The Impact of AI on Data Accuracy

𝐃𝐚𝐭𝐚 𝐪𝐮𝐚𝐥𝐢𝐭𝐲 𝐢𝐬𝐧'𝐭 𝐣𝐮𝐬𝐭 𝐢𝐦𝐩𝐨𝐫𝐭𝐚𝐧𝐭 𝐟𝐨𝐫 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐀𝐈—𝐢𝐭'𝐬 𝐚𝐛𝐬𝐨𝐥𝐮𝐭𝐞𝐥𝐲 𝐜𝐫𝐢𝐭𝐢𝐜𝐚𝐥. AI solutions, particularly those embedded in ERP systems, are designed to deliver valuable insights and recommendations to businesses. However, the 𝐪𝐮𝐚𝐥𝐢𝐭𝐲 𝐚𝐧𝐝 𝐚𝐜𝐜𝐮𝐫𝐚𝐜𝐲 𝐨𝐟 𝐭𝐡𝐞𝐬𝐞 𝐫𝐞𝐜𝐨𝐦𝐦𝐞𝐧𝐝𝐚𝐭𝐢𝐨𝐧𝐬 𝐚𝐫𝐞 𝐝𝐢𝐫𝐞𝐜𝐭𝐥𝐲 𝐥𝐢𝐧𝐤𝐞𝐝 𝐭𝐨 𝐭𝐡𝐞 𝐪𝐮𝐚𝐥𝐢𝐭𝐲 𝐨𝐟 𝐭𝐡𝐞 𝐮𝐧𝐝𝐞𝐫𝐥𝐲𝐢𝐧𝐠 𝐝𝐚𝐭𝐚. In traditional ERP implementations, businesses often found themselves achieving systems that were "on time, on budget, fully functional, and disappointing." Why? Because while the system technically worked, the data feeding it wasn't accurate enough to meet real-world expectations. Incorrect customer addresses, inaccurate inventory data, or faulty financial figures significantly compromised the value of the entire system. 𝐖𝐢𝐭𝐡 𝐀𝐈, 𝐭𝐡𝐞 𝐬𝐭𝐚𝐤𝐞𝐬 𝐚𝐫𝐞 𝐞𝐯𝐞𝐧 𝐡𝐢𝐠𝐡𝐞𝐫. AI-driven recommendations depend heavily on the accuracy and quality of data. If AI bases its recommendations on inaccurate or inconsistent data, users quickly lose trust and confidence in these insights, eventually ignoring them entirely. This lack of trust diminishes the value of AI systems, no matter how sophisticated the algorithms are. 𝐓𝐡𝐞 𝐜𝐨𝐦𝐦𝐨𝐧 𝐧𝐨𝐭𝐢𝐨𝐧 𝐭𝐡𝐚𝐭 "𝐀𝐈 𝐢𝐬 𝐠𝐨𝐨𝐝 𝐚𝐭 𝐰𝐨𝐫𝐤𝐢𝐧𝐠 𝐰𝐢𝐭𝐡 𝐛𝐚𝐝 𝐝𝐚𝐭𝐚" 𝐢𝐬 𝐟𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥𝐥𝐲 𝐟𝐥𝐚𝐰𝐞𝐝. While AI may process large volumes of data quickly, poor-quality input inevitably leads to poor-quality outcomes. 𝐀𝐈 𝐚𝐦𝐩𝐥𝐢𝐟𝐢𝐞𝐬 𝐛𝐨𝐭𝐡 𝐭𝐡𝐞 𝐬𝐭𝐫𝐞𝐧𝐠𝐭𝐡𝐬 𝐚𝐧𝐝 𝐰𝐞𝐚𝐤𝐧𝐞𝐬𝐬𝐞𝐬 𝐨𝐟 𝐲𝐨𝐮𝐫 𝐝𝐚𝐭𝐚—meaning bad data can severely degrade your results and decision-making quality. One of the longstanding strengths of SAP systems is their reliability and trustworthiness. Businesses have confidence in SAP solutions because they know the integrity of their data is preserved and accurately managed throughout the process. This reliability is especially critical in the age of AI, where the value derived is directly proportional to the quality of data provided. 𝐒𝐢𝐦𝐩𝐥𝐲 𝐩𝐮𝐭: 𝐐𝐮𝐚𝐥𝐢𝐭𝐲 𝐝𝐚𝐭𝐚 𝐢𝐬 𝐭𝐡𝐞 𝐟𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐬𝐮𝐜𝐜𝐞𝐬𝐬𝐟𝐮𝐥 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐀𝐈. 𝐖𝐢𝐭𝐡𝐨𝐮𝐭 𝐢𝐭, 𝐞𝐯𝐞𝐧 𝐭𝐡𝐞 𝐦𝐨𝐬𝐭 𝐚𝐝𝐯𝐚𝐧𝐜𝐞𝐝 𝐀𝐈 𝐬𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐬 𝐰𝐨𝐧'𝐭 𝐝𝐞𝐥𝐢𝐯𝐞𝐫 𝐭𝐡𝐞 𝐞𝐱𝐩𝐞𝐜𝐭𝐞𝐝 𝐯𝐚𝐥𝐮𝐞.

𝟯𝟴% 𝗯𝗲𝘁𝘁𝗲𝗿 𝗔𝗜 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆. 𝗡𝗼 𝗻𝗲𝘄 𝗺𝗼𝗱𝗲𝗹. 𝗡𝗼 𝗻𝗲𝘄 𝗱𝗮𝘁𝗮. 𝗝𝘂𝘀𝘁 𝗯𝗲𝘁𝘁𝗲𝗿 𝗰𝗼𝗻𝘁𝗲𝘅𝘁. That’s the headline from a controlled NL-to-SQL experiment discussed by Manoj Shanmugasundaram in Metadata Weekly. Across 522 query evaluations, the only variable that changed was context quality — and it made all the difference. Concise, high-signal context (business definitions, SQL patterns, domain rules) drove a 38% accuracy gain. Verbose, catalog-style documentation? Performance dropped and costs rose. More words diluted the signal. The biggest lift wasn't on simple or extreme queries. It was on medium-complexity ones — the joins and aggregations that make up everyday analytics work — where focused context delivered a 𝟮.𝟭𝟱𝘅 𝗶𝗺𝗽𝗿𝗼𝘃𝗲𝗺𝗲𝗻𝘁. The mindset shift: metadata was built for humans to browse. Now it also needs to work for machines to reason. Most teams are still optimizing for readability, not machine usability. If your "talk to data" initiative is stalling, it might not be a model problem. It might be a context problem. Manoj breaks down what machine-usable context actually looks like — and how to get started without rebuilding your stack. Read it in Metadata Weekly 👇

AI can now interpret your blood test results in seconds. That shift is already changing patient expectations. Access is accelerating faster than understanding. More patients are turning to AI tools to make sense of their lab results, often before speaking with a physician. The appeal is clear: speed, clarity, and control. However, interpretation is not the same as medical judgment. Accuracy is still uncertain. We don’t yet have strong, peer-reviewed evidence that these tools can reliably interpret results or generate meaningful, personalized recommendations. And in many cases, “personalized” is doing more work than the data can support. Confidence is outpacing validation. When outputs sound authoritative, patients may act on them. That can mean unnecessary testing, avoidable anxiety, or missed diagnoses. AI has a role to play. It can help patients better understand their health information and ask more informed questions. That’s progress. As these tools become more embedded in how people engage with their health, we need to ask a different set of questions: What standards of evidence should apply? What role should clinicians play? And who is accountable for the outcome? Because in health care, answers don’t just inform decisions. They shape them. #HealthAI #DigitalHealth #HealthPolicy #PatientSafety https://lnkd.in/eBqbkSbX

Everyone’s talking about AI models, but here’s the truth most overlook: Your AI is only as smart as your data. As the founder of DataGardener, I’ve seen AI transform how #businesses operate—but I’ve also seen promising models fall flat because the data wasn’t good enough. Why Data is the Real Power Behind AI Algorithms don’t work magic. They learn patterns from data. So if your data is: ✔️ Outdated ✔️ Incomplete ✔️ Inaccurate …you’ll get flawed predictions and risky decisions. No matter how advanced the model. #AI learns from patterns. The more diverse and representative your #dataset, the better your models can generalise to real-world scenarios. Two Things Every Business Needs: 1. Accuracy "Garbage in, garbage out" is real. Clean, correct data is the only way to get trustworthy insights. Insufficient data doesn’t just mean bad business—it can lead to bias, compliance risks, and lost revenue. 2. Data Volume More data = better pattern recognition. Large datasets make models more robust, less prone to overfitting. #Diversity in data ensures insights reflect reality—not just a narrow view. How Key Data Attributes Impact AI Quality: #Accuracy → Produces trustworthy, actionable results #Volume → Enables richer insights and model resilience Real-World Impact Real-World Impact At DataGardener, our clients use AI built on verified, comprehensive company data. That’s how they: Make smarter credit decisions Uncover leads others miss Mitigate risks before they become costly The difference? It’s the data. Takeaway for Business Leaders Treat your data like an asset—not a byproduct: invest in data collection, cleaning, and validation. Before chasing the next AI model, fix your foundation. Remember: AI is only as good as the data it learns from. In the age of AI, data stewardship isn’t just IT’s job—it’s a boardroom priority. Curious how high-quality data can power better AI decisions in your business? Let’s talk. Let’s build smarter—starting with the right data. #SmartData #AIDrivenDecision #Data #BusinessLeader #ComplianceRisks #CreditDecisions #AIDecisions

Not All Data is Equal: What Dr. Eli Goldratt Warned Us About in 1982 — And Why It Matters Even More Now in the Age of AI & Digital Twins In his seminal 1982 article, “Over 95% Data Accuracy—A Need or Just a Myth?”, Dr. Eli Goldratt challenged a dangerous assumption still prevalent today: That we must collect accurate, "real-time" data before we can make good decisions. In the video below from the late 1980's, Dr Goldratt proposed something radically practical instead: “Only a few data points actually matter. Focus on those few—the ones that, if wrong, would materially impact the bottom line.” 📊 He distinguished between: A data system, where all data is treated as equally important, and An information system, where data is prioritized based on the sensitivity of decisions to inaccuracies. In today’s world of AI agents and autonomous decision-making, this insight couldn’t be more relevant. We now have systems that: Depend on accurate data to make real-time decisions Learn from patterns in data to make recommendations Even act on our behalf in trading, hiring, inventory, or medical diagnostics. ⚠️ But here’s the problem: Some data (like demand forecasts) can never be 100% accurate. Some data is just too costly or impractical to measure precisely in real-time. And yet—decisions must still be made. 💡 So what do we do? We must ask: Of all the MANY data points we could measure, which FEW truly matter for making good decisions? And this is where AI and Digital Twins become powerful allies. ✅ Use AI + Digital Twins to simulate decisions and test sensitivity: Which variables cause wild swings in outcomes if inaccurate? Which don’t materially change anything—even if way off? Which data must be real-time (fractions of a second)? Which is “good enough” at weekly or monthly granularity? 🔁 Example: In short-cycle trading, being a few milliseconds more accurate than your competitors yields massive gains. But in long-term investment strategy, average weekly prices may be more than enough. We apply this principle in Digital Twins of supply chains and capital projects: 👉 Before obsessing over collecting every possible data point, we let the simulation and AI show us which variables drive the biggest differences in outcomes. 🎯 The result? Better decisions, lower cost of data acquisition, and systems that stay robust—even when not all data is perfect. This is Goldratt’s wisdom in action. ✅ Actionable Tip: Next time you’re building an AI model, dashboard, or process automation, ask not: “How accurate is the data?” but rather: “If this data is wrong, will it matter to the outcome?” How are you deciding what data points need to be accurate and which not? Would love to hear your views and questions on this... #Goldratt #TheoryOfConstraints #impossibleunless #AI #DigitalTwins #DecisionMaking #DataScience #MRP #SupplyChain #BusinessIntelligence #Simulation #Focus #InformationSystems #AgentryAI

𝟰𝟯% 𝗼𝗳 𝗔𝗜 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀 𝗳𝗮𝗶𝗹 𝗯𝗲𝗰𝗮𝘂𝘀𝗲 𝗼𝗳 𝗱𝗮𝘁𝗮 𝗾𝘂𝗮𝗹𝗶𝘁𝘆 Yet most organizations spend 80% on models and 20% on data. Your AI is only as smart as your data is clean. The pattern repeats across industries 👇 📊 𝗧𝗵𝗲 𝗗𝗮𝘁𝗮 𝗤𝘂𝗮𝗹𝗶𝘁𝘆 𝗖𝗿𝗶𝘀𝗶𝘀 Informatica's 2025 CDO survey found: ➜ 43% cite data quality as #1 obstacle to AI success ➜ 57% report data is NOT AI-ready ➜ Only 5% of organizations have comprehensive data governance 📉 𝗪𝗵𝗮𝘁 𝗕𝗮𝗱 𝗗𝗮𝘁𝗮 𝗟𝗼𝗼𝗸𝘀 𝗟𝗶𝗸𝗲 The data exists but: → Lives in 47 different systems with no integration → Uses inconsistent formats and definitions → Contains unknown biases that propagate through AI → Lacks lineage—nobody knows where it came from → Has quality issues discovered only after deployment Gartner predicts 30% of GenAI projects abandoned by end of 2025 due to poor data quality. 𝗧𝗵𝗲 𝗗𝗮𝘁𝗮 𝗘𝘅𝗰𝗲𝗹𝗹𝗲𝗻𝗰𝗲 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 Organizations achieving production AI allocate 50-70% of timeline and budget to data readiness. Here's what they build: 1. 𝗖𝗼𝗺𝗽𝗿𝗲𝗵𝗲𝗻𝘀𝗶𝘃𝗲 𝗔𝘀𝘀𝗲𝘀𝘀𝗺𝗲𝗻𝘁 Completeness: Do you have sufficient volume? Accuracy: Is the data correct? Consistency: Do definitions match across systems? Timeliness: Is data current enough for decisions? Validity: Does data conform to business rules? 2. 𝗟𝗶𝗻𝗲𝗮𝗴𝗲 & 𝗣𝗿𝗼𝘃𝗲𝗻𝗮𝗻𝗰𝗲 For every data point: Where did it originate? How was it transformed? What systems touched it? When was it last validated? You can't trust AI you can't trace. 3. 𝗕𝗶𝗮𝘀 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 & 𝗠𝗶𝘁𝗶𝗴𝗮𝘁𝗶𝗼𝗻 identify: Sample bias (unrepresentative training data) Historical bias (past discrimination baked in) Measurement bias (flawed data collection) Aggregation bias (combining incompatible data) Then engineer mitigation before deployment. 4. 𝗔𝗜 𝗚𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲 requires: Model-specific data requirements documentation Continuous data quality monitoring Automated drift detection Regular revalidation cycles 5. 𝗗𝗮𝘁𝗮 𝗣𝗿𝗲𝗽𝗮𝗿𝗮𝘁𝗶𝗼𝗻 𝗜𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 Build platforms that enable: Extraction from source systems Normalization and transformation Quality dashboards with real-time monitoring Retention controls meeting compliance requirements API access for AI consumption Data readiness is NEVER "complete." It's continuous discipline requiring dedicated ownership. The Data Excellence Test: Ask yourself these questions: ✓ Can you trace any data point from source to consumption? ✓ Can you explain its quality metrics and bias profile? ✓ Do you have automated systems detecting data drift? ✓ Can you demonstrate data governance to regulators? ✓ Do you spend more on data infrastructure than AI models? If you answered "no" to any of these, you're building on quicksand. ♻️ Repost if you've seen AI fail due to data problems ➕ Follow for Pillar 4 tomorrow: Governance & Risk 💭 What percentage of your AI budget goes to data readiness?

AI accuracy is becoming the wrong metric. Even if your GenAI model is right 95% of the time, the business risk lives in the 5% no one can detect, trace, or explain. That is not just an accuracy problem. It is an operating model problem. Here is the part most AI roadmaps skip: 1/ The wrong 5% is not evenly distributed → It clusters around edge cases, ambiguity, stale data, and novel inputs → Those are often your highest-risk, least-standardized decisions → The model may sound confident even where precedent is thin 2/ You cannot always see which 5% is wrong → A fluent wrong answer can look identical to a fluent right one → Better models often make mistakes harder to spot, not easier → Reviewers are more likely to trust polished output 3/ The cost is asymmetric → The right 95% may create efficiency → One wrong output in a contract, customer promise, pricing recommendation, or market-facing message can erase the gain → Value compounds slowly; liability can arrive all at once 4/ Do not only review outputs. Supervise patterns. → Humans in the loop check individual answers → Humans on the loop watch error patterns, drift, and exception volume → The job shifts from approving every answer to knowing when the system is sliding 5/ Build for being wrong → Define confidence thresholds and escalation rules → Route exceptions to humans with the full context attached → Log AI-generated recommendations, approvals, and actions 6/ The 5% is becoming a governance question → Regulators are moving toward documented oversight → “The model decided” will not be a sufficient explanation → Oversight you cannot document is oversight you cannot prove 7/ The 5% is your moat → Everyone has access to increasingly similar models → Differentiation lives in how well you handle exceptions → Your escalation paths, data quality, review rules, and audit trails are a competitive edge The teams winning with AI are not just chasing a more accurate model. They are accepting that the model will be wrong and building an operating system for exactly that. The real question for your next AI review is not: “How accurate is it?” It is: “What happens when it is confidently wrong, and who notices?” If you cannot answer that, you do not have an AI strategy. You have an AI bet. Need help with your AI strategy or training plan? Book a 45-minute strategy call: https://lnkd.in/gEY5pN7z Save for future reference.

🔍 Using AI? Here’s Why You Must Understand Your Data Source AI is a game-changer, but with great power comes great responsibility—especially when it comes to data. Many AI tools deliver incredible results, but if you don’t know where your data is sourced from, you’re setting yourself up for potential trouble. Here’s why: 🛡️ Data Integrity Matters: AI is only as good as the data it’s trained on. If the source data is biased, outdated, or incorrect, the output could mislead your decision-making. 🔒 Protect Your Intellectual Property: Some AI tools use open-source models or datasets. If you’re feeding sensitive, proprietary information into these tools without understanding how it’s used, you might inadvertently expose your intellectual property. 🏛️ Compliance Is Critical: Industries like finance, healthcare, and law require strict adherence to data privacy regulations. Using AI without knowing the data lineage can lead to non-compliance, fines, or worse. How to Protect Yourself and Maximize AI’s Potential: 1️⃣ Ask Questions: Before using an AI tool, ask how it sources, stores, and processes data. Transparency is key. 2️⃣ Use Closed Systems for Proprietary Data: When dealing with sensitive information, consider using AI solutions that allow for closed-loop systems to keep your data secure. 3️⃣ Validate the Output: Don’t rely solely on AI-generated insights. Cross-check results with trusted sources to ensure accuracy. 4️⃣ Train Your Team: Ensure your team understands the risks and best practices for using AI tools responsibly. AI is a fantastic tool, but it’s not a “set it and forget it” solution. Success requires thoughtful implementation, informed decisions, and a clear understanding of the technology. 💬 What’s your approach to ensuring AI outputs are reliable and compliant? Let’s discuss! #AI #DataIntegrity #DigitalTransformation #ArtificialIntelligence #AICompliance #TechLeadership #BusinessInnovation #AIEthics

64% of executives feel a sense of urgency to adopt generative AI, but even more (75%) don’t trust the data used by those models. It's a classic case of wanting to advance without taking the time to understand the fundamentals. Except this time, the risks — and the consequences of failure — are even higher. Bad data is nothing new. It’s been an unpredictable, messy problem. But an AI model that uses that bad data for training, fine-tuning, prompting, or retrieval augmented generation? That’s a catastrophe waiting to happen. An AI model using faulty data doesn’t just mirror errors — it builds on them. One mistake amplifies the next, and the next, until the output is so far from what you intended that it’s impossible to correct. Until now, data teams have kept the potential AI disasters at bay. They’ve been advocating for improving the data because they understand that AI’s success hinges entirely on the quality of data it learns from. And they’ve been in the trenches day after day, cleaning, verifying, and preparing that data so it's as accurate and reliable as possible before it ever gets to AI. But now, the rush to adopt GenAI is drowning out our data teams' warnings. And it’s creating a disconnect between what we're trying to do and what we need to do. We need to bridge this gap before we’re in too deep. Our enthusiasm for AI needs to match our dedication to data quality — and that means our data teams need a seat at the table. #dataengineering #dataquality #genai #ai

The Spanish Data Protection Agency (AEPD) has published a technical note on accuracy, suitability and data quality when AI is used to process personal data. Its central point is that data quality and the GDPR accuracy principle are closely connected, but not interchangeable. Data quality is broader, applies to personal and non-personal data, and may reflect technical, contractual or organisational requirements beyond data protection law. Figure 9 on page 13 is recommended as it shows data quality aspects to consider when deploying one or more AI systems/models in a processing operation. Accuracy under the GDPR should be assessed in relation to the purpose of the processing. Veracity, currentness, precision, granularity and update frequency are not absolute requirements. They matter to the extent necessary for the intended purpose, particularly where outcomes may affect individuals. The note also says that relevant non-personal inputs and AI-generated outputs need appropriate quality controls. This moves the discussion beyond “clean training data”. The relevant governance object is the full processing operation: inputs, transformations, model or system behaviour, safeguards, human review and outputs. High aggregate model performance may still conceal unacceptable errors affecting particular individuals. Organisations should define purpose-specific quality thresholds and output metrics, document data provenance and transformations, test relevant subgroups, monitor drift, and connect AI governance with privacy accountability. Excessive quality requirements should also be challenged where they lead to unnecessary collection or conflict with data minimisation.