How Data Integrity Affects AI Performance

I’m obsessed with one truth: 𝗱𝗮𝘁𝗮 𝗾𝘂𝗮𝗹𝗶𝘁𝘆 is AI’s make-or-break. And it's not that simple to get right ⬇️ ⬇️ ⬇️ Gartner estimates an average organization pays $12.9M in annual losses due to low data quality. AI and Data Engineers know the stakes. Bad data wastes time, breaks trust, and kills potential. Thinking through and implementing a Data Quality Framework helps turn chaos into precision. Here’s why it’s non-negotiable and how to design one. 𝗗𝗮𝘁𝗮 𝗤𝘂𝗮𝗹𝗶𝘁𝘆 𝗗𝗿𝗶𝘃𝗲𝘀 𝗔𝗜 AI’s potential hinges on data integrity. Substandard data leads to flawed predictions, biased models, and eroded trust. ⚡️ Inaccurate data undermines AI, like a healthcare model misdiagnosing due to incomplete records.   ⚡️ Engineers lose their time with short-term fixes instead of driving innovation.   ⚡️ Missing or duplicated data fuels bias, damaging credibility and outcomes. 𝗧𝗵𝗲 𝗣𝗼𝘄𝗲𝗿 𝗼𝗳 𝗮 𝗗𝗮𝘁𝗮 𝗤𝘂𝗮𝗹𝗶𝘁𝘆 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 A data quality framework ensures your data is AI-ready by defining standards, enforcing rigor, and sustaining reliability. Without it, you’re risking your money and time. Core dimensions:   💡 𝗖𝗼𝗻𝘀𝗶𝘀𝘁𝗲𝗻𝗰𝘆: Uniform data across systems, like standardized formats.   💡 𝗔𝗰𝗰𝘂𝗿𝗮𝗰𝘆: Data reflecting reality, like verified addresses.   💡 𝗩𝗮𝗹𝗶𝗱𝗶𝘁𝘆: Data adhering to rules, like positive quantities.   💡 𝗖𝗼𝗺𝗽𝗹𝗲𝘁𝗲𝗻𝗲𝘀𝘀: No missing fields, like full transaction records.   💡 𝗧𝗶𝗺𝗲𝗹𝗶𝗻𝗲𝘀𝘀: Current data for real-time applications.   💡 𝗨𝗻𝗶𝗾𝘂𝗲𝗻𝗲𝘀𝘀: No duplicates to distort insights. It's not just a theoretical concept in a vacuum. It's a practical solution you can implement. For example, Databricks Data Quality Framework (link in the comments, kudos to the team Denny Lee Jules Damji Rahul Potharaju), for example, leverages these dimensions, using Delta Live Tables for automated checks (e.g., detecting null values) and Lakehouse Monitoring for real-time metrics. But any robust framework (custom or tool-based) must align with these principles to succeed. 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲, 𝗕𝘂𝘁 𝗛𝘂𝗺𝗮𝗻 𝗢𝘃𝗲𝗿𝘀𝗶𝗴𝗵𝘁 𝗜𝘀 𝗘𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴 Automation accelerates, but human oversight ensures excellence. Tools can flag issues like missing fields or duplicates in real time, saving countless hours. Yet, automation alone isn’t enough—human input and oversight are critical. A framework without human accountability risks blind spots. 𝗛𝗼𝘄 𝘁𝗼 𝗜𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁 𝗮 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 ✅ Set standards, identify key dimensions for your AI (e.g., completeness for analytics). Define rules, like “no null customer IDs.”   ✅ Automate enforcement, embed checks in pipelines using tools.   ✅ Monitor continuously, track metrics like error rates with dashboards. Databricks’ Lakehouse Monitoring is one option, adapt to your stack.   ✅ Lead with oversight, assign a team to review metrics, refine rules, and ensure human judgment. #DataQuality #AI #DataEngineering #AIEngineering

10 of the most-cited datasets contain a substantial number of errors. And yes, that includes datasets like ImageNet, MNIST, CIFAR-10, and QuickDraw which have become the definitive test sets for computer vision models. Some context: A few years ago, 3 MIT graduate students published a study that found that ImageNet had a 5.8% error rate in its labels. QuickDraw had an even higher error rate: 10.1%. Why should we care? 1. We have an inflated sense of the performance of AI models that are testing against these datasets. Even if models achieve high performance on those test sets, there’s a limit to how much those test sets reflect what really matters: performance in real-world situations. 2. AI models trained using these datasets are starting off on the wrong foot. Models are only as good as the data they learn from, and if they’re consistently trained on incorrectly labeled information, then systematic errors can be introduced. 3. Through a combination of 1 and 2, trust in these AI models is vulnerable to being eroded. Stakeholders expect AI systems to perform accurately and dependably. But when the underlying data is flawed and these expectations aren’t met, we start to see a growing mistrust in AI. So, what can we learn from this? If 10 of the most cited datasets contain so many errors, we should assume the same of our own data unless proven otherwise. We need to get serious about fixing — and building trust in — our data, starting with improving our data hygiene. That might mean implementing rigorous validation protocols, standardizing data collection procedures, continuously monitoring for data integrity, or a combination of tactics (depending on your organization’s needs). But if we get it right, we're not just improving our data; we're setting our future AI models to be dependable and accurate. #dataengineering #dataquality #datahygiene #generativeai #ai

You can’t automate what you don’t understand. And #AI can’t optimize what it can’t trust. A lot of organizations are chasing AI in IT operations… But here’s the unspoken truth: AI isn’t failing you. Your data is. Most IT tickets are filled with: - Vague or missing short descriptions - Empty detailed descriptions - Copy-paste resolution notes - Blank or outdated implementation, testing, or blackout plans And yet we expect AI to: - Predict incident resolution - Recommend similar tickets - Cluster top issues - Detect anomalies - Auto-route and auto-resolve It’s like asking a GPS to navigate with broken satellites and incomplete maps. AI learns from your historical data but what if your past is noisy, incomplete, or misleading? Here’s the deal: The quality of your AI is only as good as the quality of your foundational data. That includes: - Description and short descriptions - CI ownership and relationships - Support & approval groups and categories - Resolutions - Implementation plan, backout plan - Accurate historical ticket data etc. Before you buy another AI tool, ask: “Is our data ready for intelligence?” Clean data isn’t a checkbox. It’s the fuel for AI precision, performance, and trust. Want real ROI from AI in IT Operations? Start with a data integrity audit, not a chatbot. #ITSM #AIOps #Data #CMDB #IncidentManagement #Automation #AgenticAI #ServiceNow #DigitalTransformation #ITLeadership

Many companies believe that implementing Artificial Intelligence is the final step to generating value. But they forget that without data governance, AI results lose credibility. Some common symptoms of weak governance include: • Duplicate customer records across different systems. • Critical fields left blank or marked as optional. • Outdated records that linger for months without review. • Conflicting definitions of what counts as “active” or “inactive.” • Parallel spreadsheets becoming the “real source” of information. What happens then? AI suggests misleading paths, misclassifies data, inflates numbers, and explains KPIs that don’t reconcile. Leaders lose confidence, and the technology that should bring clarity only creates more noise. The bottom line: without data governance, there is no trustworthy AI. Governance is not bureaucracy, it is what ensures quality, consistency, and trust so that AI can truly deliver value. Is data governance ready for the age of AI?

🚀 Every enterprise wants AI. But not everyone is ready for it. In most organizations, the biggest barrier to AI success isn’t the model, the vendor, or the cloud platform… It’s the data. Here’s why enterprise data maturity is now the single most important success factor for any AI initiative: 📊 1. AI is only as good as the data feeding it Models don’t create intelligence, they learn it. And if your enterprise data is: * inconsistent * siloed * duplicated * outdated * ungoverned …then even the best AI platforms will deliver noisy, biased, or misleading insights. Clean, connected, trusted data = reliable AI outcomes. 🧩 2. Data Governance is no longer optional AI amplifies whatever it’s trained on, good or bad. Organizations now need: * Clear data ownership * Standardized definitions * Metadata management * Access controls & lineage * Enterprise taxonomies Without governance, AI becomes a liability instead of an accelerator. 🔍 3. Contextual data > raw data AI needs context to interpret enterprise information: * Who owns the data? * What system created it? * How fresh is it? * What business process does it represent? This is where data catalogs, business glossaries, and lineage tools become critical. Context drives intelligence. ⚙️ 4. Integrated data unlocks enterprise-wide AI Siloed data creates siloed AI. To scale AI across the business, organizations need: * Unified data platforms * API-driven integration * A consistent semantic layer * Enterprise Master Data Management (MDM) When systems talk to each other, AI actually becomes predictive and proactive. 🔐 5. Responsible AI starts with responsible data Bias, fairness, privacy, explainability, all of it is rooted in how data is sourced and managed. Good data practices reduce regulatory risk and increase trust in AI systems. 🌐 6. Enterprise data determines AI ROI Companies that invest in: * data quality * data architecture * data engineering * data governance * data observability …see dramatically higher returns from their AI investments. The equation is simple: Strong data foundation → faster AI deployment → higher business value. 🧠 Final Thought AI isn’t magic. It’s math running on data.

As I called it in my predictions post on 1/21, data quality KPIs will replace model accuracy as the #1 AI risk metric. Convos on AI still obsess over model accuracy, but we are getting ahead of ourselves. If you can’t prove your training data wasn't contaminated, then what is the point in even assessing model accuracy? We know it will be wrong! Luckily, enterprises are starting to realize that 80% of AI failures trace back to stale, biased, or semantically broken data. This makes the CDO the de facto AI risk officer. So as scrutiny begins to shift upstream to data lineage, freshness, and bias sources, regulators will follow suit, requesting data lineage proofs over model performance reports. And here’s a good analogy as to why data quality will emerge as the #1 AI risk in 2026: Bad data is like sleeping on a crappy mattress. On the first night, you're just uncomfortable, but years later, you're looking at surgery and wondering why you didn't just replace the mattress. Bad data works the same way. In batch analytics, it's uncomfortable. Dashboards are wrong, reports get delayed, but in production AI, it's injury. Models make biased decisions, regulators start asking questions, and your org takes the hit publicly. No thanks!

LLMs can get "brain rot." And the damage is permanent. New research from UT Austin and Texas A&M tested what happens when you train AI models on low-quality social media content (the digital equivalent of junk food). The results: systematic cognitive decline across reasoning, memory, ethics, and personality. Here's what makes this alarming: Models trained on fragmentary, engagement-bait content from Twitter/X showed: → 24% drop in reasoning with chain-of-thought → 38% decline in long-context understanding → Increased "dark traits" (psychopathy, narcissism) → Higher willingness to follow harmful instructions The primary failure mode: "thought-skipping." Models increasingly truncate or skip reasoning chains entirely. And the damage persists even after massive remediation efforts. Scaling up instruction tuning on clean data improved some metrics but couldn't restore baseline capability. The cognitive decline is baked into the model's representations. This isn't about data volume. It's about data quality as a training-time safety problem. As LLMs scale and ingest ever-larger corpora of web data, what happens when the training diet consists of content optimized for engagement rather than accuracy? In this week's AI-Rx newsletter, I break down: → How low-quality training data causes permanent cognitive decline in LLMs → Why "thought-skipping" explains most AI reasoning failures → What this means for deploying AI in clinical settings → Why data curation is actually a safety issue, not just a performance issue → The dose-response relationship: more junk = worse cognition Subscribe here: https://lnkd.in/gqyhuVYj Drops Saturday at 9 AM EST. This research reframes how we should think about AI reliability in healthcare.

80% of AI failures are data failures. Not an exact number, but you know what I mean. The truth is, if the data underneath is fragmented, inconsistent, or impossible to trust, it doesn’t matter how good your model is - your AI project won’t move forward. That includes things like: --- Different teams defining the same metric differently --- Logic rebuilt for every new tool --- Access rules that are unclear, or quietly bypassed ... So the AI answers confidently, but not consistently. Without structured context -  shared metrics, metadata, lineage, and access policies...AI systems lose reliability fast. Especially in enterprise settings. The latest report from O’Reilly + dbt Labs maps this pattern clearly, with real data and examples from teams at Walmart, Block, and NBIM. 💡It’s a strong reminder: “Garbage in, garbage out” still applies in the age of AI - just in a different format. If you’re building copilots, conversational analytics, or agent workflows, the work doesn’t start with the model. It starts with the data you’re asking AI to reason over. 📘 Full report: https://lnkd.in/g2A3BxNq Shared in partnership with dbt Labs. Worth a read.

𝐄𝐯𝐞𝐫𝐲𝐨𝐧𝐞 𝐜𝐡𝐚𝐬𝐞𝐬 𝐁𝐢𝐠𝐠𝐞𝐫 𝐌𝐨𝐝𝐞𝐥𝐬.   Meanwhile, their AI fails because of Bad Data.  Data Quality beats Model size every time. Here are the 8 dimensions of data quality that actually matter: 1. TIMELINESS What it means: Data is updated, accessible, and available exactly when needed Why it matters: Late data limits timely actions 2. ACCURACY What it means: Data correctly represents real-world values without errors or distortions Why it matters: Prevents incorrect decisions based on wrong information 3. COMPLETENESS What it means: All required data fields are present with no critical information missing Why it matters: Missing data weakens analysis and outcomes 4. CONSISTENCY What it means: Same data values remain identical across systems, reports, and datasets Why it matters: Inconsistencies reduce confidence and usability 5. VALIDITY What it means: Data conforms to defined formats, rules, constraints, and business standards Why it matters: Invalid data cannot be processed or trusted 6. UNIQUENESS What it means: Each real-world entity is recorded once, without duplicate or repeated entries Why it matters: Duplicates distort metrics and reporting 7. INTEGRITY What it means: Relationships between data elements remain accurate, connected, and logically preserved Why it matters: Broken relationships cause system failures 8. RELIABILITY What it means: Data consistently produces dependable results across repeated use and scenarios Why it matters: Reliable data supports long-term decision making THE PRINCIPLE You can have: • GPT-5 on garbage data = garbage outputs • Basic model on quality data = reliable insights Data quality always wins. WHAT TEAMS GET WRONG They assume more data = better AI. It doesn't. They spend months optimizing model parameters while: • Training data has duplicates (Uniqueness) • Production data is stale (Timeliness) • Fields are inconsistently formatted (Validity) • Relationships are broken (Integrity) THE FAILURE PATTERN AI works in demo (curated data) → Fails in production (real data) Why?  Demo data had all 8 dimensions.  Production data has maybe 3. MY RECOMMENDATION Before scaling any AI system, audit: ✓ Timeliness: Is data fresh enough for decisions? ✓ Accuracy: Does it match ground truth? ✓ Completeness: Are critical fields populated? ✓ Consistency: Same values across systems? ✓ Validity: Conforms to schemas and rules? ✓ Uniqueness: No duplicates? ✓ Integrity: Relationships preserved? ✓ Reliability: Consistent results over time? THE STRATEGIC INSIGHT ✓ Scaling AI without data quality is like building skyscrapers on sand. ✓ The taller you build (more complex models), the faster it collapses. ✓ Teams that win invest in data quality infrastructure before model complexity. ✓ Which data quality dimension is breaking your AI systems? ♻️ Repost this to help your network ➕ Follow Sivasankar Natarajan for more insights on Enterprise AI #GenAI #EnterpriseAI #AgenticAI

THE SIX PILLARS OF HIGH-FUNCTIONING AI OPERATIONS Foundation: Your AI Is Only As Good As Your Data Governance “Garbage in, garbage out” isn’t just a saying in AI; it’s a fact. We talk a lot about the model, but the real story is always the data behind it. During my time as Dell’s Chief AI Officer, I saw this every day. The real differentiator was never the model, but the quality and governance of the data beneath it. AI success depends on many elements working together, but the quality of the data ultimately determines the integrity of the outcome.     As AI begins shaping decisions across healthcare, education, finance, and public life, data integrity is no longer a technical issue. It is a matter of trust and responsibility.     Organizations cannot scale AI responsibly unless they understand where their data comes from, how it has been handled, and whether it can be defended when challenged. Without provenance, you lose traceability. Without quality, you lose accuracy. Without governance, you lose trust.     In the enterprise, these gaps don’t cause small problems. They create systemic risk. Decisions made by AI systems are amplified at scale: a biased dataset becomes biased outcomes for millions; an incomplete dataset becomes incomplete insights powering critical business functions.     That is why Data Integrity and Provenance is the foundation of Anchor42’s Six Pillars. Rooted in established industry frameworks such as the NIST Cybersecurity Framework and the NIST AI Risk Management Framework, this pillar is designed to help leaders build AI systems they can stand behind ethically, operationally, and legally.     Tomorrow, we move to Pillar 2: Accuracy and Reliability, where we’ll discuss why even trustworthy data requires disciplined, ongoing oversight to ensure predictable performance over time.     #AI6Pillars #AIGovernance #ResponsibleAI #DataIntegrity #AILeadership #EnterpriseAI CC: Frank Murphy, David Chapman For more information visit www.anchor42.ai.