Together we built an agentic AI platform running agents across the whole organisation
From individual speed to organisational scale. We were already agent-first. The platform is what made it scale across the whole organisation
“Every engineer here had Claude Code and other coding agents, so individually everyone was faster. The question was how to make that compound across the whole organisation. Where we wanted to get to was a place where humans set direction and agents do the work: agents that run around the clock, act on signals like a failing test or an alert without anyone filing a ticket, and bring us the outcome to review rather than every line. Getting there was as much about how we work as what we build, and it needed a platform rather than another tool."
We built the platform once, so every team could take an agent from idea to production in minutes
To identify where the platform could create value fastest, together with Tibber's platform team, imeto ran a pre-study assessing the capabilities needed for agentic development. This established the best setting for the pilot and future scale.
Built alongside the pilot, the platform needed to reach the team fast. It solves the shared challenges of running agents in production, including orchestration, runtime, agent identity and the MCP gateway. Teams only configure what is unique to each agent: its goals, tools and permissions.
This takes an agent from idea to production in minutes, with the platform handling the agent registry, evals, observability, the LLM gateway and feedback collection, powered by Claude models as the primary engine behind the agents running on it.
“We have agents in production across all of our engineering teams, built by our own engineers, inside the same access, review and audit controls we already had. A new one goes live almost every working day.”
Customer issues resolved within the hour
“What it means day to day is that a complex customer issue that would take days to solve now are resolved within the hour, with the analysis done before an engineer even looks at it. An on-call engineer who gets paged at night opens their laptop to find the agent’s analysis and suggested next steps already sitting there waiting for approval.”
