A-Insinöörit wanted to make AI part of everyday work across the organisation, with AI agents working alongside employees in a variety of tasks. The aim was to manage agents like people: each agent would have a clearly defined role and responsibilities, capabilities and work experience. Together with A-Insinöörit, Arked designed an AI agent governance model and architecture, as well as a roadmap for implementation.
Challenge: Agents Were Being Created Quickly, but There Was No Governance Model
A-Insinöörit already had several AI agents in production, and the aim was for project teams to systematically incorporate agents. Agent development was to be decentralised across teams, creating a need for shared operating practices: a registry and designated owners for agents, boundaries for the data and tools agents are permitted to use, and visibility into agent quality and reliability.
In the engineering and design sector, professional responsibility always rests with a designated individual. When an agent supports design or proposal work, the person responsible must also be able to clearly see which content was produced by the agent and what information it is based on.
Solution: Managing an Agent Like an Employee
The governance model was based on the principle that an agent is a member of the workplace community and should be managed like an employee. Each agent was assigned a designated owner, a job role and a lifecycle from recruitment to retirement. This approach is familiar from everyday management, enabling the business to manage agents without technical terminology. The model specifies that agents and their owners are recorded in a central agent registry, that agents are approved before release, and that their quality is regularly monitored and improved. Access rights are role-based, ensuring that agents are used securely.
Another key requirement for A-Insinöörit was that saved content must indicate whether it was created by a person or an agent. Arked designed a solution in which any document saved by an agent is automatically labelled as AI-generated content. This makes it clear in every case which material was produced by AI and which by a person.
Outcome: A Complete Governance Model and Architecture in Four Weeks
The project was delivered by a two-person team from Arked in collaboration with A-Insinöörit’s management and experts. The result was a complete framework for agent governance. The governance model defines the rules for deploying and maintaining agents, as well as the new roles and responsibilities required. The target architecture describes how the governance model will be implemented technically. The roadmap sets out the order of implementation so that tangible benefits can be achieved as early as possible. It also includes schedule and workload estimates for budgeting purposes.
Client Feedback
“I have carried out similar work with several leading consulting firms, and I was very pleasantly surprised by Arked’s work.”
Max Levander / CTO, A-Insinöörit
