The 90% of SaaS Work Nobody Talks About – And a Platform That Eliminates It

01 /The Concept

Every cloud service developer knows the ratio: building the core functionality is 10% of the work. The remaining 90% goes into the surrounding infrastructure – user dashboards, analytics, account management.

Take an e-commerce platform, for example. Once the platform is built, store owners inevitably want sales stats and analytics with clean charts and tables – and all of that still has to be programmed.

Buster is a platform that lets developers add analytics and reporting modules to their users' dashboards quickly and simply.

What makes it interesting: users can query the data in plain English – just like they would in ChatGPT.

In practice, Buster is an AI assistant that developers drop into their user dashboards, so users can get the stats and analytics they need on demand – without the developer having to build custom dashboards from scratch.

To get started, the developer connects Buster to the databases storing their service's operational data. Buster supports MySQL, Postgres, BigQuery, Snowflake, and other popular stores. The connection takes under 60 seconds.

Next, the developer fine-tunes Buster's AI to translate natural-language user queries into the correct SQL. The AI already understands query logic in general; the developer just needs to teach it the specifics of their data model. This takes less than a full workday.

After that, the developer sets the visual style for query results – tables, charts, graphs. The building blocks are already in Buster; it's a matter of configuring display parameters and adding custom CSS.

The resulting dashboards can be embedded into user accounts with a few lines of code.

Users can then ask a question in plain English, receive a table or chart in response, and pin that block to their personal dashboard for future reference. All blocks update automatically as new data flows in.

Buster claims a developer can ship the first minimal analytics block in 30 minutes – enough to prove the technology works, then invest more time in expanding functionality and polish.

A free tier is available for getting started. The production version starts at $699 per month based on query volume. A premium tier with additional features and SLAs is available by contacting the team directly.

Founded in 2023, Buster entered Y Combinator in December of that year and received the standard $500K. Even in beta, it has landed known customers: Adobe, BambooHR, and Podium are among them.

02 /Why It Matters

Conceptually and technically, Buster resembles OpenCopilot – which was [covered here](/review/skoro-on-pojavitsja-na-kazhdom-sajte-v-internete) in the same YC batch.

The difference is scope. OpenCopilot set a broader goal: letting developers build full AI assistants for their cloud services – assistants that can guide users through tasks, or even execute tasks on their behalf. Analytics and reporting were on the list, but as one capability among many. Buster is entirely focused on that use case.

OpenCopilot's underlying pitch was to give developers "new interfaces that are better than traditional input fields and buttons."

That "better" interface is a chat, where users interact with services in natural language instead of navigating menus. It's hard not to notice that ChatGPT has meaningfully shifted people's sense of what a good interface feels like.

This is the trend worth watching: the shift in how cloud services interface with their users. From button-and-menu UIs toward intelligent chatbots you can simply tell what you want – without browsing, scrolling, inputting, and clicking.

The magnitude of this shift may be comparable to what happened with mobile phone interfaces. The iPhone arrived and everyone quickly understood that phones don't need physical buttons – tapping and swiping on a glass screen is just better. Then Siri and Google Assistant arrived, and voice commands became normal.

Gartner predicted last year that AI development would result in 10% of all customer interactions being automated by 2026 – up from just 1.6% at the time of their study.

But they were thinking narrowly: AI chatbots handling customer support tickets. What about users interacting with software itself? Isn't that also a form of customer interaction?

Users contact support when they can't figure out the interface. If AI chatbots can already answer those questions, why not embed the chatbot directly into the product? That would effectively erase the line between product UI and technical support.

03 /Opportunities

In the previous review on OpenCopilot, the case was made that every website should eventually have its own AI assistant – pictured as a widget floating at the side of the page.

But that framing may already be outdated. The AI assistant could become the primary interface of the site itself. The old functionality – pages, input fields, buttons – might end up as the sidebar, left there for power users and holdouts.

For cloud service developers, the implication is structural: the question isn't whether to add an AI assistant, but what kind. An assistant that merely answers questions already exists everywhere. The more durable competitive position belongs to assistants that anticipate what a user is trying to do and take action – which requires a fundamentally different architecture and a different training dataset.

Building those assistants is not easy, which is exactly what creates the platform opportunity: developers need infrastructure that handles the assistant's service knowledge, real-user interaction logging, and continuous retraining – none of which they want to build themselves. OpenCopilot and Buster are the early examples. More will follow. The migration of cloud service interfaces toward AI assistants looks inevitable. The space is large, the timing is right.