Data Agent
An analyst that reads your data and shows its work.
Ask in plain words. The agent digs through your data, checks what it finds, and answers with the headline, the numbers and a chart.
In plain words
What is an AI agent?
A chatbot answers from memory, so about your sales it can only guess. An agent takes action: it looks at your real data, tries things and checks the results until it has an answer it can back up.
Think of a sharp analyst who's new to your company. It gets even better once you tell it what your numbers mean. That's what Metrics are for.
Anatomy of an answer
Answers you can act on, and check.
- 1
The answer first
A headline and the few numbers that matter, in plain words.
- 2
A chart you can change
The agent picks a chart that fits. Switch it with one click.
- 3
“How I got this”
Every step, in plain words, so you can check its work.
Why did revenue dip last week?
Revenue fell 13% last week, and almost all of it came from one-time orders.
- Subscriptions held steady at $8.2k (−1%).
- One-time orders dropped from $4.1k to $2.6k (−37%).
- The drop starts Tuesday, the day the free-shipping banner ended.
One-time orders by day
✓ Revenue · certified week before last weekHow I got this· 4 steps
- Used your Revenue metric (certified)
- Looked at your orders
- Pulled the last 14 days
- Split revenue into subscriptions and one-time orders
What you can ask
If you'd ask a colleague, you can ask the agent.
No special words. Start broad, then follow up.
Understand
What happened?
- “How many new customers did I get last month?”
- “What's my average order value?”
Diagnose
Why?
- “Why did revenue dip last week?”
- “Which products get refunded most?”
Compare
Better or worse?
- “How does this month compare with last year?”
- “Which plan keeps customers longest?”
Build
Make it stick.
- “Build me a weekly sales dashboard”
- “Define “active customer” for me”
It does the work
Not just answers. Things you keep.
Ask for a dashboard and you get one. Ask it to define your numbers and you get a draft to approve.
Builds dashboards
Drafts your metrics
Remembers the conversation
Your choice of AI model
Knows when to pause
Sees all your data
Safe by design
Powerful enough to help. Fenced in so it can't hurt.
Letting AI near your business data should feel calm. These limits are built in.
Creates, never deletes
It can add charts and dashboards. It can't change or remove anything you've made.
You approve the numbers
Metrics it proposes are drafts until you approve them.
Acts only when asked
It saves things when you ask, never because something in your data tells it to.
Sees only what it needs
Your AI provider sees at most 30 rows of any result.
Nothing hidden
Every step it takes is listed, in plain words.
Private conversations
Your chats with the agent are visible only to you.
Bring your own key
Your AI, your bill.
The agent runs on your own AI key. You pay your provider directly, at cost. Orcabase never marks up your AI.
Orcabase's agent
Does the digging and shows you the answer.
Your key
From OpenRouter or Vercel AI Gateway. Stored encrypted.
Hundreds of models
Billed by your provider, at their price.
Checked before saving
A bad key never gets saved.
Prices up front
See what each model costs before you pick it.
Spend you can see
Every conversation shows its cost.
Do I need to be technical?
Which AI models can I use?
What does it cost?
What does the AI provider see?
Can it break anything?
What if it gets something wrong?
Can I use it from Claude or ChatGPT?
Hire the analyst. Skip the salary.
Add your AI key, connect your data, and ask your first question in minutes.