At a large Midwestern CPG client, a power user from R&D opened a meeting, not with slides or wireframes, but with a fully functional, browser-based application he had built himself (complete with calculations, data manipulation, visual design and interactivity) before bringing it to the analytics team to help scale. This kind of “prove-then-partner” behavior is exactly what effective business engagement looks like, and it’s no surprise it emerged from R&D, where experimentation often leads. In this case, the engineer had essentially vibe-coded an automated dashboard, offering a clear signal of where things are heading. That signal is reinforcing across the market: over several weeks of SAP Business Data Cloud Exploration sessions across the U.S., I heard similar stories in four of six locations I visited. Business users arriving with self-built prototypes is a practice spreading across industries, geographies and company sizes, and the behavior is only accelerating.
The agentic shadow analytics era
We’ve walked into the agentic shadow analytics era. Shadow analytics isn’t new. Excel and Access have played that role for 30 years. What’s new is that an LLM coding assistant can carry a non-engineer across the parts of pipeline construction, modeling, ML enrichment and front-end build that used to act as natural gates. The output is a prototype that looks and behaves like software.
The honest observation from twenty years in this field: analytics isn’t really that hard, until it is. What makes it hard? Volumes of data, performance, security, semantics and definitions, uncontrolled business processes, multiple business applications converging, regional variation, multi-language support and the physics of moving and storing data across cloud providers and zones. Notice what isn’t on that list: high-action signals, content that drives a decision, metrics tied to the process they describe and storytelling that gets the room to act. Those parts mattered before. They still matter now. But, they were never the part technology was solving.
Here’s what dissolving the technical barrier looks like up close. About a year ago, I worked with a client finishing an Ivy League MBA capstone in business analytics. The eight-week assignment (SQL modeling, machine learning enrichment and a Looker dashboard) took us an afternoon. I phoned a friend twice for SQL syntax, but not once for ML or visualization, using tools I hadn’t touched before. When you know the outcome you’re aiming for and experimentation isn’t constrained by compute, this work isn’t hard, and it left me impressed with what new analytics practitioners can do straight out of school.
I saw this coming. Around 2019, watching the first wave of RPA tools open to non-technical users, I made a bet that citizen developers, the business power-users who showcase development acumen, would build their own pipelines without waiting for a sponsor. In 2021, I hired the first end-user I saw doing it. Five years later, that hire is one of the strongest people on my team, not because of a head start on the tools, but because they kept the muscle that matters: signal selection, meaningful metrics and storytelling. The tools moved. The fundamentals did not.
What should an analytics leader do now?
- Lean into the citizen developer wave. Business users are going to build pipelines and dashboards; the question is whether they build them inside or outside a sanctioned environment. The CPG R&D engineer did it the right way by collaborating with the analytics team. Not every citizen developer will. That is the failure mode to plan for. Consider how to cultivate the culture and offer the tools to absorb the momentum as opposed to allowing it to circumvent outside of the program. Reward collaborative behavior publicly, so it becomes the default.
- Define the gold path and put notebooks on it. A gold path is the repeatable, documented, agreed-upon way to do common data and analytic work. Today, that path includes Databricks, Snowflake and Fabric notebooks alongside dashboard tools. For analytics leaders coming from an SAP-centric history, this is a bigger shift than it appears. The standard operating procedure in that ecosystem was Analysis for Office, BEx queries and manual export-and-stitch workflows. Notebooks were a separate world that lived elsewhere, used by other people. Notebooks used to be the realm of the Data Scientist and Engineer, before LLM support came along with easy collaboration. It is a better way to work, but more importantly, it leaves a governed artifact describing user intent and the steps to achieve it.
- Invest in the cleanup before the AI era forces you to. Catalogs, semantic layers, data quality and lineage are not glamorous projects but are the difference between an AI agent that returns the right answer and one that returns a confident wrong one. Genie makes this concrete: its accuracy depends almost entirely on the quality of the catalog and semantic definitions on which it sits. Teams cleaning these up now are the ones whose AI investments will return something real in the very near future. A program’s value proposition is the governed version of the truth. The shadow IT era carries the risk of propagating shadow analytics beyond anything seen before.
- Coach the durable skills. A team’s value is in knowing which question to ask, which metric drives the decision and how to tell the story that gets the room to move. Those skills do not come from a Claude or Copilot subscription. They come from time in the business and from feedback delivered by leaders who care about the answer.
Pipelines, models and dashboards are about to multiply across every function, and most of them will not be built by an analytics team. That is the opportunity. If the gold path is clear, the data underneath is clean and the builders know what a high-action signal looks like when they see one.
Business leaders may not see the full picture from where they sit, and that is fine. Share this with power users and architects and challenge them to map the gold path for the notebook and agent era.
For more information on our agentic, Microsoft and SAP consulting services, contact us.
