Your Newest Junior Staffer Doesn't Sleep (But Still Needs Managing)
Right now on my desk there is a browser tab with a half-finished prompt in it. I have been staring at it for five minutes trying to decide if I have given it enough context, or if I am about to get back something I will have to completely rewrite. This is what AI actually looks like in practice. Not a revolution. Not a threat. A tab, a prompt, and the familiar feeling of wondering whether the person you delegated to understood what you meant.
If you have ever managed junior staff, you already know most of what you need to know. The core workflow is the same: give clear instructions, provide context about the audience and the purpose, let them do the work, and then review what comes back with a generous but honest eye. The people I see struggling with AI are not lacking technical skill. They are skipping the management step. They type something vague, get something vague back, and conclude the tool is useless. But you would never hand a new hire a one-line assignment and expect a polished deliverable. The same logic applies here.
The real skill is in the course correction. Read what comes back the way you would read a first draft from someone who is smart, eager, and has been at your organization for exactly one day. It will get the tone slightly wrong. It will be confident about things it does not actually know. It will give you something structurally sound that misses the thing only you know about your community or your funder or your context. Your job is the same as it has always been: judgment, framing, relationships, and knowing what matters. That part does not automate.
Where the metaphor breaks down is where the real learning starts. A junior staffer does not come with data security risks. When you paste program data or participant information into a free AI tool, you are handing it to a company whose data practices you probably have not read. You would not give a new hire unsupervised access to your donor database on day one. Apply the same instinct here. Bias is the other piece that is genuinely new at this scale. AI reflects the structural biases of the data it trained on, which means it carries the blind spots of the entire internet, not just one person’s. And then there is automation bias, which is the subtlest risk of all: the output sounds polished and authoritative, which makes it harder to question than a rough first draft from a real person. A junior staffer’s messy draft signals “review me.” AI’s smooth draft signals “trust me.” Learning to distrust the polish is a skill you have to build.
You do need an AI strategy. That part is real. But it is not as alien as it sounds. Part of it is management skills you have been building for years: how to instruct, how to delegate, how to review, how to course correct. And part of it is genuinely new territory that deserves your full attention: data security, bias, and the discipline of questioning confident-sounding output. Start with one boring task. Manage it like you would manage a person. And take the new stuff seriously, because that is where the real work is.
Anthralytic is a strategy and evaluation studio that helps mission-driven teams clarify and amplify their impact using human expertise, data, and AI.
