People and AI agents work in the same streams, the same bases and the same records. The AI is part of the workspace, not a chat window added on the side.
For every person and every agent, set what they may create, read, update and delete in each part of the workspace. Set it per person, role, agent, base or channel, in any combination. Anything not allowed is out of reach.
Every workflow run is saved step by step. For each step the log shows what it was given, what it returned, which model version it used, how long it took and how much credit it used. If a bill is higher than expected, it can be traced to the step that caused it.
Cognition is the page for choosing the AI models an agent uses. There are eighteen named tasks, from routing and retrieval to reasoning and governance, and each one can use a different model. A model that is good value for coordinating work is too expensive for simple tagging. Usage is charged at each model's published price.
Agent work is paid for from credit bought in advance. Automatic top-up has a monthly ceiling and a waiting period between charges. Without both, a standing instruction to charge a card would have no limit.
Chats, mail, meetings, records and decisions all go into one data warehouse with access controls. Agents read from it and show where each answer came from.
Intercom
OneDrive
Notion
Google Calendar
Slack
Google ChatFrom the connector library. Webhooks and a REST API cover the rest.
Agents answer from the warehouse and show what they used: a record, a decision or a moment in a call.
Company and agent policy documents, plus a knowledge base that every agent reads before it drafts anything for someone outside the company.
Rights are set for each connected source, so an agent reads only what its role allows.
A security review asks who can sign in, where the data is kept and what happens to it. Each answer is a setting in the workspace.
Each team has separate agents, bases and streams. They all work on the same customer records, so a handoff does not need a forwarded email or a status meeting.
Every lead arrives with a reply drafted, and the pipeline updates itself.
Daya for sales →Every DM, chat and form becomes a lead, and an agent follows up.
Daya for marketing →One thread per customer across every chat app, with the reply drafted.
Daya for customer service →Orders, stock and visits on one set of records, with a digest every morning.
Daya for operations →Requests, quotes and purchase orders in one place, with every approval on record.
Daya for procurement →Stock that reorders in time, and deliveries with proof on the record.
Daya for supply chain →Late invoices chased, bills approved and the books checked every night.
Daya for finance →Hiring, onboarding and leave in one place, and a handbook that answers questions.
Daya for human resources →One plan in four views, and minutes written after each meeting.
Daya for project management →Every bug linked to the customer who hit it, and tools built by describing them.
Daya for engineering →Routine tickets arrive with the answer drafted, and new tools take an afternoon.
Daya for IT →One set of rules for people and AI agents, with every step on record.
Daya for security →Internal tools no longer have to wait in a backlog. Someone describes what they need in a conversation with an agent. The agent builds it using any data in the workspace and publishes it privately. There is no code repository, no deployment and no ticket to raise.
Which model answers, what it costs, how it sounds and how much it may use. All of it is set in the workspace.
Use several AI providers at once and bring your own API keys. Which one an agent uses is a deliberate choice, not a default.
Eighteen kinds of task, and a model for each. Set a default for the workspace, and let a single agent change only the tasks it handles differently.
Anything an agent builds or sends follows a published design system. Use the workspace default or give that agent a different one.
See how much each agent, workflow and person is using, next to their limits.
The documents every agent reads before it drafts anything. They are managed separately from the prompt, so they are easy to control.
Upload a skill or give an agent a tool, and take either away later. Everything an agent can use is in one readable list.
Advanced removes the limits on history and rows. Beyond that, we price by volume and can run Daya in a private cloud or on your own servers.
Advanced includes priority consulting. Our consultants join the workspace as members, so they follow the same permissions as everyone else and everything they do is in the audit log.
No. Seats are for the people who log in. Agents, guests in shared channels and read-only viewers are not counted.
The first agent can be live within 48 hours, because we start small. Each department after that goes live when your team approves it.
Yes. Anyone in the stream can stop a run midway, and an admin can switch the agent off. The steps it had finished stay on record.
Billing is per workspace, so everyone in a workspace shares a plan. Guests invited into a single channel or base stay free.
Yes. The Free plan covers twenty people and includes the full authority matrix, so one team can try it with real work first.
No. Every plan has unlimited agents, and agents can supervise other agents. What limits each one is its authority and spending cap.
Before anything goes live, we will map the channels, the data warehouse and who approves what.