Ways to Drive Continuous Improvement in AI Innovation

Most companies are using AI for efficiency. Some are accelerating value creation. A great case study is how Colgate-Palmolive is driving innovation. Here are specific ways they are embedding GenAI across innovation processes to substantlly improve research and product development. These come from an excellent article in MIT Sloan Management Review by Tom Davenport and Randy Bean (link in comments). 💡 AI-Driven Product Concept Generation Accelerates Ideation By linking one AI system that surfaces consumer needs with another that crafts product concepts, Colgate-Palmolive can swiftly generate creative ideas like novel toothpaste flavors. This AI-augmented workflow produces a broader product funnel and allows rapid iteration, enabling more employees to participate in the innovation process under guided human oversight. 🔍 Retrieval-Augmented Generation Enhances Data Reliability The firm’s use of retrieval-augmented generation (RAG) integrates company-specific research, syndicated data, and real-time trends from sources like Google search data. This approach minimizes the risk of hallucinations and ensures that responses are deeply grounded in verified, internal content—delivering more accurate market analysis and trend detection. 🤖 Digital Consumer Twins Validate and Refine Concepts Moving beyond traditional focus groups, the company has developed “digital consumer twins”—virtual representations of real consumer behavior. These digital twins rapidly test hundreds of AI-generated product ideas. Early evaluations show a high level of agreement between virtual feedback and actual consumer responses. This innovation speeds up early-stage concept validation and reduces reliance on slower, more limited human panels. 🔐 Democratizing AI Through a Secure Internal AI Hub Colgate-Palmolive’s AI Hub provides employees with controlled access to advanced AI tools (including models from OpenAI and Google) behind corporate firewalls. Mandatory training on responsible AI use, including guardrails and prompt engineering best practices, ensures that employees harness these tools safely and effectively. Built-in surveys and KPI tracking further enable the company to measure improvements in creativity, productivity, and overall work quality. 🌐 Bridging Traditional Analytics with Next-Gen AI for Measurable Impact By integrating traditional machine learning with cutting-edge generative AI, Colgate-Palmolive is not only boosting operational efficiencies but also driving strategic growth. This seamless blend supports tasks ranging from market research and innovation to marketing content creation—demonstrating a holistic, value-driven approach to adopting AI that is a model for other organizations.

Executives love to say “we use AI.” But their teams have no idea what that means. As a leader diving into AI strategy, you're probably drowning in vendor pitches and "AI-washing." Let's cut through the noise. After working with dozens of enterprise leaders, here's 11 actions that actually move the needle. (The roadmap most skip, but the successful follow): 1. Start with business outcomes, not AI capabilities → Map specific pain points to potential AI solutions. → Skip the "AI for AI's sake" trap. 2. Audit your data infrastructure → AI runs on data. → Review your data quality, accessibility, and governance before any major AI initiative. 3. Build cross-functional AI committees → Include IT, legal, HR, and business units. → Siloed AI projects rarely succeed. 4. Create an AI ethics framework → Define boundaries early. → Include bias detection, transparency requirements, and usage guidelines. 5. Invest in AI literacy programs → Train leaders first, then teams. → Focus on capabilities and limitations, not technical details. 6. Map process dependencies → Document which processes affect others before automation. → Prevent the "broken pipeline" problem. 7. Start small, iterate fast → Launch contained pilot projects.   → Learn from failures in controlled environments. 8. Define clear success metrics → Track technical (model accuracy) and business (ROI, efficiency) metrics. → Align KPIs with specific use cases to measure impact across user experience, operations, and revenue. 9. Build vs. buy decision framework → Create criteria for when to develop in-house vs. partner with vendors. → Include cost-benefit analysis over time, factoring in scalability and technical debt. 10. Plan for AI maintenance → Account for model drift, data updates, and performance monitoring. → Define ownership and accountability for ongoing model governance. 11. Create feedback loops → Establish channels for user feedback and continuous improvement. → Incorporate structured feedback into product roadmaps and iteration cycles. The difference between successful AI adoption and expensive experiments is rarely the technology. It's almost always the strategy and execution framework. Start with these steps, adjust for your context, and build from there. Which of these tactics resonates most with you? Share below ⬇️ ___ ♻️Found this helpful? Repost it to your network. Follow Carolyn Healey for more practical insights on AI strategy.

I have been developing Agentic Systems for the past few years and the same patterns keep emerging. 👇 𝗘𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 𝗗𝗿𝗶𝘃𝗲𝗻 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 is the most reliable way to be successful in building your 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗦𝘆𝘀𝘁𝗲𝗺𝘀 - here is my template. Let’s zoom in: 𝟭. Define a problem you want to solve: is GenAI even needed? 𝟮. Build a Prototype: figure out if the solution is feasible. 𝟯. Define Performance Metrics: you must have output metrics defined for how you will measure success of your application. 𝟰. Define Evals: split the above into smaller input metrics that can move the key metrics forward. Decompose them into tasks that could be automated and move the given input metrics. Define Evals for each. Store the Evals in your Observability Platform. ℹ️ Steps 𝟭. - 𝟰. are where AI Product Managers can help, but can also be handled by AI Engineers. 𝟱. Build a PoC: it can be simple (excel sheet) or more complex (user facing UI). Regardless of what it is, expose it to the users for feedback as soon as possible. 𝟲. Instrument your application: gather traces and human feedback and store it in an Observability Platform next to previously stored Evals. 𝟳. Run Evals on traced data: traces contain inputs and outputs of your application, run evals on top of them. 𝟴. Analyse Failing Evals and negative user feedback: this data is gold as it specifically pinpoints where the Agentic System needs improvement. 𝟵. Use data from the previous step to improve your application - prompt engineer, improve AI system topology, finetune models etc. Make sure that the changes move Evals into the right direction. 𝟭𝟬. Build and expose the improved application to the users. 𝟭𝟭. Monitor the application in production: this comes out of the box - you have implemented evaluations and traces for development purposes, they can be reused for monitoring. Configure specific alerting thresholds and enjoy the peace of mind. ✅ 𝗖𝗼𝗻𝘁𝗶𝗻𝘂𝗼𝘂𝘀 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗼𝗳 𝘆𝗼𝘂𝗿 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻: ➡️ Run steps 𝟲. - 𝟭𝟬. to continuously improve and evolve your application. ➡️ As you build up in complexity, new requirements can be added to the same application, this includes running steps 𝟭. - 𝟱. and attaching the new logic as routes to your Agentic System. ➡️ You start off with a simple Chatbot and add a route that can classify user intent to take action (e.g. add items to a shopping cart). What is your experience in evolving Agentic Systems? Let me know in the comments 👇

It's not sexy to say, but most of AI transformation has nothing to do with AI. There are 10 steps in the sequence of making an internal process or external product AI-native. Only 1 step is AI, and ironically, the other 9 steps are the far harder part. Step 1: Identify the problem - Find the manual process worth automating. turn your brain off autopilot & turn on your "suck meter". - Funny enough, your company becomes more efficient just by mapping out your processes even if you don't introduce AI. Step 2: Understand the workflow - Map how people actually work today. grab an 8.5x11 piece of paper or Excalidraw and create a flow chart of the workflow from beginning to end. - Least sexy part, but generally where the people driving transformation (FDE, GTM engineer, etc) should spend the majority of their time. Step 3: Collect the data - Gather sample inputs, documents, edge cases - Example: for my content machine ai workflow, I gathered past slack messages/notion transcripts to test automated ideation Step 4: Build the prototype [The AI Part] - Whether its engineer-led or SME-led the goal is to test your hypothesis that there's a better way of doing things for yourself as customer zero. Don't worry about code cleanliness, don't worry about scalability. Step 5: Test & iterate - Before you take the process from single player (only you using it) to multiplayer (many users), you want to beat it up with as many rounds of work & feedback + edge cases as possible. Turning every process into a self-improving loop before scaling is key. Step 6: Integrate with systems - Point-in-time data is good for testing the workflow, but live data is necessary before going into production. Step 7: Roll out & train - Whether the new process lives on a live link, on GitHub or an internal library, next step is hand-holding your peers/users through the onboarding process of your new workflow/product. Step 8: Drive adoption - Embed the workflow in your culture where adoption is tracked, ideas & feedback are celebrated, and new/creative use cases become social currency in your business. Step 9: Empower contribution - Treat your new process like an opensource project. Allow users to become contributors. Whether they are literally pushing code or are simply empowered to add ideas/feedback to a kanban board that gets serviced by engineers, make everyone feel like a builder. Step 10: Measure & capture value - If you're in the experimental phase of AI adoption in your company, fuck ROI. The goal is to empower people to throw a lot of shit at the wall & see what's worth focusing on. You don't need to be scientific during this process. - If you're in the scale-up phase of AI in your business, and you need to realize hard ROI, you need to reskill employees attached to this process, undershoot your approved hiring roadmap, or measurably increase ACV/conversion rate/sales cycle speed.

Stop bolting AI onto broken workflows. It’s the most expensive mistake enterprises are making today. When you layer AI over an inefficient legacy process, you just automate chaos and accelerate bad results. I just finished reading “The Algorithm” by Jon McNeill (former President of Tesla and COO of Lyft) during my vacation. While the book draws heavily on operational marvels at Tesla and SpaceX, its core truth is a blueprint for any leader driving digital transformation. To build a truly "Data Ready, AI First" organisation, we must embrace McNeill's 5-step framework to fundamentally reimagine workflow, not just bolting AI to it. The 5 Steps of the Algorithm (And Why AI Comes Last) 1. Question every requirement: Challenge the status quo ruthlessly. If a rule or policy exists "because we've always done it that way," it’s time to hunt it down and interrogate it. 2. Delete every possible step: The most efficient step is the one that doesn't exist. Focus on the friction points and design them out entirely. 3. Simplify and optimise: Map your process cold. Strip away everything the customer doesn’t care about or pay for. If you can't eliminate a back-end step, make it completely invisible to the user. 4. Accelerate cycle time: Speed up the new process until it breaks. The breaking point is your best diagnostic tool as it tells you exactly where the next bottleneck lies. 5. Automate LAST: This is the golden rule. Smart teams want to code solutions immediately. But if you automate before you simplify, your code becomes concrete which is incredibly difficult and expensive to change. Hold off on the AI until the workflow is pristine. Embedding the 3 Cultural Practices for Enterprise Scale Efficiency is a math problem; scaling it is a culture problem. McNeill outlines 3 practices we can embed within our organisations to drive continuous improvement: Widen the Aperture: Don't just look at your core product; look at the entire customer journey. What are they doing before and after they touch your service? Expand your definition of the product to solve for their entire experience. The Weekly Cadence: Create relentless urgency and accountability. Establish a rhythm where teams report directly to leadership on the top 2–3 most pressing systemic problems weekly. High visibility drives high velocity. Eat Your Own Dog Food: Leaders must create rapid, firsthand feedback loops. Experience your own product, use your own data tools, and feel the friction your clients feel. When leadership is close to the ground, the entire organisational flywheel spins faster. The Bottom Line In an era of epochal technological shifts, maintaining the status quo is a losing strategy. The winners won't be those who buy the most AI licenses; they will be the ones who use this transition to ruthlessly optimise how work actually gets done. Don't pave the cow path. Reimagine the journey. How is your organisation ensuring you simplify before you automate?

A tale of AI and agentic progress in two parts. AI platforms improve at a steady rate, but many are caught off guard when it seems like they suddenly reach par with human capability and reliability levels. Every AI platform and product roadmap must integrate the paradigm of continuous capabilities maturity, or the business will suddenly fall behind. That’s how so many get caught in the hamster wheel of continuously playing catch-up. Wait-and-respond roadmaps don’t work because the pace of progress is too rapid. Technical maturity leaps are no longer rare events. Capabilities maturity cycles have gone from once every 10 years to every 2 years. LLMs went from completely unreliable to high-value tools between 2022 and 2024. Agents went from completely unreliable to high-value tools between 2023 and 2025. Maturity cycles will accelerate as AI tools and agents augment people in new ways. I have taught continuous transformation and capabilities maturity since 2019, but it’s still not common knowledge. AI product roadmaps must be forward-looking and prescriptive to keep up with AI’s progress trend. That means AI product managers must work with technical teams to understand what’s feasible today and what’s most likely to be technically feasible in 1, 2, and 3 years. The roadmap must use a maturity model approach to products and platforms. Infrastructure and tool selection must also follow a maturity model. The business must evaluate how well the vendor’s roadmap aligns with its own. It’s not enough to support today’s business and customer needs. Will the vendor support the business as it progresses along its capabilities maturity model? This should be evaluated for model and information maturity progression. Is the vendor delivering updates rapidly enough? Typically, that means every 6-8 weeks for foundational model updates and every 3-6 months for platform capability upgrades. Vendors should be transitioning from a customer-product model to a platform-partnership model where monetization is tied to outcomes. Business mindsets must adapt to the new reality of continuous capabilities maturity progression and transformation.

Ever asked ChatGPT to “make it more engaging” for the 12th time? That’s not AI’s fault. It’s your system’s. We’ve reached the phase of AI adoption where most teams are using it... But very few are using it well. The difference? ✅ Systemised prompting ❌ Random experimentation In Q1, we built an internal AI Prompt Library and it’s quietly become one of our most effective productivity tools. Not a fancy app. Just a structured Notion board with every proven prompt we use from support emails to investor memos. Why it works: 1. Prompt Consistency = Better Output LLMs are highly sensitive to how you ask. A single word tweak can change everything. We standardised all prompts with a simple format: Context → Role → Task → Style → Output Format No guesswork. Just quality on repeat. 2. Version Control = Fewer Headaches Generative models change behind the scenes. What worked in May might underperform in July. We tag all prompt updates with a date and reason. Helps us keep quality high and feedback loops clean. 3. Quarterly Reviews = Continuous Improvement Every 90 days, we review the whole library: ✅ What’s working 🛠 What needs tweaking ➕ What’s missing It’s a mini sprint retro for AI. Real-world impact: 🧠 20–30% time saved on strategy, writing, and research 📉 32% drop in duplicated or low-quality output 📈 Higher trust across the team for AI-assisted work 🧩 Easier onboarding for new hires If you’re using AI in your team… Don’t just prompt. Build a prompt system. It’s not about finding the perfect question. It’s about documenting what works — and making it repeatable. For founders: ✅ Build a shared prompt library ✅ Version and log changes ✅ Review it like a product ✅ Treat prompts as systems, not hacks Anyone else started building their own AI playbook or workflow?

The Kaizen of Software Development: Small AI Improvements, Big Results When most people hear the term “Kaizen” (改善), they think of factories and assembly lines. However, continuous improvement isn’t limited to manufacturing. It applies equally to how we develop, test, and deliver software to our customers. Initially, we weren’t utilizing AI in our development process. It felt experimental and unproven. But step by step, we began integrating conversational AI and the results have directly improved the customer experience: - In coding, AI accelerated bug fixing and optimization. Customers see faster product updates and fewer disruptions. - In QA, AI-generated test cases helped us catch edge cases early, resulting in more reliable releases and fewer issues reaching customers. - In documentation, AI transformed technical specs into clear, accessible guides. Customers can now find answers quickly and onboard smoothly. - In support enablement, AI-assisted reviews and FAQs ensure that our knowledge base remains current, providing customers with consistent and accurate information. Every step begins with trial and error. The first attempts weren’t perfect, but that’s precisely how Kaizen works. Small experiments, consistent learning, and steady improvement ultimately compound into faster releases, higher quality, and better experiences for our customers. This is Kaizen in action: continuous, incremental improvements that add up to better products and better experiences for our customers. 💡 What’s one minor improvement you’ve made that had a significant impact on your customers? #Kaizen #ContinuousImprovement #AI #CustomerExperience #ConversationalAI

One of the biggest lessons I've learned working with Fortune 5 companies on AI transformation is that successful AI adoption isn't about having better models. It's about building better systems for innovation. Two methods we use to drive innovation at the speed of AI: Innovation Pods; Vibe to Vetted process 1. Innovation Pods Instead of treating innovation as everyone's side job, we create dedicated, cross-functional teams around each core business capability with the goal of driving and scaling AI innovation. Each pod typically includes: • A program manager who owns the roadmap and backlog. • Subject matter experts who understand the business and define user requirements. • FDE's who build and deploy solutions. These teams meet regularly, prioritize opportunities, validate ideas with users, and continuously improve how work gets done. More importantly, they become the owners of that capability. They don't just build tools—they create documentation, playbooks, and repeatable processes that allow innovation to scale across the business. 2. Vibe to Vetted The second lesson is that democratizing development only works when there's a shared foundation. The goal is simple: anyone in the business should be able to build and extend capabilities—not just software engineers. But freedom without standards creates chaos. We've created a common set of guardrails and process that includes shared skills, Git repositories, design systems, AI agents, design system, security guardrails, and engineering standards. Whether someone builds in Cursor, Claude, Antigravity, or another AI-native IDE, every project follows the same architecture, uses the same UI components, meets security requirements, and can be easily maintained by the next team because it has all of the correct documentation. Enterprise AI isn't a technology challenge. It's an operating model challenge. The organizations that build repeatable systems for innovation will outperform those that simply deploy more AI tools. Put these into practice for you and watch how fast innovation takes off.

𝗪𝗵𝘆 𝗣𝘀𝘆𝗰𝗵𝗼𝗹𝗼𝗴𝗶𝗰𝗮𝗹 𝗦𝗮𝗳𝗲𝘁𝘆 𝗜𝘀 𝘁𝗵𝗲 𝗦𝗲𝗰𝗿𝗲𝘁 𝗔𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲 𝗶𝗻 𝗔𝗜 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻 Most AI projects don’t fail because the model is inaccurate, the data is messy, or the tools are complex. They fail because employees don’t feel safe enough to 𝘶𝘴𝘦 the AI. People fear looking incompetent. People fear judgment. People fear being replaced. People fear being the first to try something new. And until leaders address this, AI adoption will always be slow, silent, and resisted. This is the hidden truth behind most stalled AI initiatives. 𝗧𝗵𝗲 𝗛𝘂𝗺𝗮𝗻–𝗔𝗜 𝗪𝗼𝗿𝗸𝗳𝗹𝗼𝘄 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 𝗧𝗵𝗮𝘁 𝗔𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗪𝗼𝗿𝗸𝘀 Not a new platform. Not another training session. Not a bigger budget. A 𝘤𝘶𝘭𝘵𝘶𝘳𝘦 where people can experiment, question, and learn without fear. Here is the integrated strategy that forward-thinking organisations are using. → 𝗣𝘀𝘆𝗰𝗵𝗼𝗹𝗼𝗴𝗶𝗰𝗮𝗹 𝗨𝗿𝗴𝗲𝗻𝗰𝘆 Explain why AI matters, but remove fear-based narratives. Replace 𝘵𝘩𝘳𝘦𝘢𝘵 𝘭𝘢𝘯𝘨𝘶𝘢𝘨𝘦 with 𝘰𝘱𝘱𝘰𝘳𝘵𝘶𝘯𝘪𝘵𝘺 𝘭𝘢𝘯𝘨𝘶𝘢𝘨𝘦. → 𝗖𝗼𝗮𝗹𝗶𝘁𝗶𝗼𝗻 𝗼𝗳 𝗧𝗿𝘂𝘀𝘁 Involve respected frontline employees, not only senior leaders. Adoption spreads faster when it’s modelled by peers. → 𝗖𝗼-𝗗𝗲𝘀𝗶𝗴𝗻 𝗪𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀 Ask employees how AI can support their tasks. When people help build the process, they commit to the outcome. → 𝗦𝗮𝗳𝗲-𝘁𝗼-𝗙𝗮𝗶𝗹 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗟𝗮𝗯𝘀 Create controlled environments where teams can test AI, make mistakes, and try again. Without judgement. Without penalties. → ↳ 𝘛𝘩𝘪𝘴 𝘪𝘴 𝘸𝘩𝘦𝘳𝘦 𝘤𝘰𝘯𝘧𝘪𝘥𝘦𝘯𝘤𝘦 𝘪𝘴 𝘣𝘶𝘪𝘭𝘵, 𝘯𝘰𝘵 𝘪𝘯 𝘗𝘰𝘸𝘦𝘳𝘗𝘰𝘪𝘯𝘵 𝘵𝘳𝘢𝘪𝘯𝘪𝘯𝘨. → 𝗜𝘁𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 Use simple feedback loops Try → Reflect → Redesign → Try again. This approach builds skill, trust, and continuous improvement. 𝗪𝗵𝘆 𝗧𝗵𝗶𝘀 𝗪𝗼𝗿𝗸𝘀 Because AI adoption is not a technical shift. It is a behavioural shift. If people feel safe, they experiment. If they experiment, they learn. If they learn, they adopt. If they adopt, AI scales. Culture drives capability. Safety drives innovation. And together, they drive adoption. 𝗔 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻 𝗙𝗼𝗿 𝗟𝗲𝗮𝗱𝗲𝗿𝘀 𝗪𝗵𝗼 𝗔𝗿𝗲 𝗦𝗲𝗿𝗶𝗼𝘂𝘀 𝗔𝗯𝗼𝘂𝘁 𝗔𝗜 Is psychological safety built into your AI strategy? Or is your organisation still relying on training alone to change behaviour? One approach creates genuine transformation. The other creates quite a resistance. 𝘐𝘧 𝘵𝘩𝘪𝘴 𝘢𝘭𝘪𝘨𝘯𝘦𝘥 𝘸𝘪𝘵𝘩 𝘺𝘰𝘶𝘳 𝘵𝘩𝘪𝘯𝘬𝘪𝘯𝘨, 𝘵𝘢𝘱 𝘭𝘪𝘬𝘦 👍, 𝘧𝘰𝘭𝘭𝘰𝘸 Jyothish Nair 𝘢𝘯𝘥 𝘴𝘩𝘢𝘳𝘦 ♻️ 𝘵𝘰 𝘦𝘭𝘦𝘷𝘢𝘵𝘦 𝘵𝘩𝘦 𝘤𝘰𝘯𝘷𝘦𝘳𝘴𝘢𝘵𝘪𝘰𝘯 𝘰𝘯 𝘈𝘐 𝘢𝘥𝘰𝘱𝘵𝘪𝘰𝘯, 𝘰𝘳𝘨𝘢𝘯𝘪𝘴𝘢𝘵𝘪𝘰𝘯𝘢𝘭 𝘵𝘳𝘢𝘯𝘴𝘧𝘰𝘳𝘮𝘢𝘵𝘪𝘰𝘯, 𝘢𝘯𝘥 𝘵𝘩𝘦 𝘧𝘶𝘵𝘶𝘳𝘦 𝘰𝘧 𝘩𝘶𝘮𝘢𝘯 𝘈𝘐 𝘤𝘰𝘭𝘭𝘢𝘣𝘰𝘳𝘢𝘵𝘪𝘰𝘯. #AIAdoption #ChangeManagement #PsychologicalSafety #DigitalTransformation #FutureOfWork