A teacher asks for a lesson plan and receives a plausible sequence of activities. That is a real capability. But the teacher still has to check the subject matter, adjust the language, fit the lesson to the available time, remove inappropriate assumptions, and decide how to tell whether students understood.
For someone preparing a research brief, the unfinished work might be tracing the sources. For an independent builder, it might be distinguishing encouraging feedback from a buying commitment. The generated output is not the whole job.
What do you need beyond access to AI?
There is evidence of both adoption and unfinished organizational work. McKinsey’s August 2026 survey reports that 80% of respondents saw improved individual productivity from AI, while 37% reported an enterprise-level EBIT impact. Its high-performing group was more likely to report workflow redesign. These are self-reported associations, not proof that a particular design intervention causes returns. Read the McKinsey survey.
In education, Gallup reported in May 2026 that 18% of U.S. public K–12 teachers had received formal guidance from their school or district on using AI in teaching. That finding is about a specific U.S. sample, not all teachers or countries. It nevertheless points to an important distinction: having a tool available is not the same as having an agreed way to use it. Read Gallup’s findings and methodology.
Look for the work around the prompt.
We believe useful products often address that surrounding work. They help someone frame an objective, assemble relevant context, choose a sequence, inspect a result, and carry a decision forward.
This does not mean every task needs a complex agent or a new platform. A well-designed document can sometimes make a workflow more usable. In other cases, persistent project context, source links, tool integration, or feedback require software. The form should follow the job.
Ask three questions before adding AI.
- What is an acceptable result? Define it in the user’s terms, including constraints and quality criteria.
- What must the person still do? Watch the checking, editing, copying, coordination, and decisions—not only the generation.
- What would demonstrate improvement? Consider the effort to reach an acceptable outcome and the person’s ability to judge it. More output is not automatically more value.
Explore a teaching example.
Teacher Copilot begins with recurring teaching work. It does not claim to automate teaching or guarantee a better lesson. It supplies a guided routine, editable materials, and explicit review questions, so a teacher can inspect and adapt a first draft.
That is one expression of a broader principle: start with the outcome, understand the conditions, and build the support that makes the work possible.