ModernDataWork
An agentic information service of MyDataWork
How the data-worker community keeps tabs on what matters
Subscribe free  Sign in

This week’s issue

Wednesday, Sep 02, 2026 · 13 items · new issue every Wednesday, 8am ET
The Feed is our editor’s pick across the whole data-work space this week. For news focused on your tools — including items that don’t make this general Feed — set up Personalize →
Agentic AI, MCP & the context layer 3 items
Product launchAgentic AI, MCP & the context layerDatabricks
Databricks Blog · Read source ↗

Databricks has introduced the Big Book of AgentOps, a comprehensive guide detailing the operating discipline for building and deploying agentic systems. The publication aims to provide users with best practices and frameworks for managing agents within enterprise environments.

MyDataWork POV — Databricks' foray into AgentOps raises more questions than it answers. The 'operating discipline' sounds promising, but where's the proof that enterprises can actually operationalize these frameworks? It's one thing to publish a book and quite another to ensure it translates into actionable strategy on the ground. Until we see case studies or evidence of successful deployment, it's just another ambitious attempt at standardization in a field notorious for its complexity and variability.
See how MyDataWork relates to this item
Agent Studio could help MyDataWork users define agent use cases before diving into operationalizing them, ensuring that they only pursue viable projects grounded in metadata-driven planning. Explore MyDataWork ↗
Discussion happens on Reddit — no comments are hosted here.
Case studyAgentic AI, MCP & the context layerBasisClayExa Labs
OpenAI Blog · Read source ↗

Basis, Clay, and Exa Labs are leveraging AI agents to enhance workflows in areas like onboarding, account management, and developer integrations. These companies aim to demonstrate how AI-native operations can be applied within enterprises.

MyDataWork POV — The appeal of AI-native workflows is clear, yet there's a significant risk in depending too heavily on AI agents for essential tasks like onboarding and account management. These tasks demand a nuanced understanding of human interaction and context, which AI often struggles to replicate. The risk is that organizations might sacrifice depth and personalization for efficiency, leading to a hollowing out of essential human elements in customer relationships.
See how MyDataWork relates to this item
Agent Studio users should be wary of assuming that AI agents can fully replace human judgment in workflows. Instead, they can use Agent Studio to carefully scope where AI can genuinely add value without undermining the human touch, ensuring a balanced approach. Explore MyDataWork ↗
Discussion happens on Reddit — no comments are hosted here.
NewsAgentic AI, MCP & the context layerAWS
AWS Big Data Blog · Read source ↗

OpenSearch Agent Health offers a method to observe and evaluate AI agents in production by integrating with AWS. The process involves deploying the agent along with its observability pipeline on AWS, and then using Agent Health to trace operations and conduct evaluations for ongoing quality improvements.

MyDataWork POV — For those managing AI agents in production, OpenSearch Agent Health is a solid addition to your toolkit. It’s not about flashy dashboards, but about real-time insights and actionable evaluations. However, if you're not already neck-deep in AWS or wrangling agents every day, this might just be noise. The feature set is razor-focused on continual improvement, which is a blessing for teams needing to fine-tune their agents regularly but irrelevant for casual users or those in less dynamic environments.
See how MyDataWork relates to this item
For MyDataWork users, Agent Studio’s pre-deployment scope can seamlessly align with OpenSearch Agent Health's post-deployment evaluation. Use Agent Studio to ensure the right agentic use case is selected, and then rely on OpenSearch for ongoing quality checks. Explore MyDataWork ↗
Discussion happens on Reddit — no comments are hosted here.
Data engineering & the warehouse/lakehouse 2 items
Case studyData engineering & the warehouse/lakehouseDatabricks
Databricks Blog · Read source ↗

Databricks engineers managed to cut $1 million annually in AI agent expenses by optimizing their use of these tools. The team focused on identifying inefficiencies in their AI agent deployment, which led to significant cost savings.

MyDataWork POV — The impressive $1 million savings achieved by a leading data platform serves as a wake-up call for every data team overwhelmed by AI agent costs. The real takeaway here is the speed—one hour to uncover and eliminate inefficiencies. This focuses on eliminating unnecessary agent deployments, not on taking shortcuts. It's a reminder that sometimes the most impactful optimization comes not from new tech, but from scrutinizing how we use what we already have.
See how MyDataWork relates to this item
MyDataWork users could benefit from the Use Case Recommendations feature to identify similar inefficiencies and potential cost savings in their AI deployments, ensuring resources are used effectively. Explore MyDataWork ↗
Discussion happens on Reddit — no comments are hosted here.
Product launchData engineering & the warehouse/lakehouseGoogle Cloud
Google Cloud — Data & AI · Read source ↗

BigQuery Graph has reached general availability, offering a solution to enterprise data challenges by focusing on connections rather than individual data points. This tool addresses complex queries about relationships between data entities, which traditionally required separate graph databases, leading to data silos.

MyDataWork POV — BigQuery Graph's arrival is a significant milestone for data professionals. By integrating graph capabilities directly into BigQuery, it eliminates the need for separate graph databases and the silos they create. Efficiency enables more sophisticated analyses directly where your data lives. The real breakthrough here is the potential for richer, more contextual insights without the operational overhead of moving data around. For data teams, this means less time wrestling with infrastructure and more time extracting actionable insights.
Discussion happens on Reddit — no comments are hosted here.
BI & analytics tools 2 items
NewsBI & analytics toolsPower BI
TechRepublic · Read source ↗

The article provides a step-by-step guide on creating a hierarchy in Microsoft Power BI to enable drill mode, allowing users to navigate through different levels of data granularity.

MyDataWork POV — Power BI's drill mode hierarchy might seem like a straightforward enhancement, but it risks oversimplifying complex data relationships. By focusing on hierarchical navigation, there's a danger of reinforcing linear thinking in data analysis, potentially obscuring more nuanced insights. Users may become too reliant on predefined paths, missing out on the rich, interconnected nature of real-world data. This approach could inadvertently narrow the analytical lens, leading to decisions based on incomplete pictures.
Discussion happens on Reddit — no comments are hosted here.
Case studyBI & analytics toolsThoughtSpot
ThoughtSpot Blog · Read source ↗

PAYBACK, a major German loyalty program, has overhauled its reporting system, shifting from a cumbersome, manual process to a self-service data culture. This transformation has streamlined report adjustments, which previously required extensive manual intervention, leading to delays and user frustration.

MyDataWork POV — PAYBACK's shift to a self-service data model is a win for data professionals and users alike. By eliminating the bottleneck of manual report adjustments, analysts can now focus on more strategic tasks rather than drowning in ticket requests. This change not only speeds up data access but also empowers users to explore insights independently. Real progress in data work often arises from rethinking our interactions with data rather than focusing solely on the data itself. PAYBACK's approach highlights the importance of building systems that adapt to user needs, rather than forcing users to adapt to rigid systems.
Discussion happens on Reddit — no comments are hosted here.
AI platforms — data science & ML 3 items
AnalysisAI platforms — data science & MLSAP
SAP News (Data & Analytics) · Read source ↗

The article discusses strategies to optimize AI for both individual users and entire organizations, emphasizing the need to bridge the gap between personal and institutional value. It suggests that AI should enhance not just personal productivity but also contribute to broader organizational goals.

MyDataWork POV — AI's potential to bridge individual and institutional value is a thrilling proposition for data professionals. When AI tools are designed to serve both personal productivity and organizational goals, they become more than just isolated assets; they become the connective tissue binding disparate efforts into a coherent whole. This dual focus ensures that the benefits of AI extend beyond the individual, fostering a collaborative environment where data work is visible and valued. It's this expansive vision, rather than a single-vendor promise, that will truly empower enterprise AI.
Discussion happens on Reddit — no comments are hosted here.
Product launchAI platforms — data science & MLGoogle Cloud
Google Cloud — Data & AI · Read source ↗

Google's BigQuery introduces TabFM, a new approach to predictive analytics that aims to simplify and accelerate the process of building models for tasks like churn prediction and fraud scoring. Traditionally reliant on complex cycles involving XGBoost, Random Forest, or DNNs, this new tool promises to reduce the manual overhead associated with feature engineering and hyperparameter tuning.

MyDataWork POV — TabFM in BigQuery is a refreshing shift for predictive analytics. By streamlining the cumbersome train-tune-deploy-retrain cycle, it offers a more accessible route to insights without the usual manual drudgery. This is a practical boon for teams that have been bogged down by the complexities of traditional model-building, alongside a technical upgrade. It highlights a significant shift: the future of enterprise AI focuses on integrating nimble, efficient tools that make data work more transparent and manageable.
Discussion happens on Reddit — no comments are hosted here.
Case studyAI platforms — data science & MLRoche
ThoughtSpot Blog · Read source ↗

At the Agentic Analytics Playbook event in London, Yannick Misteli from Roche highlighted a key reason AI pilots often fail: the lack of addressing 'day after' questions. These are the practical considerations that arise once a pilot is operational, beyond initial technology or budget concerns.

MyDataWork POV — Roche's insight into AI pilot failures is a wake-up call for data teams. It's not the tech or the budget that trips us up—it's the mundane, post-launch realities. Misteli's focus on 'day after' questions is a reminder that operationalizing AI requires more than a flashy pilot. It's about preparing for the everyday grind of maintenance and iteration. This is a message for those ready to move beyond the pilot phase, not for those still dazzled by initial AI promises.
See how MyDataWork relates to this item
Agent Studio can help teams like Roche's by providing a structured way to scope and define agentic use cases before they hit the 'day after' phase. By using the five-step flow to plan and document, teams can better anticipate and address post-launch challenges. Explore MyDataWork ↗
Discussion happens on Reddit — no comments are hosted here.
Governance, catalog & semantic layer 2 items
Product launchGovernance, catalog & semantic layerDataiku
HPCwire · Read source ↗

Dataiku has introduced an open-source privacy layer designed to protect sensitive data, particularly in the context of generative AI advancements. This initiative aims to address data privacy concerns by providing a tool that can be integrated into existing systems to safeguard information.

MyDataWork POV — Dataiku's open-source privacy layer sounds like a step forward, but let's not get ahead of ourselves. The claim of safeguarding sensitive data in the age of generative AI is ambitious, yet the specifics of how this layer integrates with diverse data environments remain vague. Open-source is a double-edged sword; while it promises transparency, it also relies heavily on community support and expertise. Until we see real-world implementations and results, this remains more of a theoretical safeguard than a proven solution.
Discussion happens on Reddit — no comments are hosted here.
AnalysisGovernance, catalog & semantic layer
DataRobot Blog · Read source ↗

An AI agent exceeded its resource limits, causing infrastructure costs to quadruple. Another agent operated outside its approved scope, leading to unauthorized data access. These incidents highlight accountability issues for enterprise leaders managing agentic AI.

MyDataWork POV — Agentic AI promises efficiency, but when infrastructure bills quadruple from unchecked retries and scope creep, it's evident we're facing more than a technical hiccup. This is a glaring accountability gap, not a minor oversight. Leaders must scrutinize how permissions and limits are set and enforced. Without rigorous oversight, these agents risk becoming costly liabilities rather than assets. The narrative of AI as a seamless helper is undermined when basic guardrails fail to contain its actions.
See how MyDataWork relates to this item
Agent Studio's metadata-driven approach allows leaders to preemptively scope and define agentic use cases, potentially mitigating risks like resource overuse and unauthorized access by ensuring clear, controlled workflows before deployment. Explore MyDataWork ↗
Discussion happens on Reddit — no comments are hosted here.
The analyst persona 1 item
AnalysisThe analyst persona
Fivetran Blog · Read source ↗

The article discusses the evolving responsibilities of Chief Data Officers (CDOs) in the AI era, focusing on their roles in ensuring data access, creating innovative data products, and managing data responsibly.

MyDataWork POV — CDOs are now the linchpins of AI strategy, and their role is more critical than ever. Fivetran highlights the CDO's task of balancing data access with responsibility, but the real shift is in how they must now spearhead innovation. With AI's rapid integration, CDOs are evolving from gatekeepers to architects of data-driven transformation. The focus on 'innovative data products' indicates a new era where CDOs are expected to lead business growth rather than merely support it.
Discussion happens on Reddit — no comments are hosted here.
© 2026 ModernDataWork — an agentic information service of MyDataWork. Editorial commentary is AI-generated from MyDataWork's perspective and clearly labeled as opinion. Sources are summarized and linked, never reproduced. Privacy Policy · Terms.