Dashboards answer the questions you predicted when you built them. The real questions show up later, and they are almost always ad-hoc: why did sessions spike on Monday, which campaign drove that, did the new landing page actually change anything.
Answering those usually means building another report, exporting a CSV, or pinging an analyst. d8a.tech now includes Ask AI: a chat that talks directly to your analytics data and produces insights without leaving the dashboard.
What Ask AI Can Do
Ask AI works a question the way an analyst would. You ask it in plain language, and the agent decides how to answer: which data to pull, how to aggregate it, and how to present the result. Answers come back as text, inline charts, and tables, directly in the chat.
Under the hood, it combines two kinds of access that analytics chatbots usually treat as either-or: the reporting API and raw read-only access to the underlying database tables. On top of that, it has access to the property settings, so it can investigate with the actual setup in mind: anonymization settings, the identifiers in use, and applied traffic filters. More on how that works below.
The Ask AI panel. The agent can read the current dashboard context, answer follow-up questions, compare segments, and render charts in the chat.
Preloaded With Your Dashboard Context
Every chat starts with the context of the dashboard you are currently viewing. The metadata forwarded to the AI engine includes the selected date range, the active metric, and any filters you have applied.
That makes deep dives immediate. You do not need to explain what you are looking at. Even a lazy prompt works: ask “any news?” and the agent already knows you are looking at users over the last 28 days, pulls the data, and summarizes what stands out.
A two-word prompt is enough. The agent inherits the dashboard’s date range and metric, pulls the trend, and lists the takeaways.
The context is not a black box either: you can preview exactly what was forwarded to the agent at any time. And when the context changes because you edited the dashboard, a new session picks up the fresh state in one click.
How It Talks to the Data
Under the hood, the agent has two ways to reach your data:
- The reporting API. The same API that powers the dashboards. For standard questions like trends, comparisons, and breakdowns, the agent calls it the way any client would.
- Direct read-only database queries. For questions the reporting API was never designed to answer, the agent writes SQL against the same tables your BI tools can use.
The reporting API path covers most everyday questions:
“Take the sessions metric for last week and compare it with the previous one.” One prompt, one chart, and a day-by-day breakdown.
The SQL path is what makes the agent genuinely useful for analysis. When a question crosses report boundaries, the agent queries the tables directly:
A custom funnel built from an event parameter. No report covers this, so the agent inspects the schema, writes its own read-only SQL, and reports drop-off per step.
This works because of a decision we made long before the agent existed: the database schema is flat, documented, and designed to be queried by humans and BI tools. It turns out that what is easy for an analyst to query is also easy for an LLM to query. If your data model needs a tribal-knowledge decoder ring, no amount of prompt engineering will save the agent. Clean, documented data is what AI-ready actually means.
Built With Security First
Giving a language model direct access to run database queries is the part where most teams should get nervous. We designed for it from the start.
Our cloud is a multi-tenant environment, so isolation is strict:
- One property per session. Each AI agent session has access to exactly one property. There is no way to reach across tenants, and no shared context between properties.
- Read-only by construction. The agent’s database access is read-only. It can query your data, but it cannot modify, delete, or export it anywhere.
- Inspectable context. What the agent receives is visible in the session context view, not hidden in a prompt you cannot see.
This is especially important because the model runs queries directly. Tenant separation is not enforced by the prompt asking the model nicely. It is enforced below the model, at the access layer.
EU-Based LLMs by Default
By default, the agent relies on EU-based LLM models. The default model is fully open source and runs on EU infrastructure, so your data is never shared with non-EU companies.
If you chose d8a.tech because you care about where your analytics data lives, an AI feature that quietly ships your data to a US inference endpoint would undo that choice. So we did not build it that way. Unless you deliberately switch to a model of your own choosing, your questions, your dashboard context, and the query results the agent reads all stay within EU-based model providers.
Bring Your Own Model
The agent’s intelligence does not live in one specific model. The crucial part, the orchestration, the data access, the safety boundaries, lives in the d8a.tech core. The LLM API is a swappable component that the core manages.
That design has a practical consequence: specific accounts can integrate their own LLM APIs. If your organization has an existing agreement with a model provider, or runs its own inference, Ask AI can use it.
In our testing, any model scoring at least 40 on the Artificial Analysis Intelligence Index produces good quality insights. For reference Claude Sonnet max scores 51.7, Kimi K2.6 scores 54 and Fable 5 scores 64.9. All of them work well with our setup, which means you can trade off cost, speed, and provider preference without losing the feature.
What To Do Next
Ask AI is available now. Open any dashboard, start a chat, and ask the question you would normally turn into a report request.
Start with something you already half-know the answer to. Watching the agent pull the same numbers you would have pulled, faster, is the quickest way to build trust in the answers you did not predict.
If you are not on d8a.tech yet, start with a free cloud account. It takes a few minutes, and the database schema docs show exactly what the agent, and you, can query.