Episodes

Your pipeline isn’t slow because R forgot how to math — it’s slow because compute is in the wrong place and data is moving too much. This episode shows how switching from local compute (R on your laptop) to SQL compute (R runs inside SQL Server, next to the data) flips hours into minutes. Then we t…

Most “works-on-my-machine” bugs come from drift: different CLIs, runtimes, and extensions. Dev Containers kill drift by turning your environment into code: a devcontainer.json that pins tools, extensions, mounts, and security posture. Use Templates to standardize, Features to add tools, pre-builds …

Flying Business Central without telemetry is flying IFR with the instruments off. Turn it on, push signals into Azure Application Insights, and read them with Microsoft’s Power BI apps to catch spikes, deadlocks, and extension issues before users file tickets. The gotcha that burns most admins? App…

The Copilot buttons you see in Business Central are just the trailer. The movie is in System.AI, where you can register custom Copilot capabilities, control where/when they appear, and wire metaprompts so Copilot speaks in your company’s voice. Ship safely by piloting in a sandbox, guarding keys (I…

Most bots are parrots: fluent, ungrounded, risky. RAG (search + LLM) fixes that by answering from your tenant with your permissions. You’ve got two Microsoft lanes: Copilot Studio (low-code, blazing fast to value) and Azure AI Foundry (code-first, full control, governance). Use a simple lifecycle—E…

A dataflow that “worked today” can be a ticking time bomb tomorrow. Schema drift, hardcoded values, and brute-force reloads turn quick wins into fragile, costly pipelines. The fix is simple discipline: modular M, parameterization, Delta Lake with schema control, and partitioning + incremental refre…

Is Fabric’s Digital Twin Builder a chaos-tamer or just another layer to babysit? We break down what the Digital Twin Builder (in Fabric Real-Time Intelligence) actually does: a semantic canvas for modeling, an ontology to keep definitions straight, low-code mapping of messy sources into OneLake, an…

The Azure AI Foundry Trap isn’t the platform—it’s treating it like plug-and-play. Multimodal demos crumble on messy real data; RAG without hybrid (keyword + vector) search and semantic re-ranking hallucinates; and “agents” spiral when you don’t scope identity, actions, or observability. The cure: g…

Autonomous agents stop suggesting and start acting. That’s power—and risk. Give them memory (Cosmos DB), connectors, and broad scopes, and they’ll execute workflows without waiting. The cure is simple but non-negotiable: toggle act vs. suggest, lock scopes with least-privilege RBAC, enforce data cl…

PBIP turns monolithic PBIX files into text you can version—opening the door to Git. But visibility isn’t discipline. Real collaboration comes from PRs as gates, automated checks that catch mistakes before humans review, and deployment pipelines that promote changes DEV → TEST → PROD without finger-…

This episode argues that Syntex being folded into SharePoint Premium is not a naming joke — it’s Microsoft consolidating content processing, content experiences and governance into one platform so Copilot can stop guessing and start delivering real answers. SharePoint Premium = Brain (content e…

This episode lays out the star schema “pro move” for Power BI: center your model on lean fact tables and surround them with flattened dimensions so VertiPaq can filter fast and calculate cleanly. Ditch the one-giant-table “digital spaghetti”—it’s why slicers crawl, totals go weird, and DAX feels ho…

In today’s busy business world, you face many challenges when using copilot technology. Knowing these problems is important for getting the most from your investment. Recent numbers show different acceptance rates in industri...

This rant-to-roadmap episode calls out classic Power BI alerts as clunky, dashboard-bound “sticky notes” and shows why Fabric Data Activator is the grown-up alternative: event-driven rules that trigger real workflows (via Power Automate, tickets, Teams posts, etc.) without littering your tenant wit…

Running BI without ALM is file-roulette. This episode shows how to escape “Final_v9.pbix” chaos by treating Power BI models as code, using Git for real history + branching, and deploying via pipelines so changes move Dev → Test → Prod without late-night rollbacks. Tie it all together with tickets s…

This episode turns SQL from “scary” to useful. You’ll learn why SELECT is a safe, read-only way to ask questions, how the four core clauses (SELECT, FROM, WHERE, ORDER BY) snap together, where T-SQL adds Microsoft-style extras (e.g., TOP, TRY…CATCH), and how to avoid rookie traps (SELECT *, missing…

This episode shows how to turn messy SharePoint lists → clear Power BI dashboards your team actually uses. You’ll connect with the SharePoint Online List connector (the right way), tame ugly fields in Power Query (choice/person/lookups, UTC → local), and publish + embed the report back into SharePo…

Model-driven apps aren’t here to dazzle—they’re here to survive schema change, scale security, and ship to production. In this episode you’ll watch a blank Dataverse table become a themed, role-aware app with custom forms, views, relationships, security roles, and automation. We contrast the “flash…

Dataverse migrations aren’t “move some rows on Friday.” They’re an org redesign: data model, security, automations, integrations, licenses, and cutover. This episode shows how to pick Dataverse for the right reasons, audit what you have, map legacy lists/custom tables to a relational schema, keep f…

Copilot doesn’t fix messy data—it amplifies it. This episode shows how “data goblins” (duplicates, stale archives, bad labels, abandoned folders) turn Copilot into a confident liar, why trust collapses on the first wrong answer, and how to flip the script: micro-scoped pilots, visible wins, and cle…

No single tool “wins” your analytics deployment. GitHub owns truth, Fabric moves packages, Azure DevOps enforces gates. Treat Bronze/Silver/Gold as code, or your Medallion becomes a CSV swamp with 3 a.m. dashboard calls. This episode lays out a practical stack and the three non-negotiables—isolated…

Power Automate approvals shouldn’t be a single “Start and wait” glued to an email. Build them like software: dynamic approvers, restartable stages, clean audit logs, modular child flows, and sane reminders/escalations that don’t spam. This episode shows how to turn fragile, one-off flows into resil…

Leaders think “Copilot = flop” because they’re staring at sign-ins, not behavior. This episode shows how to separate door counts (Entra) from work signals (Copilot/Insights), identify where real prompts happen, and defend ROI with evidence—not vibes. You’ll leave with a 3-step adoption playbook, CF…

Gateways don’t fail mysteriously—they fail predictably across network, credentials, and architecture. This episode turns a “worked in test, died in prod” horror story into a repeatable deployment playbook: lock down outbound/FQDN rules, fix service-account/credential mapping across every source, an…