Give each AI agent a role, such as finance, support or research, and describe a job in a sentence. The agent builds the workflow for it, so one team of agents can have many workflows running quickly.
A finance agent, a support agent, a research agent: each with a brief and tools.
A sentence becomes a workflow with readable steps.
Refunds, payments, bookings and briefs, each run by the agent for that role.
Small internal apps, built by describing them.
Connectors, webhooks and a REST API for the rest of the stack.
Fully automatic or with approvals, with every run logged.
An agent persona is a name, a brief, a personality and a set of tools: a finance agent, a support agent, a research agent. Describe one in a sentence and it appears on the org chart as a draft. Approve it, and it is ready for work. Each persona can use a different AI model.
Tell an agent what needs doing, in a sentence. It writes the workflow as readable steps: what starts it, what each step reads and writes, and where it branches. A workflow starts on a schedule, a condition, a webhook or an event in another app.
Business process automation rarely stops at one workflow. Each agent adds the workflows its role needs: routing refunds, reconciling payments, booking visits, writing the morning brief. They run side by side on the same records, and each one shows which agent answers for it.
Some processes need a screen. Describe the tool in a sentence and an agent builds it using any data in the workspace, then publishes it privately. Reply with a change and the app is updated.
Workflows read from and write to the tools a business already runs: the CRM, the shop, the accounting software, mail and calendars. Webhooks and a REST API cover anything the connector library does not.
A person approves a new workflow once, before it goes live. After that it can run from start to finish, or with an approval on any step that needs a person. Either way, every run shows each step as it happens, can be stopped at any point and lists what each step cost.
Each agent works inside its permissions and spending limit. Beyond that, the agents can run unattended, or ask for a person's approval on the steps that matter most.
It can be set up without our help. Pro and Advanced also come with consulting, and our team can do it with you.
A role and a brief, in a sentence.
Mail, the shop, the ledger or the CRM.
One for each job in the role.
Permissions, a spending limit and any approvals needed.
From the connector library. Webhooks and a REST API cover the rest.
Google Calendar
Slack
Google Chat
OneDrive
NotionUsing AI agents to carry out routine work: reading what comes in, updating records, drafting replies and starting the next step. In Daya each agent has a role and builds the workflows for it, on shared records.
An ordinary automation follows fixed rules. An AI workflow can also read a message, a document or a call transcript and work out what it means, so it covers steps that used to need a person.
Running a process such as approvals, onboarding or month-end as a defined sequence with owners and deadlines. Here each process is a set of workflows on shared tables, so every case is visible, along with where it has stopped.
Describe the job in a sentence and the agent builds the workflow as readable steps. Each agent can build a workflow for every job in its role, so a few agents cover many processes.
No. Someone approves a new agent and each new workflow once, before it goes live. After that a workflow can run unattended inside the agent's permissions and spending limit, or with an approval on any step.
No. Describe an agent's job or a tool in a sentence. A developer can extend things later through the API, but building needs no code.
Free for up to twenty people. Describe one agent, and let it build its first workflows.