Chat with Alba on
Instant answers, wherever you message.
Platform · Agents
Personal agents for every role.
Each agent combines a defined working role with selected knowledge and services. The channel is the entry point; the configuration decides the scope of assistance.
The digital twin
An agent shaped around your role.
A digital twin of working context, responsibilities and available capabilities.
A personal agent becomes useful when it knows what it is there to help with. In Smart Mind, the digital twin is a configured working companion: it has an identity, a role, goals, assigned knowledge spaces and available actions. These elements give the agent a practical frame for interpreting a request. They also give the organization a clear way to decide what support each employee should receive.
Personalization with a shared foundation
Two employees can ask similar questions and still need different answers. A sales colleague might need an explanation suited to a customer conversation, while an operations colleague needs the internal procedure behind it. Their agents can draw on different Spaces and use different services. The platform remains shared, but the working context is specific.
A role that can be described
The quality of an agent starts with an understandable brief. Who does it assist? Which outcomes should it support? Which knowledge is relevant? What actions may it invoke? Answering these questions makes the digital twin concrete, and as a role changes, its knowledge and action assignments can be reviewed.
Personalization model
Identity and role
Goals and knowledge
Permitted actions
Agent design
A clear definition. A useful working scope.
Five configuration choices connect the agent to the organization.
An agent definition is the link between a general AI capability and a particular business responsibility. Identity establishes who the agent represents. Role describes the kind of assistance it should provide. Goals explain the outcomes it should help achieve. Knowledge spaces identify the information it can draw upon, while available actions describe the services it can use. Together, these choices establish a working scope that can be explained, tested and refined.
Configure meaning before mechanics
A useful role description goes beyond a department name. It explains the requests the agent should address, the audience it serves and the form of assistance expected. A goal such as preparing a clear policy-based response gives more direction than a broad instruction to be helpful. Assigned Spaces provide context; available actions add specific capabilities.
Keep assignments intentional
Access to a shared platform does not mean that every agent needs every source or service. A focused configuration is easier to understand and evaluate. This also helps separate a public-facing agent from an employee assistant: both use the same architectural model while working with different information and responsibilities.
The agent brief model
Who it assists
What it should achieve
What it may use
Reasoning and execution
From a request to a useful result.
The planner organizes the work. The executor invokes the selected actions.
A message gives the agent a starting point. The agent interprets that request in relation to its role and goal, considers its assigned knowledge and determines which available capabilities are relevant. The planner organizes the required work into actions and parameters. The executor then carries out the authorized invocations and returns their results. This separation connects conversational intent to specific services without asking the language model to invent an implementation.
Plan around the actual task
Consider an employee asking for a response based on an internal procedure. A useful plan may first retrieve the relevant knowledge, then use a drafting capability with the request and supporting context. An incomplete request may require clarification first, and a request outside the assigned role needs a clear boundary.
Use the declared capabilities
Execution connects the plan to the services assigned to the agent. The agent selects from available actions; the service layer handles how those actions are performed. Human review remains part of the workflow wherever the output requires an employee’s judgment or acceptance.