Intention, discernment, and AI
The rise of large language models is revealing what matters
The work of the world is common as mud.
Botched, it smears the hands, crumbles to dust.
But the thing worth doing well done
has a shape that satisfies, clean and evident.
—Marge Piercy, from “To be of use”
Three weeks ago, I launched the preview phase of ArtsManaged Compass, an AI-fueled arts management thinking partner now open by invitation (with plans to go public this month).
I’m building this system in part to bridge a gap in arts management support – context-rich, field-informed, user-centered guidance at a reasonable cost. But I’m also building it as a form of action-learning – making sense of generative AI by making something with it rather than observing from a distance.
The project has left me continually amazed at the range and rising capabilities of large language models. In 100 days from concept to launch, through at least 80 working sessions, Claude Code helped execute, test, and troubleshoot a robust, responsive tech stack at speed, cost, and competence that was unimaginable a year ago.
Along the way, I discovered that AI systems are brilliant bureaucrats and technocrats. They seek and mimic patterns at extraordinary speed and complexity. As a result, they can compress or collapse many of the more tedious components of an arts manager’s workday: the clerical, analytical, pattern-heavy administrivia. But they are newbies and novices in at least two domains: intention and discernment.
Intention sets a productive direction or outcome. Discernment knows what “good” and “done” look like. So far, LLMs are crap at both.
Of course, intention and discernment have always been part of arts management practice: directing attention, naming outcomes, noticing and adapting to the gaps and opportunities as the process unfolds. But they’ve been entangled in and obscured by the execution.
This changes the dynamics of making things – at least making digital things. Complex execution used to be the bottleneck. Now it’s increasingly fast and cheap. The new bottleneck is knowing and naming things worth making, and ensuring they’re made well.
From the ArtsManaged Field Guide
Function of the Week: People Operations
People Operations involves designing and driving systems and practices that attract, engage, retain, and develop people within the enterprise (also called human resources).
Framework of the Week: Ladder of Control
The Ladder of Control is a communications tool for supervisors and their direct reports to help calibrate reporting relationship across different kinds of work. The ladder offers seven levels of authority, from the least agency ("Tell me what to do…") to the most agency ("I've been doing…").
Photo by Nathan Dumlao on Unsplash