After twelve years building business intelligence systems for enterprise companies, I started asking a different question. Not "how do we visualize the data better?" but "how do we remove the need for humans to stare at dashboards at all?"
The dashboard trap
Every company I have worked with has the same architecture. Data flows from source systems into a warehouse. From the warehouse into curated datasets. From datasets into dashboards. From dashboards into meetings. From meetings into decisions. Eventually.
The latency is days. Sometimes weeks. The dashboards answer yesterday's questions with last week's data. By the time a human sees the number, acts on it, and the action flows back through the system, the opportunity is often gone.
This is not a visualization problem. It is an architecture of attention problem.
Dashboards assume the right human will look at the right number at the right time and make the right decision. That assumption fails constantly. The right human is in a different meeting. The right number is buried on page three of a report they stopped reading months ago. The right time was Tuesday at 3pm, and nobody noticed until Friday.
The agent shift
In 2024 I started experimenting with a different model. Instead of building dashboards that wait for human attention, I started building agents that act when conditions are met.
The first prototype was simple: a revenue operations agent that monitored pipeline health in real time. When it detected an anomaly, a pattern that historically preceded a miss, it did not send a dashboard link. It drafted an email to the relevant rep with context and suggested next steps, and offered to book a coaching conversation with their manager.
The rep received a helpful nudge, not surveillance. The manager received a summary, not a report. The system did not wait for anyone to notice. It acted.
What changed
The technology stack that makes this possible now exists. Large language models can interpret unstructured data. MCP can connect models to enterprise systems in a structured, governed way. Agentic frameworks can manage multi-step workflows with memory and context.
What is still missing, in most organizations, is the data architecture underneath it.
Agents need more than access. They need context. They need to know what normal looks like. They need causal relationships, not just correlations. They need to know which actions have been tried before and what happened.
Building that context layer is the next generation of data leadership. It is not about bigger warehouses or faster queries. It is about encoding organizational knowledge into systems that can act on it.
The resistance you will face
When I started talking about this shift, the resistance was immediate and predictable.
BI teams felt threatened: if an agent writes its own SQL, what is the BI developer for? Managers felt uncomfortable: if a system can identify problems and suggest actions, what is the manager's role? IT felt overwhelmed: agents that write and execute code sound like a security nightmare.
All of these concerns are valid. All of them miss the point.
The BI developer's job was never to write SQL. It was to translate business questions into data answers. The agent elevates that role: the developer becomes a curator of organizational context and a designer of decision systems rather than a builder of reports.
The manager's job was never to read dashboards. It was to develop people, remove obstacles, and create the conditions for success. The agent gives managers back the hours they were spending staring at numbers.
And the security concern is real, which is why governance architecture for agentic systems will be one of the most important fields in enterprise technology over the next five years.
Where I am now
I no longer build dashboards. I build decision systems. A decision system has three layers:
- Signal layer. What is happening right now, captured from every relevant source, structured and unstructured.
- Context layer. What it means, given our history, our strategy, our constraints.
- Action layer. What should happen, who should know, what they should do.
The dashboard sits awkwardly between the first two layers. It shows signals without context and waits for human translation. An agentic system closes the gap. It moves from signal to action with minimal human latency. It does not replace judgment. It accelerates it.
The question for data leaders
If you run a BI or data team, the question is not "how do we make our dashboards better?"
It is "how do we make our dashboards unnecessary?"
Not by eliminating transparency, but by encoding it into systems that act on what they see. Systems that know what matters, know what to do, and learn from what happens next.
That is the future of business intelligence. It is closer than most organizations think. If you want to see it running in a browser instead of reading about it, try the demos.