I've kept busy over the last few months building what I hope will become a valuable platform. While we are still a few weeks away from a larger reveal, the signs are promising and progress has been swift.

At the core of this solution is an agentic framework: an overarching service layer designed as a dedicated, highly optimised, and extensible expert system. Its aim is to transform data into actionable insight and generate knowledge and action plans that improve outcomes.

Agentic frameworks have been a major focus of mine for some time. In recent months they have started to emerge not only as abstract patterns, but as tangible products. As these frameworks develop, I keep seeing design choices that often hit the target, but still miss the bullseye of what agentic systems can truly deliver.

Three Principles for Strong Agentic Frameworks

  1. Context. Frameworks must be able to share and enrich context. They should learn from their actions and operate not only as data receivers, but as data transmitters, discovering and integrating new data sources as they work. Context is not static; it evolves, and it should inform decision-making at every level of the framework.
  2. Authority. Agents within these frameworks must provide genuine expertise and domain-specific knowledge. Relying only on generic large language model agents running sequential transactional tasks is not enough. Each agent should be purpose-built, with predefined and finely tuned capabilities. That enables better routing, more authentic responses, and higher confidence in execution.
  3. Outcomes. Outcomes should drive the framework as both inputs and outputs. The steps required to reach those outcomes should be generated dynamically, always aiming for the most efficient and effective path. Orchestration is critical: a controller or orchestration agent should continuously validate whether outcomes are being achieved and optimise the workflow in real time.

These three principles are not independent. Context informs authority by showing agents what domain they are operating in. Authority shapes outcomes by ensuring actions are grounded in genuine expertise rather than generic inference. And outcomes feed back into context, giving the system a learning loop it can use to improve over time. Getting all three right simultaneously is where most frameworks currently fall short, and where the real design work lies.