.jpeg)
Imagine opening your AI assistant in the morning and asking, “What happened on the website yesterday?” And instead of a generic answer, you get a real analysis. With your data, your metrics, your context.
It's not a chatbot that looks at a dashboard. This is a direct connection between your analytics stack and an AI model that understands and classifies your data and recognizes connections that you would never have noticed in the dashboard.
The biggest problem when you let go of AI on raw data: It knows the numbers but not the meaning. 500 visitors to /pricing — is that a lot or a little? Without context, any answer is worthless.
That is why bchic has a built-in Knowledge Catalog. It works like this: There are stored descriptions, benchmarks and interpretive aids for every metric, dashboard and important KPI. The AI therefore not only knows that the bounce rate is 67%, but also that this is a normal value for a landing page in the B2B sector, but a warning sign for a product page.
That makes the difference between an answer like “The bounce rate is 67%” and “The bounce rate on the product page is 67%, which is well above your 90-day average of 52%. Since the last deployment on Tuesday, the figure has risen. Possible cause: Charging time has worsened.”
Specifically, the catalog contains:
The result: You're not asking a generic language model that knows analytics terms. You're asking a model that understands your specific setup.
Analyses don't happen in a vacuum. Anyone evaluating a campaign today launched a redesign last week and installed a technical deployment three days ago. Without this context, each individual analysis is only an excerpt.
Short-term memory solves just that: it stores recent interactions, analyses, and insights and automatically provides them as context for every new question.
An example: On Monday, you ask “How does the new landing page perform?” and receive an analysis with comparative values. On Wednesday, you're asking “Has that improved?” and the AI knows what you're referring to. It automatically compares Monday's values with the current status without you having to specify the landing page again or manually set the time period.
That sounds like a small thing. In practice, it's the difference between a tool that you survey and an analyst who thinks along.
What tracks short-term memory:
This creates a coherent picture over the days instead of isolated data points.
It's not a feature that you have to configure before it becomes useful. Here are real queries that are working from now on:
Traffic analysis:“How did traffic develop this week?” provides an overview with daily values, comparison with the previous week and trend classification.
Campaign evaluation:“What are the benefits of the Google Ads campaign?” shows UTM-filtered visitor numbers, engagement metrics, and a comparison to organic traffic.
Page performance:“Which sites have the highest bounce rate?” provides a ranking with context. Not just the number, but also whether the value is above or below the average.
Change Impact:“What did Tuesday's deployment bring?” accesses the change impact analysis and automatically provides the before/after values.
Target group analysis:“Where do our visitors come from?” Breaks up traffic by country, device, browser and referrer.
Funnel analysis:“What does our conversion funnel look like?” shows the interruption rates between defined steps and identifies the largest drop-off.
Compare:“How is this week performing compared to last month?” Put two time periods side by side, with percentage changes per metric.
These aren't demo prompts. These are the questions we ask ourselves every day.
New tools usually fail in one of two places: Either they are so complex that only the technical lead can operate them. Or they're so open that you don't know where to start.
bchic MCP is neither of them.
Setup in less than two minutes: You generate an API token in bchic under Settings > Integrations, enter the MCP URL in your AI assistant, and you're done. No SDK, no library, no deployment.
Guardrails instead of free text chaos: The AI doesn't access an open database and write SQL. It uses predefined, tested analysis functions with clear parameters. This means: The answers are reproducible, the metrics are correct, and no one accidentally receives incorrect figures because a prompt was poorly formulated.
Reading rights, not writing rights: The MCP connection can query and analyze data, but it can't change anything. No tracking code is modified, no settings are overwritten, no data is deleted. If you want, you can create change events, but that's it. Full control remains with you.
GDPR from the start: The AI connection processes aggregated analytics data, not personal information. Since bchic works cookieless and does not collect any PII, there is no data protection problem on the AI side either. No additional processing agreement, no new legal basis required.
The honest answer: Because we needed it ourselves.
We built bchic from the start with the claim that analytics data should not be acidified in a dashboard. Dashboards are good for monitoring. Decisions require interpretation. And interpretation doesn't scale when it takes place in a single person's mind.
The classic analytics world works like this: collect data, build a dashboard, set up a meeting, someone presents figures, the team discusses, and at some point a decision is made. This process takes days. Sometimes weeks.
With an AI interface, this becomes: ask a question, get an answer, decide. In minutes, not days. Not because the AI is smarter than the team, but because it eliminates the manual part of analysis: collecting data, comparing time periods, recognizing patterns, creating context.
It's not a feature. It's a different way of working with data.
What you're seeing today is the beginning. The MCP interface is the basis for everything that comes along.
Proactive insights: Instead of waiting for questions, the system detects anomalies and actively informs you. Traffic slump on Sunday evening? You'll find out Monday morning at 8, not Wednesday in the Weekly.
Automated reports: Weekly summaries that consist of contextualized analyses rather than standard tables. What happened, why did it happen, what should you do next.
Deeper links: Change impact analysis, funnel data, traffic sources and target group segments as a coherent picture. Not five separate dashboards, but an answer to “Where are we losing money right now and why?”
Multi-source analysis: In the long term, we not only want to make bchic data analyzable, but also enable links with CRM, ad platforms and business data. So the answer to “Is this campaign worth it?” contains not only visitor numbers, but real revenue.
MCP integration is now available to all bchic customers with the Growth Plan or higher. No upgrade, no extra charge, no feature gate.
If you don't use bchic yet: Create a free account and be live in five minutes. Cookieless, GDPR-native, and now with a brain that thinks for itself.
