Jobs & dignity.

The augmentation-first staffing plan

One page, written before you buy anything, that answers the only staffing question that matters: when AI saves this person time, where does the time go — and who decided?

Your staff have heard the headline version: AI is coming for jobs. Here is what the record actually shows so far. In McKinsey's global AI survey, 32% of respondents in 2025 expected AI to reduce head count within a year; a year later, only 14% reported an actual reduction, and two-thirds reported little or no AI-related change in total employment. Expectation ran more than twice as fast as reality. That is not a promise it stays that way — 39% of respondents now expect reductions in the coming year — but it does mean you have time to decide on purpose instead of by panic. This playbook is how a small team uses that time.

Time to run it: about 90 minutes for a team of up to ten, once. Cost: $0. Review: 30 minutes, quarterly.

Inventory tasks, not jobs

For each role, list the week's actual tasks in plain words: drafting thank-you letters, reconciling the donation spreadsheet, answering the same five intake questions, sitting with a family through a benefits application. Jobs are not automated; tasks are. A role is a bundle, and most bundles mix routine work with judgment work. You cannot make a dignity-preserving decision about a bundle you have never unpacked.

Sort every task into three columns

• Augment: AI drafts, sorts, or summarizes; the person reviews and decides. (First drafts, meeting notes, schedule conflict checks.)
• Automate with a human check: routine and verifiable, with a person sampling the output. (Reminder emails, data entry between systems you already use.)
• Never delegate: the work where a person is the point. (The hard conversation, the judgment call on a family's case, anything a donor or client would feel betrayed to learn a machine handled alone.)

If a task does not fit a column, it stays human by default. The burden of proof sits on delegation, not on the person.

Write the augmentation plan, one per person

For each staff member, one short paragraph, written with them, not about them: the task AI will help with, the tool, the training they will get, and — this is the step most organizations skip — where the saved time goes. Named in advance: more client hours, the backlog that never gets touched, training, or a saner week. Time saved with no named destination does not vanish; it gets quietly harvested, and staff can tell the difference.

This step exists because the sector's own data says it is the missing one. In the 2026 State of Nonprofit AI report from NTEN and Bridgespan (917 nonprofit workers surveyed, late April to mid-June 2026), 98% reported using AI in some capacity — but 37% said their organization provides no staff training on the tools, and 58% reported no AI roadmap at all. Use is nearly universal; support is not. A one-page plan per person closes exactly that gap.

Decide the dividend in writing

One sentence, signed by the executive director, kept with the plans: "Time AI saves in this organization goes first to [mission work / training / workload relief], and no staffing change is made on AI grounds without [the review in Step 5]." We are not asking you to promise no job will ever change — no honest leader can. We are asking you to decide, in advance and in writing, who shares the productivity gain. That decision, made calmly, is the whole difference between augmentation and slow-motion replacement.

The 30-day check

Sit with each person and ask three questions: Did the tool actually save the time? Did the saved time go where the plan said? Did any "never delegate" line get crossed? Adjust the plan, not the person. If a tool saved nothing, drop it and say so — a failed tool is a purchasing lesson, not a staff failure.

The one action: this week, run Step 1 and Step 2 for a single role — yours. Unpack your own bundle first. Leaders who sort their own tasks before anyone else's get believed when they say the plan is about dignity, not head count.

The steelman for the skeptics: "This is a layoff plan with better manners." It can become one — any plan can be abused. The difference is mechanical, not rhetorical: the dividend is decided in writing before the savings appear, the "never delegate" column is explicit, and the person whose job it is co-writes the plan and sits in the 30-day check. If your organization will not do those three things, do not call what it is doing augmentation.

Published Friday, October 16, 2026. The Issue #6 newsletter edition wraps this playbook and lands in inboxes separately — browse the newsletter archive. This page is the playbook itself, frozen as published: playbooks never change after they go up. If a better way appears, it gets a new number, not a rewrite.

Free forever. One playbook every Friday.

Back to the shelf · ← Playbook No. 05 — Volunteers, scheduled · Playbook No. 07 — Boards, briefed →