Most impact teams and organizations are under pressure to figure out AI, but few have cracked how to drive widespread and deep adoption across their teams. The barriers are rarely about the technology itself — more often it’s staff hesitation, unclear use cases, or a lack of practical training that stalls progress.

On November 4, leaders from across the NationSwell community came together to surface the approaches that are working best — from training and change management to addressing staff hesitation — so leaders can move past experimentation and toward practical use.

Some of the most salient takeaways from the event appear below:


Key takeaways:

Start with a bottleneck the team can name. Ask the people who do the work where it routinely gets stuck, whether that’s pulling data for a monthly report or waiting for colleagues across time zones to approve a proposal. Ranking those bottlenecks by the value of solving them and the effort required can help teams decide where to start.

Measure the work before changing it. Record how long a task takes, or how quickly the team can respond before introducing AI. A baseline makes it easier to measure whether the new process saves time once you account for reviewing outputs and maintaining the tool.

Test a workflow before connecting it to core systems. A grant due diligence pilot began as a shared prompt that staff used to document what they look for when assessing financial or reputational risk. Testing and refining it separately gave the team time to see where it helped before integrating it into its grants management system.

Give staff time to experiment and share what works. Peer-led coffee chats can offer colleagues a place to demo a workflow others can reuse. Meeting-free blocks or weeks can give people room to try something that might take several rounds to get right, especially when experimentation would otherwise compete with their usual workload.

Decide where human judgment enters the process. AI can check whether applications are complete or pull together a first pass at due diligence, but for consequential decisions, build in a review step and make clear who is responsible for weighing the evidence and making the call.

Keep a trail back to the source. When AI helps summarize or check data, a colleague should be able to trace a claim back to the original dataset, method, and citation. One approach is to give staff human-verified data summaries with guidance on how to cite and use the findings in reports or grant materials.

Review AI-assisted work before passing it along. A quick read can miss a brief that’s technically accurate but buries the point, or an analysis that lacks the context a colleague needs to act on it. Before sharing, ask whether you understand the work well enough to explain it and would feel comfortable putting your name on it.

Use adoption as a chance to develop staff judgment. Giving an early-career colleague ownership of an AI agent can create a leadership opportunity. They learn to define what good output looks like, spot where the agent falls short, and decide how it needs to change as the team’s work evolves.