They wanted agents everywhere. First they wanted to know what could go wrong.
ASICS is running proofs of concept with AI agents and retrieval augmented generation, and expects AI to become part of how their people work. Before scaling any of it, they wanted an environment where agents run securely, where the bill holds no surprises, and where there is visibility into what the agents are doing.
A small internal team had built that environment themselves. They were not sure they had the experience to see every pitfall coming. Budget overruns, data leaks, access and authorisation, observability. All of it sat with one team.
So they asked Brighting for an outside read on the platform they had, and on the capabilities it would need before agents scale.
ONE FRAMEWORK, 42 CAPABILITIES, TWO WEEKS
Brighting has a capability framework for scaling agentic AI at enterprises. It covers 42 capabilities across agent discoverability, autonomy, intelligence and trust. We matched every one of them against the platform ASICS was building, together with their team.
Score the maturity
Every capability got a maturity level based on what was actually in the platform, not on what was planned.
Name the impact
For each gap we described what an immature capability costs once agents run at scale.
Verify in the code
A three-week deep dive followed: code review, documentation review, and proofs of concept to test the findings.
Propose the mitigation
Each gap came with a target maturity and a concrete way to close it.
Four domains carried most of the risk
Agent discoverability
Knowing which agents exist, what they are allowed to touch, and who owns them.
Agent autonomy
How much an agent may decide on its own, and where a human stays in the loop.
Agent intelligence
The models, context and retrieval an agent uses, and how its output is evaluated.
Agent trust
Access and authorisation, guardrails, cost control and observability across every agent.
Two documents, two audiences
Management presentation
The state of the platform in a form the management team can act on. Where it works, where it is ready to scale, and what is missing.
Capability breakdown
All 42 capabilities, one by one: current state, expected state, position on the maturity scale, and the proposed way to close the gap.
Our own framework, and the runway behind it
The learnings that come with a fast developing space such as AI are all processed and proven in our Agentic Runway. Using it will save any organization months of work for a fraction of the cost.
Capability Framework for Scaling Agentic AI
42 capabilities, a maturity scale per capability, and a mitigation for every gap.
Brighting Agentic Runway
The platform we are using to take ASICS from the assessed state to the target state.
More of our work
All cases →
Building your own agent platform?
We score it against 42 capabilities and tell you what to fix before you scale. Two weeks, with your team in the room.