--- title: "AI assistant - ask your company’s knowledge, and let it do the work | Floral" description: "Floral’s in-app chat searches, analyses and acts across your whole knowledge engine - meetings, sources, contacts, deals and tasks. Works with your permissions, asks before it writes, and connects to any AI model through MCP." canonical: https://floral.so/features/ai-assistant --- # AI assistant - ask your company’s knowledge, and let it do the work | Floral > Floral’s in-app chat searches, analyses and acts across your whole knowledge engine - meetings, sources, contacts, deals and tasks. Works with your permissions, asks before it writes, and connects to any AI model through MCP. Source: https://floral.so/features/ai-assistant Core feature · Digital colleagues # Ask your knowledge engine - and let it do the work Everything your company knows, one question away. The Floral assistant sits inside your workspace with access to the whole knowledge engine: meetings, summaries, briefings, transcripts, your own documents and data sources, plus companies, contacts, deals, tasks and notes. It finds, it analyses, and it finishes the job - with your permissions, and never without your approval. [Sign up](https://floral.so/signup)[Let's meet](https://floral.so/book-a-demo) What it does ## Find, analyse, act Most AI tools stop at retrieval. The assistant reasons across everything your company knows, then reaches for the same tools your team uses to actually change something. ![Scattered documents becoming answers and completed work](https://floral.so/img/assistant-illustration.png) ### Find across everything Ask in plain language. Floral searches by keyword and by meaning across meetings, summaries, briefings, transcripts and every source you have uploaded. ### Analyse, not just retrieve Compare accounts, pull the pattern out of a quarter of conversations, summarise what changed since the last meeting. It reasons across the workspace, not one document at a time. ### Get the work done Create the task, write the note, update the record, draft the follow-up - and chain several steps together until the job is finished. ### Grounded in your own material Every answer shows the meeting, summary or source behind it, so you can open the evidence and check the work. ### Works as you, not above you It acts on behalf of the person using it with exactly their permissions. Membership and access are re-checked on every request, and every action is logged. ### Approval before every write Reading is free. Anything that changes a record stops and asks first, so the assistant can be useful without being risky. Open by design ## Connect it to any AI model, through MCP Floral runs its own MCP server with 58 tools. Claude, ChatGPT, Codex and any other MCP-capable client can reach the same knowledge and the same tools, with the permissions of the person using it - and a connection made over OAuth is read-only. The model you prefer today does not have to be the model you are stuck with next year. That is the whole point of building on an open protocol rather than a single vendor's assistant: your knowledge stays in your workspace, your data stays yours, and the AI layer on top stays replaceable. MCP access is enabled per workspace, and the same tools are available through the public API. [See the API and MCP server](https://floral.so/features/api-and-mcp)[Let's meet](https://floral.so/book-a-demo) Without being asked ## Floral suggests the next step on its own You should not have to know the right question. When a meeting summary is approved, Floral works out what follows from it and brings that to the person it belongs to. ### Tasks from what was agreed Every to-do captured in an approved summary becomes a proposed task, with the deadline and priority it was given in the conversation. ### Insight for the individual Suggestions are personal, not a team firehose. Each person gets a short review inbox with what is relevant to their own accounts and work. ### Records that should move When a conversation implies a company detail changed or a deal moved on, Floral proposes the update - with the exact quote it is based on. Related features ## Pairs well with [AI meeting summariesApproved summaries are what the assistant reasons over and what suggestions are built from.Learn more ](https://floral.so/features/ai-meeting-summaries)[Workspace AI contextTeach the assistant your vocabulary, your industry, and what to ignore.Learn more ](https://floral.so/features/workspace-ai-context)[Per-user feature accessThe permissions the assistant inherits are the ones you set here.Learn more ](https://floral.so/features/per-user-feature-access) [All core features](https://floral.so/features)[All CRM add-on features](https://floral.so/features/crm) ## Frequently asked What can I actually ask the assistant? Anything your company has recorded. It searches by keyword and by meaning across meetings, summaries, briefings, transcripts and the documents and data sources you have uploaded, plus companies, contacts, deals, tasks and notes. Ask what a customer said about pricing last quarter, what changed since the last meeting, or which accounts have gone quiet. Can it do things, or only answer questions? It can do things. The assistant reaches the same tools your team uses - create a task, write a note, update a record, move a deal - and chains them to finish a job. Every write asks you for approval before it happens. Can it see things I am not allowed to see? No. The assistant acts on behalf of the person using it with exactly their permissions, and never more. Feature access, record access and workspace membership are re-checked on every single request, and every action it takes is logged. Can I use it from ChatGPT or another AI tool? Yes. Floral runs its own MCP server with 58 tools, and it works with Claude, ChatGPT, Codex and any other MCP-capable client. It acts with your own permissions, and a connection made over OAuth is read-only. That is the point of building on an open protocol: you are never locked to one vendor’s model. MCP access is enabled per workspace, and the public API is there when you need more. Does Floral suggest things on its own? Yes. After a meeting summary is approved, Floral proposes the next step to the person it belongs to - follow-up tasks from the to-dos that were agreed, records that look out of date, and insight worth knowing. Suggestions land in a personal review inbox where you edit, apply or dismiss them. Nothing is written without a person deciding. Where does the answer come from? From your own material. Answers show the meeting, summary or source they are built on, so you can open the evidence and check the work rather than trusting it blindly. Insights ## Read more [Insights →](https://floral.so/insights) - April 2, 2026 · 10 min read ### [Sales knowledge base best practices: the four components of one your team will actually use](https://floral.so/insights/sales-knowledge-base-best-practices) Most sales knowledge bases die within a year. Here are the four components that make a sales knowledge base self-sustaining, and the honest criteria for evaluating tools. - December 16, 2025 · 7 min read ### [How to cut new sales hire ramp time in half](https://floral.so/insights/how-to-cut-new-hire-ramp-time-in-half) The average B2B sales rep takes 8 to 12 months to reach full productivity. That is too long. Here is how leading sales teams are using meeting intelligence to give new hires instant access to institutional knowledge, cutting ramp time dramatically. ## Put your company’s knowledge one question away Sign up, connect a calendar, and ask the assistant something only your company would know. [Sign up](https://floral.so/signup)[Let's meet](https://floral.so/book-a-demo)