What if the most important question about artificial intelligence is not how intelligent the machine becomes
but how wisely the surrounding system is designed?
In this conversation with Omid of Goodfolio, we move beyond the familiar language of automation and efficiency to examine the deeper architecture of enterprise AI: the people using it, the incentives shaping it, the data feeding it, and the forms of responsibility that remain when machines begin participating in organizational decisions.
Drawing from a background in quantitative finance and systems thinking, Omid describes the “messy operational world” where AI must actually function across distributed workforces, external partners, regulatory environments, global supply chains, and relationships that extend beyond the conventional organization chart.
Here, AI is not simply a tool added to an existing workflow.
It becomes part of the system itself.
We explore why a healthy AI architecture should preserve human judgment, taste, intuition, and domain expertise while allowing machines to reduce friction elsewhere. Omid also challenges the simplistic idea of keeping a “human in the loop” when that human is reduced to an exhausted reviewer approving an endless stream of machine-generated material.
The deeper task is not merely to place a person somewhere in the process.
It is to decide where human attention creates the most meaning.
In this conversation, we explore:
* Why enterprise AI is a systems problem rather than only a technical problem
* The hidden complexity of distributed workforces and external partner networks
* Why AI adoption remains uneven across regulated industries
* What a genuinely human-centred AI architecture looks like
* The limits of using AI for financial advice and other high-stakes decisions
* How guardrails, inputs, expectations, and accountability shape AI governance
* Why “human in the loop” can create a new bottleneck
* The difference between AI as surveillance and AI as genuine support
* Why employees resist systems that appear designed to harvest and replace their knowledge
* How organizations can measure trust through sustained adoption and engagement
* Why the people affected by AI should be present from the beginning of its design
* How AI exposes missing policies, weak data, broken workflows, and institutional blind spots
* Why coordination may matter more than intelligence
* The hidden costs of rapid implementation, technical debt, security, and token usage
* What companies should preserve as automation expands: judgment, intuition, taste, and identity
AI acts as a mirror.
It reveals the processes an organization only imagined it had, the values it consistently rewards, and the distance between the system drawn on paper and the one people actually inhabit.
The technology may accelerate what is already present.
But we still retain some agency over what, precisely, we choose to accelerate.
The future of AI will not be determined by capability alone.
It will also be shaped by whether we design systems that treat people as obstacles to efficiency or as participants in the intelligence of the whole.
Get full access to thegreengage at thegreengage.substack.com/subscribe