CAPABILITY · AI ASSISTANT

Ask your fleet.
In natural language. With real data. In seconds.

AI assistant based on Anthropic's Claude Sonnet 4.6. Real-time access to your tenant data. Answers complex questions, generates correlations, writes remediation scripts, and prepares reports for your board.

See pricing Request demo

"How many devices in Madrid have Office 2016?"

In any traditional tool, that question requires: going to Inventory, applying filter by site=Madrid, applying filter by software=Office 2016, counting results. 5 minutes. And for each follow-up question, you repeat the process.

With the Itamite assistant, you write the question and get the answer in 3 seconds. And you continue: "Which ones? When does their license expire? What is their average security score? Generate me the script to update Office 365 on those devices." The assistant remembers context and goes deeper.

REAL USE CASES

What administrators already do with the assistant

Incident investigation

"Barcelona site score dropped 15 points this week. What controls changed?" The assistant analyzes site history, detects the affected control (pending patches after Patch Tuesday), lists specific devices and proposes corrective action.

Audit preparation

"Are we ready for next month's NIS2 audit? What controls are we missing?" The assistant reviews current NIS2 compliance status, lists gaps by regulation article, and proposes prioritized remediation plan.

License economic analysis

"How much are we spending on software licenses no one uses? Top 5 with cost." The assistant cross-references contracted licenses with real usage measured by agent, calculates annual waste, and lists top 5 savings opportunities with specific figures.

Complex searches

"Give me all laptops with Windows 10 unpatched for over 60 days AND with BitLocker disabled." The assistant generates the multi-criteria filter, executes the query and returns the list with key data (hostname, user, last login).

Script generation

"Generate me the PowerShell script to enable BitLocker on these devices with AD recovery." The assistant generates the complete PowerShell script, commented, with error handling and verification command. Ready to review before executing.

Reports for management

"Quarterly executive summary: posture, compliance, incidents. No technical jargon." The assistant generates 3 paragraphs in plain language to present to the board, with specific figures, no technical jargon.

PRIVACY AND CONTROL

Your information, your rules

Access only to tenant data

The assistant only sees data from the tenant you're operating in. Cannot access other tenants' data or cross-data. Your Itamite role permission also applies to the assistant.

BYOK applied to assistant

If you have BYOK active, queries to the assistant are encrypted with your KMS key before sending to Anthropic. Anthropic receives encrypted text, not plaintext. Critical for regulated sectors (healthcare, finance, public sector).

Zero training with your data

Anthropic does NOT use Itamite prompts to train its models. Contractually agreed and auditable. Assistant usage logs are deleted after 30 days by default, configurable to 7 days in Enterprise plan.

The assistant does NOT execute actions

By design, the assistant only analyzes and responds. NEVER executes commands, modifies configuration, or sends messages. You execute actions from the console after reviewing the assistant's proposal. Zero risk of "the bot did something I didn't want".

FAQ

Assistant questions

What AI model does it use?

Claude Sonnet 4.6 from Anthropic. State-of-the-art model in technical reasoning and long context understanding (1M tokens). When Claude Opus 5 comes out, we automatically update at no additional cost for Enterprise customers.

How many queries can I make?

Business: 100 conversations/month per tenant. Professional: 500/month. Enterprise: unlimited. A "conversation" is a continuous session of related questions, not each individual message.

Does it work in other languages?

Yes. Ask in any language and it responds in the same. Tested and optimized for: Spanish, English, French, German, Italian, Portuguese, Dutch, Polish. For less common languages, quality may vary.

What about sensitive information?

Queries are transmitted encrypted TLS 1.3 to Anthropic. If you have BYOK, also encrypted with your KMS key. Anthropic processes and responds without saving content for training. Logs in Itamite are kept according to your policy (default 30 days). For ultra-regulated sectors, consider the Sovereign on-prem plan.

Can the assistant be wrong?

Yes, no LLM is infallible. For factual questions about your fleet (figures, lists, configurations), the response is highly reliable because it comes from your real data. For technical opinions or recommendations on very recent regulation, cross-reference with official sources before acting. Generated scripts and SQL must be reviewed before executing.

Is there an on-premise model option?

Yes, in Sovereign / On-prem plan. We deploy the assistant with a local AI model (Llama, Mistral or equivalent) running in your infrastructure without external connection. Quality ~80% of Claude Sonnet 4.6, sufficient for most operational cases. Indicated for defense, critical banking, high-category public sector.

The assistant that understands your IT fleet

30-min demo with your real data. We show you what you can ask it.