Dashboards aren't the future of BI. Your data foundation is.

For years, the promise of business intelligence (BI) was self service: Give business users dashboards, reduce their reliance on IT, and put more data in the hands of the people making decisions.

It worked. But it also created a new problem.

Enterprise dashboard sprawl took over. Most enterprises now juggle dozens, sometimes hundreds, of dashboards and static reports built to answer a particular isolated point-in-time question. When the business needs a new answer, another report gets added to the pile. And as organizations introduce generative AI and agentic workflows, simply recreating legacy dashboards with newer technology misses the bigger transformation opportunity.

In the latest episode of Unlocking Gemini, TTEC Digital’s Caleb Johnson sat down with Sandra Parhami, Looker Specialist at Google Cloud. They explored how BI is evolving from static reporting to conversational, AI-powered experiences — and why unified business logic matters far more than the dashboard itself.

From self-service BI to agentic BI

Business intelligence has already gone through several major shifts.

The first moved organizations away from relying exclusively on technical teams to retrieve data and build reports. The next generation of BI tools made self-service analytics increasingly accessible to business users.

Now, AI is driving another transition. "We are seeing the biggest shift we have probably ever seen in BI,” says Parhami.

Organizations are beginning to ask how AI can proactively surface insights, answer questions in natural language, and eventually support autonomous workflows.

This requires a different way of thinking about BI.

As Johnson notes, many organizations approaching data estate modernization still start with the same request: “We have 50 dashboards today. Give us a new platform with those same 50 dashboards.” But recreating the old environment on new technology leaves much of AI’s potential untouched.

Instead of requiring a new dashboard every time someone asks a new question, emerging BI experiences allow users to interact directly with their data through conversational interfaces and generate insights on demand.

The real cost of another dashboard

Every new dashboard also carries what Parhami describes as a kind of “tax.”  

Technical teams must scope the request, locate the raw data, define the appropriate metrics, build the report, and commit to maintaining it. Multiplied across departments, self-service BI quickly devolves into an IT bottleneck.

AI offers a way to break that cycle, but only if the underlying data and infrastructure can be trusted. As organizations scale the number of questions their data systems need to answer, handling each one as a separate reporting project becomes increasingly difficult. Most companies aren't planning to scale their data teams at the same rate as demand for insights. The goal becomes to create a reusable foundation capable of answering questions the organization hasn't asked yet.

In the AI era, the semantic layer becomes the star

Dashboards aren't disappearing. Visualizations remain an important way for people to understand information. But the role of the BI platform is expanding.

Employees will increasingly query business data directly within AI environments, like Gemini Enterprise, rather than navigating traditional dashboards. For these conversational models to provide accurate outputs, they must ground themselves in a centralized semantic layer. Enter Looker's semantic layer. By leveraging LookML, organizations can codify business logic centrally, creating consistent definitions for metrics and rules across the enterprise. Instead of sales defining “revenue” one way and marketing another, the organization can establish a single, immutable source of truth that downstream applications consume consistently.

“What Looker does really well, our bread and butter, [is] that proprietary modeling layer language, LookML,” Parhami explains. “It's aged really nicely in the time of AI.”

While it’s not a new capability, AI makes its value clearer. Consistent definitions and permissions are now a foundational AI requirement.

AI adoption depends on data trust

According to Parhami, trust is a huge factor influencing AI adoption. “I think everyone has data, but what people are really lacking is trust in the data.”  

When definitions are standardized through the semantic layer, employees can interact with AI knowing the information they receive is grounded in governed business logic.

The semantic layer therefore becomes more than an analytics architecture decision. It's part of creating an environment where people are willing to use AI in their everyday work.

Start with a data dictionary, not another dashboard

Of course, establishing a solid data foundation can be difficult.  

Years of dashboards, data silos, and technology decisions often leave business logic scattered across systems. Definitions may live inside individual reports, warehouses, or even individual teams.  

Parhami's recommendation is to start by separating that logic from the dashboard itself.  

Doing so requires an upfront investment, but it changes the economics of everything that follows. Instead of repeatedly rebuilding business logic for individual reports, organizations create reusable building blocks that can support dashboards, conversational analytics, and AI workflows alike.

From there, organizations can adopt what she calls a “data dictionary mindset,” consolidating definitions and establishing a trusted source of truth.

“The second you make that step forward, you enter a brand-new world where all of your efforts are exponentially compounding on each other,” Parhami says.

Agentic BI starts with data that is defined, governed and trusted well enough for AI to act on it. That’s where the semantic layer matters: it gives teams a shared foundation for faster decisions, smarter workflows and BI experiences that can keep pace with the questions the business hasn’t even thought to ask yet.

[cta-1]