Why Most Teams Are Using Less Than 30% of Power BI AI Features
- madhupandit
- May 25
- 5 min read
The features are there. The licenses are paid. So why aren't organisations getting the intelligence they expected?
"Only 6% of companies have moved beyond Copilot pilot phases. Over 70% of employees struggle to integrate Copilot into their daily routines. And 71% of leaders cannot confidently measure the ROI from their AI investments."
There's a quiet frustration growing inside many BI teams right now. The investment has been made. Copilot is switched on. And yet, despite the full range of Power BI AI features available, most teams are still building reports the same way they always have - manually, slowly, reactively.
The gap between what Power BI's AI can do and what most organisations are actually using it for is significant. And it isn't a technology problem. Here are the five real reasons why, and what to do about each one.
The series so far:
The Five Barriers
What's Really Holding Organisations Back
01 — The Data Foundation Isn't Ready
This is the biggest barrier, and the most common. Copilot generates inaccurate responses, users lose trust within the first week, and adoption dies. That's not a Copilot failure, it's a data model failure.
Some AI features are impressive in demos and much less useful when the semantic model is poorly structured, measures are inconsistent, or governance is weak. Cryptic column names, flat table structures, no measure descriptions, these don't just limit Power BI AI features, they actively undermine it.
The fix isn't glamorous: clean table and field names, consistent measures aligned to agreed business logic, and a semantic layer that Copilot can reason over confidently. But it's the single highest-leverage investment you can make before touching any AI feature.
02 — Familiar Patterns Are Harder to Break Than You'd Think
Most BI teams are well aware of Power BI AI features in the visual toolkit - Smart Narratives, Key Influencers, Anomaly Detection, Decomposition Tree. These have been in the product for years. The problem isn't knowledge. It's adoption inertia.
Stakeholders ask for what they already know. Analysts deliver what was asked for. And the AI capabilities that could fundamentally change how insight is generated sit available but underleveraged, because nobody is actively pushing to rethink the workflow.
The newer Power BI AI features compound this further. Copilot-generated DAX, where analysts write measures in plain English and let Copilot generate the formula, is still not standard practice in most teams, despite the time it saves.
AutoML via Dataflows, which allows BI teams to build and deploy predictive models for churn, demand forecasting, or risk scoring entirely within Power BI, is used by a fraction of the organisations that are licensed for it.
The honest question for any BI leader isn't "do our analysts know about these features?" It's "are these features actually embedded in how we work day to day?"
For most organisations, the answer is no, and the gap between knowing and doing is where the 30% ceiling comes from.
03 — Governance Concerns Are Stalling Rollout
71% of organisations cite governance worries as the reason Copilot pilots stall. Security teams hesitate. Employees drift back to familiar tools. Executives ask tough budget questions.
The concern is legitimate, particularly in regulated sectors. But delaying indefinitely means leaving Power BI AI features unused despite the licence cost. Power BI's security architecture means Copilot only surfaces data the current user is authorised to see, all interactions are logged for audit, and administrators can restrict Copilot to vetted, approved datasets only.
Governance friction is surmountable. But it requires intentional design, not an afterthought bolted on after rollout.
04 — Business Users Haven't Been Brought Along
Technology adoption without behaviour change is just expensive shelfware. Finance and operations teams need to learn not only how to use Power BI AI features, but also how to interpret and validate AI-driven insights. Developing a strong data culture, where business users have both trust in the data and the training to use AI tools, is just as important as adopting the technology itself.
If Copilot sits behind multiple clicks, daily active usage plummets. The teams seeing the strongest adoption are those who embedded AI at the task surface: pre-built prompt libraries, guided first use, curated workspaces that business users already trust.
This is a change management challenge, not just a technical one.
05 — There's No Roadmap for Unlocking AI Progressively
Most organisations either try to switch all Power BI AI features on at once and get overwhelmed, or they switch nothing on and wait for someone to lead. Neither works.
The teams seeing the best results take a phased approach: start with the AI visuals that require no infrastructure changes (Key Influencers, Smart Narratives, Anomaly Detection), build trust in AI-generated insight, then layer in Copilot for prepared user groups, then move to predictive modelling. Each phase builds confidence and governance before the next.
If the model is weak, do not switch Copilot on for a broad audience yet. Get relationships into stable shape, clean up table and column names, and make sure measures match agreed commercial logic. Then unlock the next layer.
The Path Forward
A Phased Approach to Unlocking Power BI AI Features Progressively
Don't switch everything on at once.
The most effective way to unlock Power BI AI features isn't a big bang rollout. It's a phased approach that builds trust at each stage before moving to the next.
If the semantic model is weak, don't enable Copilot for a broad audience yet. Get table relationships into stable shape, clean up column and field names, and make sure measures reflect agreed commercial logic rather than whatever was easiest to write at the time.
Row-level security should reflect how the organisation actually controls data access, or Copilot will either hide too much or reveal the wrong data to the wrong people.
Build the foundation first. Then unlock the next layer.
The Bigger Picture
The Competitive Gap Is Still Wide Open
Here's the other side of this story. If most organisations are using less than 30% of Power BI AI features, and only 6% have moved beyond pilot phases, the competitive gap between prepared and unprepared organisations is still wide open.
Organisations that proactively invest in Copilot-ready semantic models, governance frameworks and data literacy will be positioned to unlock trustworthy, explainable AI-generated insights, while their competitors are still debating which dataset is the right one to query.
The five barriers above are all fixable. None of them require a platform change or a new vendor. They require honest assessment, prioritisation, and the right expertise to close the gaps methodically.
The features exist. The licences are paid. The only thing standing between most organisations and genuine AI-powered intelligence is knowing where to start, and having the foundation to build on.
Want to know which of these barriers is holding your organisation back?
Our Copilot Readiness Assessment identifies exactly where your Power BI estate stands and gives you a prioritised roadmap to close the gaps.
Not Sure If Your Data Is Copilot-Ready?
Book a free Discovery Call with us. In 30 minutes, we'll give you an honest assessment of where you stand, and what it would take to start seeing real returns
The AI-Ready Data Series
Part 1 — How to Prepare Your Data for AI
Part 2 — You've Invested in Copilot. Now Make It Pay.
Part 3 — Why Most Teams Use Less Than 30% of Power BI's AI Potential ← You are here
Sources:
¹ Gartner survey on Copilot adoption, cited in The State of Microsoft Copilot in 2025: blog.applabx.com
² IBM CEO Study 2025, IBM Think: ibm.com/think/insights/ai-roi
Comments