Common objections
We already have Copilot, ChatGPT, etc.
Keep them. Datafi gives them governed access to systems outside Microsoft's or Google's cloud, like your ERP, TMS and on-prem databases. Copilot is an integration partner, not a competitor.
Connecting AI to our data is too risky.
Start read-only, with one team on one dataset. Every request is checked against that person's row-level access and logged, and you widen access only when satisfied.
We'll build it ourselves.
A production MCP server is a big project alone, before per-user governance, business context, observability and upkeep. Datafi gives you all of that from day one.
Product
What does Datafi actually do?
It connects your line-of-business systems to your people's AI tools. Each question is checked against the person's access, answered live from the source with business context, and logged.
Which AI tools does it work with?
Microsoft Copilot, ChatGPT, Claude, Gemini, Grok, Llama, any compatible MCP client, your own model, and Datafi's business chat and browser extension.
Can different teams use different AI tools?
Yes. Each person can use the tool they prefer, at the same time, with the same per-user rules and one Control Tower log across all of them. Add or switch tools later without redoing access rules.
Do we need to move or copy our data?
No. Datafi connects to your systems where they are, and the AI works from live source data.
What are Converse and Activate?
Converse lets you ask questions, get answers and generate reports, with Chat, Teams and Cards. Activate adds agent and app builders inside your AI tool, plus Datafi Studio to refine agents visually. Compare them.
What are Basic and Standard licenses?
A Basic license lets someone ask questions and use published agents and apps. A Standard license also lets them build, deploy and refine those agents and apps. Converse needs only Basic licenses; Activate needs a mix of both.
Does it depend on which AI tool we use?
No. Converse and Activate work the same in Copilot, ChatGPT, Claude, Gemini or Datafi's business chat, on your complete business data.
Where should we start?
Teams with Copilot or ChatGPT usually prove Converse first, then move to Activate. Teams without an AI partner often start with Activate.
What happened to Bedrock?
Bedrock is now simply Datafi. Nothing about how it works has changed.
What is Control Tower?
Where you manage oversight, observability and governance: policies, connections and the AI activity log.
What doesn't Datafi do?
It doesn't replace your AI tools, systems of record or reporting warehouse. Agents hand decisions to a person where a person should decide.
Security and control
How is access controlled?
Per user and per row. Each AI request is checked against the asker's identity and permissions, so the AI never sees more than they do.
What gets logged?
Every request and answer: who asked, what, which data was used and what came back. It's the record auditors, boards and cyber insurers want.
Where does Datafi run?
Datafi Hosted runs in Datafi's cloud, for accounts up to 100 users. Customer Hosted runs in your own cloud and is required for more than 100 users.
Pricing and getting started
How much does it cost?
Per user: $20 per Basic license and $40 per Standard license per month at list, with volume discounts up to 25%. Minimum 50 users, or $12,000 a year. Your license spend also becomes credits, at $0.004 each. Full pricing is published.
What isn't included in the price?
Services like the launch workshop, your own LLM costs, and cloud costs if you host in your own cloud, paid directly to those providers.
How long until we see value?
Your first use case reaches production after a launch workshop of 3 days or less. Most teams know within weeks whether to expand.
Can we start smaller than a full rollout?
Yes: one team, one dataset, read-only. Add users as it proves out; the price per user falls as you grow.
No questions match that search. Try another word, or ask us directly.