Before You Automate a Process, Prove You Understand It
Automation multiplies encoded assumptions. Observe real work, test the account and choose the smallest useful intervention.
Read guideBrilliant Quest / insights
Brilliant Quest studies how people, processes, information and systems interact inside organisations. These field notes explain recurring patterns we believe are useful when improving operations or introducing AI.
11 entries
Automation multiplies encoded assumptions. Observe real work, test the account and choose the smallest useful intervention.
Read guideHuman Middleware is often an architectural signal, not a people problem. Separate repetitive bridging from the judgement you need to preserve.
Read guideA diagram is a representation. Connect claims to evidence and keep disagreement visible before treating the map as an operating model.
Read guideChoose deterministic rules, AI interpretation and human judgement by the nature of the task, not by the novelty of the tool.
Read guideBusiness identity is more than a label. Establish relationships between records before letting automation or agents act across systems.
Read guideA person repeatedly transfers, reformats, interprets or reconciles information between systems that cannot communicate effectively.
Read patternResponsibility or information moves between people or teams without an explicit, observable acceptance event.
Read patternThe same real-world business object is represented by different names or identifiers across systems.
Read patternSeveral systems or documents appear to contain the authoritative version of the same information.
Read patternA process works because a small number of people understand contextual rules that are not represented elsewhere.
Read patternA simple process accumulates special cases and workarounds until nobody understands the whole system.
Read patternBring one outcome and explore what may be getting in the way.