Agents
Autonomous workers with a job, acting inside your systems
A Riff agent owns a piece of your operation, like order confirmation or supplier chasing. It works its queue around the clock inside your systems, reports to the people who run that work, and gets better every night.
Tested tools and procedures you can read let agents act in the ERP, hold steady through model upgrades and earn autonomy over time.
Operational work, end to end
An agent carries real cases across days and systems, not a single exchange.
Complex work, carried through
It reads the order or claim, checks it against agreements, stock and history, and carries the next step out, start to finish.
Every channel, both directions
Email, Teams, Slack, phone and portals, on whichever channel the other side already uses.
Acting inside your systems
Agents trigger your systems' own business functions: create the sales order, confirm the date, post the goods receipt.
Learning that compounds
Every conversation and every action adds to what all agents know.
- In conversation
Taught by the people it reports to
Corrections in the conversation become procedure, and what one agent is taught, every agent knows.
- Every action
Every interaction adds context
Every email, read and action adds to the shared context, so agents know more of your business the longer they run.
- Every night
A review every night
Each night the agent reviews its day and proposes better instructions, tools or skills. Its people accept or reject them.
Built to work with people
Agents and people share the same cases, the same views and the same language.
Shared working surfaces
Agents and people work the same cases in the same Interfaces, each seeing what the other has done.
Adapted in plain language
Tell an agent a new rule or exception and it becomes procedure, versioned like everything else.
Makes its own tools when asked
On instruction it builds the tool or skill a task needs, tested and reviewed before it reaches production.
Predictable by design
Behaviour you can inspect, test and keep stable as models change.
Procedures you can read and test
Plain-language procedures, tested against historical cases before release, keep behaviour stable through model upgrades.
Deterministic tools
Every system action runs through maintained integration code, built once and shared by every agent.
Versioned and promoted
Agents, procedures and tools are versioned, tested in staging and promoted, reversibly.
Trust, earned step by step
Autonomy is granted by the people the agent reports to, one task type at a time.
- Day one
Zero autonomy on day one
A new agent proposes and a person approves every action and message.
- Over time
Room to act, widened at your pace
As it proves itself, its people let it act on its own, per task and per tool, reversibly.
- Any time
Escalates when unsure
Outside its procedures or confidence, it brings the case to a person with full context, and learns from the answer.
Build your own agents by chatting with AI
Create a blank agent in the studio and talk to it.
Procedure written in conversation
Tell it the role, the queue, the rules. It drafts, you refine.
Access to a large set of tools
Maintained by Riff: system actions, domain tools, third-party services.
Shared skills and agent tools
Built for one agent, available to all.
Its own identity, with SSO
A user in its own right, with a role IT manages.
Scoped access to your integrations
Requested when needed, granted at the scope IT allows.
Its choice of model
Frontier labs or open weights, per agent and per task.
Generates its own custom tools
When a task needs one that does not exist yet.
Hooked up to channels
Email, Teams, Slack, phone, portals.
Connected to interfaces
Working the same cases as people and other agents.
What Riff is not
- Not an ERP replacementRiff sits on top of your systems of record.
- Not a general-purpose assistantAgents have specific operational roles.
- Not a workflow toolAgents have judgment and context, not just trigger-action rules.
- Not an analytics platformThe context layer serves operations, and your BI tools plug into it.
We leave capability, not dependency
Our forward-deployed team works as part of your organisation from day one, deploys the first agents, and stays as the backstop until your own capability exists. Your builders learn on real use cases with our team alongside, the platform carries the standards, and your domain experts end up owning, extending and building agents around your own strategic goals.