Practical AI
As AI has evolved, our AI strategy has evolved. I’m sharing our journey, and I’d like to hear yours.
The analytics industry is moving fast. Clients expect sharper insights, tighter timelines, and broader visibility into their data. At Allegro Analytics, we've spent the past year deliberately building AI into how we work so we can deliver on that expectation. Here's an honest look at where we started, where we are, and where we're heading.
Phase 1: Laying the Foundation
We started with a single AI vendor early last year. Rather than naming specific providers, what matters is the approach: most enterprise AI tools are capable of the same core features I'll describe. The initial priority was two-fold.
First, we used AI to optimize our own internal operations, sharpening how our team operates day-to-day. Second, and more importantly for our clients, we focused on the semantic layer. That means working alongside clients to build clearly and consistently defined business metrics and the dimensions that support them. Clean, well-defined data is the prerequisite for any AI feature to work effectively. Without it, AI tools return ambiguous or misleading answers. With it, clients can confidently use their AI licenses to query data and trust what they get back.
We also built the capability to set up, configure, and optimize whatever analytics platform a client uses. That flexibility matters, because no two clients are alike.
Phase 2: Getting Closer to Client Data
Our internal AI progress raised a practical challenge: to move faster and deliver deeper insights for clients, we needed to work directly with their data. Our enterprise AI license covers our own environment, not theirs. So we adjusted. AI access to client data environments is now written into our contracts, with appropriate security and governance guardrails in place.
This shift expanded what's possible in both discovery and delivery. In the discovery phase, we can surface patterns and anomalies faster. In delivery, we can produce broader analysis with more context. The result is better work in less time.
Phase 3: Building What’s Next
This is where potential meets practice. The current phase encompasses three major areas: project enablement, automated translation of analytics content across platforms, and web application development.
Our team is rapidly expanding its AI skill set and building informed opinions about which tools perform best for specific use cases. We're also building a custom prompt library, purpose-built and tested for our work and quality standards. These structured inputs ensure consistency across teams and are designed around analytics workflows and the output standards our clients expect.
The pace of progress has been remarkable. We're accomplishing things today that weren't feasible a year ago, or even six months ago.
A few examples worth highlighting:
Dashboard documentation. AI generates comprehensive documentation for every component of a dashboard file: Business Executive Summary, KPIs, filters and actions, data source details, data dictionary, and calculated fields. What used to take hours now takes minutes, with consistent structure and formatting.
Dashboard prototyping. Early-stage designs can now be produced reflecting client branding, template style, narrative structure, business terminology, and chart type selection. This compresses the time from kickoff to first review significantly.
Administration agent. By connecting our AI environment to internal tools, we've built an agent that handles work like automated task creation and closure in our project management tool, email monitoring with action notifications, document review, and communications drafting. This keeps the team focused on high-value work.
Web applications. To pressure-test what's possible, I ran a personal proof of concept. As a novice boater, I wanted a smarter way to plan days out on the water. I built a simple Boat Day form where I enter a date, location, and time. The app returns an itinerary, a local weather forecast for that time window, a boating conditions assessment, and a customized, interactive packing checklist with progress bar to done. The output can be shared with others via a generated URL. It's a small use case, but it illustrates what's accessible now.
What Hasn’t Changed
AI accelerates delivery, but it doesn't replace business context or judgment. Human expertise remains essential: knowledge of business goals, data structure, narrative vision, and expected outcomes still drives the intelligence of the prompt and iterations. Testing against real expectations, catching context or details the model misses, and knowing when to push back on AI-generated output are skills that matter more as AI becomes more capable.
Phase 4: Stay Tuned
We don't have a complete picture of what Phase 4 looks like yet, and that's exactly the point. The roadmap is evolving in real time based on what we're learning. What we know for certain is that the gap between what's possible and what clients are currently getting is still wide, and closing it is our focus.
If your organization is navigating similar questions around AI adoption, data readiness, or analytics modernization, we'd welcome the conversation. The goal is the same for all of us: better decisions, faster.
