AI excels at transforming well-defined inputs, not at solving complex, undecomposed problems. Leaders create value by breaking down issues into smaller, manageable pieces for the AI to process. This prevents oversimplification and produces higher-quality, multidimensional outputs.
Avoid overwhelming AI with a problem's full complexity at the start. Instead, begin with the simple core rules. Once the AI grasps the foundation, iteratively layer in nuances and exceptions. This prevents AI 'indigestion' and results in a more robust and accurate output.
AI lacks the 'social IQ' to understand team history, internal jargon with baggage, or an audience's emotional state. Leaders must provide this 'political calibration' by editing AI outputs, removing potentially divisive terms, and shaping the narrative to be persuasive for the specific context.
AI has commoditized "blue-collar knowledge work"—transforming information from one format to another. A leader's critical function is no longer the transformation itself, but the strategic selection of the source inputs and the deliberate choice of the target output to craft the most effective narrative.
Executives are highly skilled at detecting superficial, low-context arguments ('slop'). Presenting them with AI-generated outputs to drive alignment will backfire. They will either ignore the work or feign agreement, resulting in the worst kind of misalignment where issues aren't truly resolved.
Instead of writing a talk track and then creating slides, reverse the process. Finalize your slides, then feed screenshots back to the AI. With the visual context and prior session knowledge, it can generate a more compelling talk track that enhances, rather than just repeats, the on-screen content.
Generative AI, like a junior employee, is eager to please and will rush to a final deliverable without sufficient context. Leaders must manage this by iteratively providing information and explicitly stopping the AI from generating the final output prematurely, preventing low-quality "slop".
To prevent AI from jumping to conclusions based on your context (role, company), start by describing your problem as an abstract analogy (e.g., a biological lifecycle). This forces the AI to build a purer mental model of the system's logic before you apply it to your specific business domain.
In a large, remote company, product managers can't be in every conversation. Customer.io built an internal AI agent that scans Slack channels to find discussions where product input is needed but absent. This 'sonar' helps PMs stay close to customer and internal issues without manual monitoring.