This edition starts with a question many SaaS teams are quietly asking: if AI can generate code so quickly, can it also generate the embedded analytics infrastructure your customers will rely on?

Vibe-coding may make it easier to generate dashboards, charts, and analytics prototypes, but embedded analytics is not just a collection of visual components. It is an infrastructure challenge involving multi-tenancy, data pipelines, security, self-service, governance, personalization, and long-term scalability.

That distinction matters now because AI is making software creation feel faster than ever. But speed does not remove architecture. It only makes architectural considerations easier to overlook.

The tension between SaaS and AI remains at its peak. Many software developers now believe they can build almost any application using AI, open-source libraries, and vibe-coding tools — regardless of complexity.

Later this year, we may hear many success stories… or we may learn otherwise.

At one extreme, SaaS companies could start losing subscribers as businesses leverage AI to build their own custom applications, while vibe-coding platforms emerge as the next generation of unicorns.

On the other side of the spectrum: maybe not so fast.

As far as building analytics capabilities is concerned, I have observed again and again that even the brightest software engineers overlook some caveats:

Coding a dashboard is simple and quick. However, creating a canvas for power users to securely build their own rich dashboards in a multi-tenant environment that offers self-service and personalization capabilities for everyday users, in a managed style, is a different ballgame.

Additionally, the visual outputs, such as tables and charts, are the tip of the iceberg. The visual analytics performs well on top of a solid data foundation, where all the data pipelining resides. Should you build the data pipeline in a code-centric, hard-coded manner, or as something that power users can manage in a no-code style?

What you read in italic in the last three paragraphs comes from an email from 20 years ago! I could write the same today, use “vibe-coding” instead of “coding,” and add something like “adopting Agentic AI that amplifies the need to govern data.” However, the core message still remains true:

If you are a SaaS company and your core focus is not analytics, don’t try to code or vibe-code an embedded analytics platform using some chart libraries. Instead, choose a complete, modern analytics platform that provides 90% of what your product needs and lets you vibe-code the remaining 10% to extend it.

Building software is one thing. Building scalable, secure, reliable, and continuously evolving software platforms is something else entirely.

The architecture of an embedded analytics platform plays a critical role in determining how well it can support current and future AI capabilities. While surface-level features may evolve quickly, the underlying platform defines what is possible, how easily new capabilities can be integrated, and how reliably they can scale. For SaaS companies, understanding architectural fundamentals is essential to making durable platform choices.

Not all features are created equal. Some are built on flexible, extensible foundations; others are constrained by hardcoded logic or legacy design. It’s important to look beyond what a platform can do today and ask how it does it. For example, does “talk to your data” rely on a brittle prompt template, or is it backed by a modular, model-agnostic architecture? The answer reveals how easily that feature can evolve.

Platform limitations often surface gradually. A platform that lacks multi-tenant context may struggle to support ad hoc queries like “make me a dashboard” in a SaaS environment. One that doesn’t support API callbacks may block future agent-based automation. These constraints may not be obvious during a demo, but they can become blockers as AI expectations grow.

It is difficult to directly evaluate platform capabilities. It is much easier to evaluate key features that are dependent on a solid architectural foundation. Use the availability of these features as a strong indication of platform capability or weakness. 

Finally, it’s important to assess platform risk holistically. This includes not only technical limitations but also roadmap transparency, ecosystem maturity, and alignment with your product strategy. A platform that looks capable today but lacks architectural depth may become a liability tomorrow.

The logic makes sense. If customers can get the answers they need inside your product, they have fewer reasons to leave it. They do not need to export data, open a separate BI tool, or upload company information into an outside LLM just to answer questions your product should already help them answer.

But simply adding dashboards does not create retention.

To actually drive retention from analytics, you need to embed insight into the customer’s workflow. Analytics should appear at the moments when decisions are made, questions come up, and actions need to happen.

Retention starts with frequent use. Analytics that customers only check occasionally cannot become essential to daily work. When insight is already in view, or one step away, it stops being something users have to remember to check and becomes something they naturally rely on.

This is where AI makes the opportunity even stronger. Conversational AI makes analytics easier and more accessible because users can ask questions about their data directly inside the workflow. Instead of leaving the product to run a report somewhere else, they can get answers where the work is already happening.

But access alone is not enough. The analytics also has to reflect how each customer actually works.

A real estate brokerage focused on luxury residential sales may care about buyer liquidity, deal certainty, and time-to-close. A commercial leasing firm may care more about renewal rates, vacancy duration, and tenant stability.

Both companies may use the same software, but they do not define success the same way.

If your analytics cannot adapt to those differences, customers will look elsewhere for answers. They will create spreadsheets, use outside tools, or build their own reporting habits around your product instead of inside it. Once that happens, your product becomes easier to replace.

Retention is not a by-product of having analytics. It is the result of designing analytics to be indispensable.

The goal is not simply to add analytics. The goal is to make your product the place customers go to understand what is happening, ask better questions, decide what to do next, and take action without leaving their workflow. That is what turns analytics from a feature into a retention driver.

SaaS teams should be cautious about equating faster code generation with durable platform creation. A dashboard can be built quickly, but a secure, scalable, multi-tenant analytics experience that customers can personalize and rely on over time is much harder to build.

The real risk facing product teams: Using AI to recreate only the visible layer of analytics while underestimating the foundation underneath it. For most SaaS companies, the smarter path is not to vibe-code an embedded analytics platform from scratch. It is to choose a strong analytics foundation, then use AI and custom development to extend the parts that truly make the product different.

See you in the next edition,

Arman Eshraghi
The Embedded Intelligence Brief
Connect on LinkedIn