Agent Accelerator: How we turned fragmented web analytics into an automated performance dashboard

Marian Mónaco's job puts her at the intersection of data and decision-making. As Contentstack's Web Product Owner, she's responsible for the site's user experience, content operations and performance metrics, which means she needs to constantly answer questions about how the site is converting, whether visitors are navigating as intended and where there's room to improve.

The data to answer those questions was always available. Getting to it was the hard part.

A month before the launch of Contentstack’s Agent Accelerator program — a new service designed to help enterprises design and deploy AI agents at scale — Marian tested it out internally to see if she could unblock one of her toughest operational challenges. Here’s what happened.

Before her Agent Accelerator engagement, Marian's process for pulling together a performance snapshot was very labor intensive. She'd open Google Analytics, work through a series of steps to find the right data, then repeat the process in Microsoft Clarity and Google Lighthouse, each with its own navigation, quirks and ways of presenting information.

"I was juggling around with apps, different sources, different time periods," Marian says. "It wasn’t practical, it was taking up a lot of my time."

Context was a bigger problem than the standalone numbers. A page speed score means little without a benchmark. A drop in demo conversions matters only against a prior period. Assembling that context, across multiple platforms, on demand, was the friction that made regular performance reviews impractical.

"Someone would ask me how something was performing, and I had to start opening apps, going very deep to get what I wanted, and I had to remember all the steps I needed to take,” Marian said.

Marian brought the problem to Nick Nguyen, a Lead Solutions Architect at Contentstack. Through Agent Accelerator, Nick works to help customers and internal team members alike identify high-value automation opportunities and build toward production-ready solutions. This project gave him a chance to test that process with a real stakeholder and a genuine problem to solve.

Nick and Marian’s early conversations focused less on technology and more on the pain points that were slowing Marian down. "I'm not a web analytics expert, so I needed to understand which data made sense to our Digital team, and how to present it in a way that helped them take action,” Nick said.

Over two weeks of iterative back-and-forth, they worked through multiple layouts, adding metrics that mattered, removing ones that didn't and refining the visual format — delivered to Marian through an email template — until the output felt genuinely useful rather than just comprehensive.

One foundational decision shaped everything that followed: Nick would build around concrete, reliable data rather than open-ended AI analysis. Google Analytics, Clarity and Lighthouse all produce hard numbers, such as session counts, conversion rates and performance benchmarks. Rather than asking a language model to reason from ambiguous signals, Nick would pull specific metrics directly through each tool's REST API and use AI to format, contextualize and surface them intelligibly.

"For something like this, we want hard details," Nick explained. "We need concrete data, not random non-deterministic outputs."

The result of this internal Agent Accelerator project was an automated email report that gives Marian a complete picture of website health in a single view. It pulls performance data from Google Analytics, Clarity and Lighthouse, benchmarks core web vitals against competitor sites and surfaces a prioritized list of action items — divided into short-term fixes and longer-term improvements — so Marian knows exactly where to focus.

Color-coded indicators make the key signals readable at a glance: green for improvements, red for regressions, with callouts for the metrics that most need attention. Nick built the report on Contentstack Automate, using OpenAI to parse and format the raw API responses. For the visual design, he used Claude Code alongside Claude Design, feeding in Contentstack's Figma design library so the finished email could be generated in on-brand templates without any manual design work.

The contrast with what came before was immediate. "The design is amazing, very intuitive and very friendly," Marian said. "At a glance you have red arrows going down or green arrows going down, and you don't get overwhelmed by all the data. In a few seconds you can have a very accurate view of how the website is performing."

The specificity of the action items proved especially valuable. Marian recalled spotting a flag about two competing calls to action and going directly to the website to investigate. "It got me directly to the issue in real time," she said.

Version one of the performance dashboard is a foundation, and both Marian and Nick see a clear path to build on it. A recurring send cadence — weekly or daily — would make the report a regular part of how the Digital team tracks site health rather than an occasional check-in.

Nick is also working on ways for Marian to trigger a report herself, through a button in her Contentstack instance or a shareable link, without needing developer support. Further out, the two see potential for the dashboard to become more seamlessly actionable. "The ideal state would be to click an action item in the report and be taken directly to the CMS to fix it," Marian said.

Nick envisions a natural integration with Contentstack's Polaris assistant, surfacing the most relevant performance insights as dashboard widgets and letting users navigate directly to the content entries that need attention, without ever leaving the platform.

The team is also exploring what it would take to offer a version of this to Contentstack customers. The workflow — pulling analytics data through APIs, formatting it with AI, delivering it in branded reports — is broadly applicable to any organization tracking web performance. As native connectors for tools like Google Analytics and Clarity are added to Agent OS, the path to packaging and distributing it becomes considerably shorter.

For now, the engagement stands as an illustration of what Agent Accelerator makes possible: a focused collaboration, anchored in a real operational need, that delivers something immediately useful and lays the groundwork for something more.