You run a six-person nonprofit. The training line in the budget is a rounding error, or it does not exist at all. Staff need to get more capable with the tools you already pay for, and you have no money to send anyone on a course.

The first thing to notice is that “we have no training budget” assumes training has to be a purchased event. For a small team using mainstream tools, the formal curriculum problem is mostly already solved by the vendors themselves. The binding constraint sits somewhere else: protected time to practise, and one person who shares what they have figured out so the same lesson does not get learned six times.

The training your team needs is mostly already paid for. It came bundled with the tools. What you need to add is a 20-minute calendar block and a name.

The training-budget problem is real. The fix sits elsewhere

Most small nonprofits have no training budget worth the name, and adding one is rarely the right move.

Bridgespan, citing NTEN, reports that only 1% of nonprofit technology budgets goes to training. The same article finds that more than half of nonprofit leaders say staff lack the expertise to use or even learn about AI. The funding gap is real, structural, and not about to close.

Adding a training line will not fix it. GivingTuesday’s AI Readiness Survey shows that the strongest predictor of AI readiness is when an organisation acquires dedicated technical personnel, typically around 15 staff. Below that threshold, tech ownership is informal by default. There is no IT director to hand a training plan to. The plan, whatever it is, will be carried out by someone whose primary job is something else.

That changes what “training” usefully means at your size. The work is not buying curriculum. It is creating the conditions in which the curriculum that already exists actually gets used.

What free training is already attached to your stack

Before buying anything, exhaust the free learning hub for the tools you already pay for.

If you are on Microsoft 365, Microsoft Learn for Nonprofits publishes free, structured paths covering Microsoft 365, Copilot, Power Platform, fundraising and volunteer management. The Introduction to AI Skills for Nonprofits path is the obvious starting point if your team has Copilot in their licences and is not yet using it.

If you are on Google Workspace, Google.org runs Get Time Back with AI, a free resource on using Gemini for grant writing, fundraising and event planning. The broader Google Workspace Learning Center covers the rest of the suite.

TechSoup Courses offers nonprofit-specific seminars (30 to 60 minutes) and longer 100/200/300-level courses developed with subject-matter experts. Access tiers vary by course, but a meaningful share is free or low-cost for TechSoup-eligible orgs.

Mailchimp Academy offers free interactive courses and certifications in email marketing, automation and the platform itself. Joining the free Mailchimp & Co programme opens the catalogue. Canva runs a similar free school for design and brand kits.

Two practical points. First, none of this is hidden. It is one search away from any staff member who looks. The reason it is under-used has nothing to do with access; the missing pieces are permission and time. Second, paid external nonprofit training often duplicates what the vendor already offers free. Treat the vendor hub as the default and external paid courses as the exception, not the other way round.

Why 20 minutes a week beats a one-day workshop

Short, repeated, protected practice on real work outperforms an annual training event, especially for a small team.

The cadence: short, repeated, protected

A JMIR meta-analysis found that spaced digital education outperformed massed delivery for clinical behaviour change, with knowledge gains maintained up to two years after the trial. The evidence is strongest in clinical training, but the underlying mechanism (spacing beats cramming) is well-established across adult learning.

Twenty minutes weekly is a heuristic, not a study-derived figure. It is short enough that it cannot legitimately be cancelled for “we are too busy this week,” and long enough to actually work through one lesson or one workflow. Plan for a quarter, not a week. Real proficiency with a new tool builds across months of repeated use on real tasks, not in a single sitting.

The manager move: visible endorsement, not delivery

Gallup found that employees whose manager visibly supports AI use are 8.8 times more likely to say AI gives them more opportunities to do their best work. Manager visibility, not training delivery, is the lever.

For a six-person org, the ED is the manager. You do not need to run the sessions. You need to put the calendar block on the schedule yourself, name it (“Practice block, Tools”), and protect it the same way you protect a board call. If the block gets quietly eaten by something else two weeks running, the message staff receive is that practice is the lowest-priority thing on the list.

What to actually practise: workflows, not features

Pick the recurring task. The board paper that takes a day. The donor thank-you that backs up. The grant draft that eats Sundays. Practise on real work, not toy exercises. The vendor learning hub supplies the technique. Your own workflow supplies the context. A staff member who uses a Copilot lesson to draft next month’s board paper has done two things in 20 minutes. A staff member who watches the same lesson in the abstract has done one.

How to pick the one person who shares what they learn

The designated sharer is a role with a calendar slot and a name attached. Not a job title and not a lunch-and-learn.

Pick by curiosity, not seniority. The right person is whoever already pokes at new features without being asked. They get a 30-minute monthly slot in a team meeting, not a separate event, to walk through one thing they have figured out. Rotate annually so it does not become a permanent burden, and so the role does not calcify into “the tech person.”

Protect their time. The 30-minute slot is the visible part. The unseen part is the hour or so they spend a month preparing it. Take an hour off something else. Do not invent a title. Do not write a job description. Just name them in the meeting agenda and refer to the slot by their name for the year they hold it.

Lunch-and-learns fail because attendance is optional and the slot is owned by no-one. The designated sharer slot is owned, scheduled, and short. If your team has tried lunch-and-learns and watched them die, this is the version that works at your size.

What to do when staff are using AI tools you have not approved

Assume staff are already using AI tools on personal accounts. Frame the next conversation as learning, not enforcement.

WalkMe and SAP report that 78% of employees admit to using AI tools their employer has not approved, while only 7.5% have received extensive AI training and 23% have received none. Shadow AI is the baseline. The training vacuum is the cause.

The instinct to clamp down with a policy backfires for two reasons. Staff who have been quietly using ChatGPT to draft donor emails will stop telling you. You lose the chance to teach them what to keep out of the prompt, such as donor names, financial figures, and beneficiary case notes, and what is fine. You also lose the chance to learn what is genuinely useful from the people doing the actual work.

A light-touch policy that names the rule (no personal or beneficiary data into a free AI tool) and invites disclosure (“tell me what you are using and we will work out together what stays in”) gets you further than a ban. The point of the policy is to make it safe to talk about.

How to know it is working without a formal evaluation

Three signals tell you the practice block is paying off. None of them require a survey.

Watch for the “how do I” interruptions to drop. When a staff member who used to ask the same question every fortnight stops asking, they have either solved it or given up. Walk past their desk. If the workflow has changed, they have solved it.

Watch for the same workflow showing up across different staff. If the comms coordinator’s prompt for shortening board updates starts appearing in the operations manager’s drafts, the designated sharer slot is working. The lesson is travelling without you having to push it.

Watch for the recurring artefact getting faster. The board paper that took a day in January should take half a day by April. If it still takes a day, something in the practice cadence is broken, probably the protection of the block, or the choice of what to practise.

If none of those three are moving after a quarter, stop and re-diagnose. The problem is rarely the curriculum. Usually the calendar block is being eaten by something else, the designated sharer slot has not actually happened, or the workload underneath is so high that 20 minutes a week is genuinely unavailable.

The board sentence at the end of the year, if this works, is short. We are using a 20-minute weekly practice block and a rotating internal sharer instead of paid training in 2026. That is the whole programme.

Frequently asked questions

What if no one on my team is willing to be the designated sharer?

Make the slot smaller before you make the role formal. Ask whoever already pokes at new features to walk through one thing for ten minutes at a regular team meeting, with no preparation expected. If a willing person genuinely does not exist, the role rotates fortnightly and everyone takes a turn covering one workflow they know well.

Are TechSoup courses actually any good?

The short seminars are uneven and the longer courses are stronger. The catalogue is worth scanning when you have a specific topic, less worth treating as a curriculum to work through. For most small nonprofits, the vendor learning hub for the tools you already pay for will outperform a TechSoup course on the same topic.

Should we pay for ChatGPT Plus or Microsoft Copilot if we want staff to learn AI?

Probably not as a training expense. If the rest of your team is on Microsoft 365 with Copilot already in the licence, that is the cheapest path to learning. Pay for ChatGPT Plus for one person who is doing the most AI-adjacent work, then let them share what they learn through the designated sharer slot. Avoid licensing the whole team early.

How do I justify protected practice time to a board that wants every hour billable to programme?

Reframe the time as preventative. Twenty minutes a week against software you already pay for protects the spend by making it actually used. Bring the board the artefact that gets faster (board paper, grant draft, donor email) and the time saved on the recurring task. The numbers there are usually larger than the protected practice time.

When does it make sense to actually buy external training?

When you have a specific job-relevant gap that no vendor hub covers, when a regulatory or compliance requirement names a course, or when you are about to make a tool decision and need an unbiased outside perspective. Buying generic upskilling for a small team rarely pays back. Buying a targeted course for the one person whose job will change most usually does.