Abstain, amplify, or automate. Sort any piece of work in five questions, starting with the only category most AI roadmaps leave out.
Plenty of models sort work into what AI should automate and what it should assist. Far fewer name the work it should stay out of entirely. That line is where the credibility of everything else comes from, so this model draws it first.
Name a task, answer five questions about it, and add it to the board. Do this for eight or ten tasks and you have the beginnings of a roadmap your leadership team can actually argue with.
Runs entirely in your browser. Nothing is sent anywhere.
Name the task and answer all five to add it.
Work where the stakes, the ambiguity, or the human relationship mean AI should not be involved at all. Deciding this in advance, in writing, is what makes the other two categories trustworthy. An organization that has never named its abstentions has not made a decision about AI; it has simply not hit the consequences yet.
A person’s discernment is the point of the work. AI expands reach and speed, and the human stays in the loop by design rather than as a formality. Most professional work belongs here and gets misfiled into automate.
The outcome is deterministic and the thinking happened when the rule was written. Hand it over completely and stop spending attention on it. The failure mode here is supervising something that does not need supervision.
Sorting work into what AI automates and what it assists is common ground, and several similar taxonomies exist. The abstention category is the part worth arguing about. It is also the part that turns an AI roadmap into a position rather than a purchase order.
This model is one of three on the Frameworks page. Use it, cite it, or argue with it.
You can fill this board in twenty minutes on your own. The hard part is the meeting where four people hold four different views about what should stay human, and nobody has the authority or the framing to settle it. That conversation is most of what I am hired for.
I facilitate the version of this with leadership teams, and the output is a written set of abstentions your organization can actually stand behind.
Human-led. Strategy, design, and editorial standard are Lexi Gilbert’s. Delivery, including imagery, is supported by AI systems of her own architecture and, where applicable, by human collaborators. Authorship and final judgement remain hers.
What I am seeing in the work across AI, visibility, and leadership. Written by me, sent when there is something worth saying.
Check your inbox for a confirmation email. If it isn’t there shortly, look in Promotions or Spam.
Change how this page reads. Your choices are remembered on this device and apply to every page.
Stored on your device. Nothing is sent anywhere.