinessStrategy Creospan Inc.

Unlock the power of AI to simplify your database queries with natural language commands! #Creospan #GenerativeAI

Had a great time at the Wharton Reunion Weekend, catching up with friends and attending inspirational sessions on building a personal brand with Prof Americus Reed II. AI took the main stage! Wharton is #1 for AI student opportunities and experiences. Prof Ethan Mollick's keynote on Generative AI was just fabulous! Peyman Askari Narayan Chintalapati Pratima Poplai Betty Ryberg Surya Khoday Trish Skoglund Adriana Muñoz William "Billy" Greineisen, Ph.D. #GenerativeAI #Wharton