AI
Product News
What's new in the CloudQuery Platform AI Assistant
TL;DR
When we launched the AI Assistant, it let you ask questions about your cloud in plain English and get answers or SQL back. This update turns those answers into things you can keep and act on. You can now build reports straight from a conversation, revisit past chats, manage policies without leaving the assistant, and attach your own business context so answers fit your environment. Under the hood, the assistant now runs on a new agent platform that's more reliable under load. We've also shipped an early proof of concept for understanding how your cloud resources relate to one another.
A few months ago, the AI Assistant became the home page of the CloudQuery Platform. You could ask a question in plain English and get an answer directly, or get SQL to run yourself. That solved the first problem: getting from "I have a question" to "I have an answer" without writing a join.
But an answer is often just the start. You want to save it, share it, turn it into something that runs on a schedule, or pick up a conversation from last week. This release is about closing that second gap by turning ad-hoc answers into reusable, shareable work.
Your cloud data is already in CloudQuery. But getting an answer to "how many unencrypted RDS instances do we have across all environments?" still means switching to the SQL Console, writing a join or two, and hoping you have the right table names.
That friction is what the new AI Assistant removes.
What Is the CloudQuery AI Assistant? #
The AI Assistant is now the home page of the CloudQuery Platform. Instead of a dashboard of charts, you get a conversation window - type a question about your infrastructure in plain English and get an answer back.
This is different from the SQL query writer in the SQL Console. That tool stays where it is, unchanged. The AI Assistant is a separate feature built on a new architecture, designed to be extended over time. The two will eventually converge, but for now they're independent.
We built this because of a pattern we kept seeing: teams had all their cloud data normalized and queryable in CloudQuery, but the step from "I have a question" to "I have an answer" still required SQL fluency. Not everyone on a platform team writes queries every day. The AI Assistant closes that gap.
Key Takeaways
- You can now build and update reports directly from a conversation
- Session history means past conversations are saved, searchable, and browsable
- Custom context lets you steer answers toward your own environment and terminology
- You can create and query policies conversationally, without leaving the chat
- The assistant now runs on a new agent platform that's more reliable under load
- An early relationship-awareness proof of concept gives the assistant a better sense of how resources connect
Build Reports Straight From a Conversation #
Previously, building a report meant doing it by hand and wrestling with our report YAML specification, which could pose a real barrier if you weren't already comfortable with it.
Now you can ask the assistant to build a report directly from an ad-hoc question, or add a chart to a report you already have. Reports are private by default until you choose to share them, so you can refine something before anyone else sees it.
The benefit is simple: a one-off question becomes a reusable, shareable asset without you ever leaving the conversation. Ask "how many unencrypted volumes do we have by account?", like the shape of the answer, and turn it into a report you can revisit next quarter.
Pick Up Where You Left Off With Session History #
Before, the assistant didn't remember previous conversations. If you closed a tab or came back the next day, that work was gone, and you'd often end up asking the same question again.
Now, past AI conversations are saved, searchable, and browsable. You can go back to an answer you got last week instead of reconstructing it from scratch, and you don't lose your place when a tab closes.
Custom Context, Now Even Smarter #
The assistant already understands your cloud asset inventory schema. It knows what an EC2 instance is. What it couldn't know is what your organization means by its data: which tags map to which teams, who owns what, or which accounts are production versus sandbox. That knowledge tends to live in people's heads.
Custom Context: information you attach to a conversation to steer the assistant's answers toward your own environment, ownership model, and terminology.
Now you can define and attach custom context to a conversation, so the assistant's answers reflect how your infrastructure is actually organized. The payoff is more relevant answers without re-explaining your environment specifics every time you start a chat.
Turn a Question Into a Policy Without Leaving the Chat #
Policies are how you take a good question and make it run on a schedule. Getting there used to take a detour. You'd generate a SQL query with the assistant, then leave to go define a policy separately.
Now, if you have the right permissions, you can do it all in one place. Have an ad-hoc conversation about an issue, generate the query, and turn that query into a policy without ever leaving the conversation. It removes a manual, easy-to-forget step from a workflow teams run all the time.
More Reliable Under the Hood #
Not every improvement is something you click. We fully retired the legacy conversations backend and moved the assistant onto our new agent platform.
The practical change: hard timeouts have been replaced with iteration limits and background polling. In plain terms, a long-running question keeps working in the background instead of failing in front of you. The result is fewer dead-end errors and more consistent behavior when the system is busy.
Relationship Awareness (An Early Proof of Concept) #
Until now, the assistant worked out how resources relate to one another—say, which security groups are attached to which resources—by reading the database schema and the underlying data.
We've started giving it something better: access to an ontology and taxonomy that describe those relationships directly. This is an early look at what happens when the assistant understands the shape of your infrastructure, not just its contents.
A note on scope: This is a proof of concept. It's scoped to what's shown in the interface, it doesn't yet cover the full ontology, and it currently supports AWS only. Expect it to grow from here.
Is This Safe for Enterprise Use? #
The safety model from the original release still holds. Data privacy mode remains the default: the assistant works with your schema to generate SQL, and access to your actual infrastructure data is an explicit, admin-controlled decision. The new features follow the same rules. Reports are private by default until you share them, and policy management respects your existing permissions, so only users who are already allowed to manage policies can do so from the chat.
The move to the new agent platform also means we can keep a closer eye on quality and reliability as capabilities expand.
What the AI Assistant Still Can't Do #
This is an evolving product, and a few things remain out of scope for now. Relationship awareness is a proof of concept limited to AWS and to what's shown in the interface, not the full picture. As always, the assistant is scoped to your cloud infrastructure; it isn't a general-purpose web tool. And the SQL Console isn't going anywhere: if you prefer writing queries directly, that path stays exactly as it was. The AI Assistant is an additional way in, not a replacement.
See the AI Assistant in Action
We can walk you through the new reporting, policy, and custom-context features and help you figure out the right setup for your team. Or explore the AI Assistant docs to get started on your own.
Frequently Asked Questions #
Can I build a report without knowing the report YAML format? #
Yes. You can now ask the assistant to build a report directly from a question you've asked, or to add a chart to an existing report. No YAML is required. Reports stay private until you choose to share them.
Are my past AI conversations saved now? #
Yes. Session history saves your past conversations and makes them searchable and browsable, so you can return to an earlier answer instead of re-asking or losing work when a tab closes.
What is custom context, and how is it different from before? #
Custom context is information you attach to a conversation to describe your specific environment: what your tags mean, who owns what, which accounts are production. Previously the assistant could only infer structure from the available data; now you can tell it directly, which produces more relevant answers.
Can I create a policy from the AI Assistant? #
If you have the right permissions, yes. You can generate a query in conversation and turn it into a policy in the same place, without switching to a separate workflow.
Why does the assistant feel more reliable now? #
The legacy conversations backend was retired in favor of a new agent platform. Hard timeouts were replaced with iteration limits and background polling, so long-running questions no longer fail visibly; they keep working in the background.
What is relationship awareness, and which clouds does it support? #
It's an early proof of concept that gives the assistant access to an ontology and taxonomy describing how resources relate, rather than inferring relationships from schema and data alone. It's scoped to what's shown in the interface, doesn't yet cover the full ontology, and supports AWS only for now.
