Aevi AI Assistant

Aevi is Aezona's AI-powered assistant, built into CostLens. Ask it anything about your cloud costs, recommendations, AWS best practices, or how to use CostLens. Aevi has full context about your accounts and recommendations — it's not a generic chatbot.

Aevi runs on Claude models from Anthropic. You can choose which model answers your chat questions, and every answer shows which model wrote it.

What you can ask Aevi#

• "Why is my EC2 spend so high this month?"
• "Explain this recommendation in simple terms."
• "What is an IAM Role vs access keys?"
• "What does rightsizing an RDS instance mean?"
• "How do I reduce my S3 storage costs?"
• "Summarize my top 3 savings opportunities."
• "What are Reserved Instances and are they worth it?"
• "Which account has the most waste?"

Which models Aevi uses#

Model	Best for	Cost per question
Claude Sonnet 4.6 (default)	Detailed analysis, comparisons, multi-step questions	Higher — roughly 4× Haiku
Claude Haiku 4.5	Quick, simple questions and fast replies	Lower

Both models read the same account data before answering, so for a simple factual question they often give similar answers. The difference shows on harder questions and in cost: a Haiku question uses much less of your monthly Aevi budget.

Choosing a model#

1. Open Aevi — the Aevi page in the sidebar, or the Ask Aevi side panel.
2. Under the message box, click the current model's name (for example, Claude Sonnet 4.6 (default)).
3. In Select a model, pick a model. Each one shows what it's best for and whether it costs more or less. You can type in Search models… to filter the list.
4. Your next message is answered by that model.

Your choice is remembered in your browser, so it stays selected when you come back. If your plan doesn't include a model, it isn't offered in the list.

Seeing which model answered#

• Every Aevi reply shows a small "Answered by Claude Sonnet 4.6" (or Haiku 4.5) label underneath it. The label is kept when you reopen a past conversation.
• You can also ask Aevi directly — "Which AI model are you?" — and it answers truthfully, for example: "I'm Aevi, powered by Claude Sonnet 4.6 from Anthropic."
• Aevi never shares its internal instructions, prompts, or any API keys. If asked, it politely declines and returns to your cost question.

Replies from before this feature was introduced don't have a label.

Aevi usage limits#

Each organization has a monthly Aevi budget in US dollars. Every chat question, insight refresh and AI explanation uses a small amount of it, based on the real price of the model that answered. The budget resets at the start of each calendar month.

Plan	Monthly Aevi budget	Included in your plan	When the budget is used up
Trial	—	—	Aevi chat isn't included on Trial. Upgrade to use it.
Personal	$2	$2	Aevi pauses until next month
Pro	$5	$2	Aevi pauses until next month. Usage between $2 and $5 is billed on your invoice.
Business	$10	$10	Aevi keeps working. Usage above $10 is billed on your invoice.
Enterprise	$10	$10	Aevi keeps working. Usage above $10 is billed on your invoice.

Personal and Pro are available to individual workspaces; Pro, Business and Enterprise to company (organization) workspaces. See Billing & costs for how Aevi usage appears on your invoice.

The usage meter#

The chat header shows how much of this month's budget is used, for example $0.08 / $2.00 (4%). It turns amber at 80% and red when the budget is used up. Click it to open Settings → Aevi for details.

Choosing Claude Haiku 4.5 for simple questions is the easiest way to make your budget last longer.

What happens when you reach the limit#

• Chat: sending a message shows a "Monthly AI usage limit reached" pop-up with your usage, and no new AI request is made, so nothing more is charged.
• Dashboard insights: Aevi Insights keep showing the last saved snapshot instead of generating new ones.
• Rule-based features keep working: recommendations, billing, anomalies and everything else in CostLens are not affected — only the AI-generated layer pauses.

On Business and Enterprise this doesn't happen: Aevi continues and the extra usage is billed.

How admins change limits#

Organization admins manage Aevi in Settings → Aevi:

• Turn Aevi on or off for the whole organization.
• Change the monthly limit — when your plan allows self-service, you can set any limit up to the ceiling shown on the page. Otherwise the limit shows "Managed by CostLens"; contact CostLens support to change it.
• Set a per-user limit so one person can't use the whole organization's budget.

Changing plans updates the Aevi budget to the new plan's default automatically, unless your limit was set by hand.

AI explanation on recommendations#

Every recommendation detail page includes an AI Explanation panel generated by Aevi. It explains in plain English why the resource is wasteful, what impact the change will have, and any risks to be aware of. You can regenerate this explanation at any time using the Refresh AI Explanation button.

Insights on the Dashboard#

The Aevi Insights card on the Dashboard is a standing set of AI-generated observations about your spend — anomalies, unusual patterns, and quick wins — organized by category (Compute, Storage, Database, Networking, Serverless, Observability, Other) and each carrying a confidence level (High / Medium / Low).

Unlike the chat, these insights aren't regenerated on every page visit. CostLens persists a snapshot per account (and a separate consolidated one across all accounts) and only regenerates it when the underlying data actually changed — a new sync, a new analysis run — rather than on a timer. Click the refresh icon on the card to force an immediate regeneration on demand.