Context: A Small Organization at a Reset Point

 

World Salmon Council does something specific and hard to replicate: it gets kids to rivers. Volunteers, many of them retired science teachers, run field stations where students test water quality, study macroinvertebrates, and watch salmon spawn. The impact is outsized relative to the organization's size. The structure is not.

 

At the time of the engagement, the organization had roughly one and a half employees and a working board carrying real operational load. A new executive director had recently stepped in — three years out of college, in her first leadership role, following a period of transition. Funding was grant-dependent, and much of it reimbursement-based, which meant spending money before receiving it. The board had identified diversification toward recurring individual donors as the long-term answer, but that plan had been sitting for years without the capacity to execute it.

 

The board also carried a specific frustration. A previous consulting engagement had produced a thoughtful set of strategic assets, but nobody had the bandwidth to implement them. When AI came up, several board members said a version of the same thing: we cannot afford to end up with another beautiful deck.

 

 

The Real Obstacle Was Not Skill

 

My conversations before the workshop surfaced three distinct hesitations, and none of them was that the technology looked hard to use.

 

The first was capacity. As one board member put it, the organization was so lean it did not feel it had the luxury to experiment. Overwhelm was a live concern, not a theoretical one. Adding a demanding new initiative to a one-person staff carried real risk.

 

The second was dependency. Another board member asked it directly: if we build our operations around your expertise, what happens when you are gone? For an organization with no technical staff, a consultant-shaped hole is a structural risk, not a hypothetical.

 

The third was values, and it ran deepest. This is an organization that exists to protect cold, clean water. AI infrastructure consumes substantial energy and water. During the second session, the executive director described researching a proposed data center in the Columbia River Gorge that would draw cooling water from a tributary with an active salmon run. Her summary of the tension was blunt: her own success as a leader using these tools makes the ocean hotter. That is not a communications problem. It is a real conflict, and it deserved a real answer.

 

 

The Work

 

Redesigning the format around the constraint

The engagement was originally scoped as a single half-day leadership workshop. After my alignment conversation with board leadership, I concluded that format was wrong for this organization, and it was wrong for a reason worth naming.

 

A board that has just told you it does not have the luxury to experiment will not respond well to a four-hour block. And one board member had asked the sharpest question in the whole engagement: what happens if we workshop an idea for several hours and then never get trained on step two? If the session ends in philosophical alignment with no concrete next move, and there is no internal resource to carry it forward, what was the point?

 

I split the half-day into two 90-minute sessions two weeks apart, with a required experiment in between. That structure did three things a single workshop could not. It kept the time ask small enough that a stretched board would actually show up. It made step two part of the design rather than an afterthought, because every participant had to try something and report back. And it gave the second session real material to work with — people were making decisions about AI based on their own two weeks of experience, not on a facilitator's description of what the technology can do.

 

Attendance was voluntary throughout, and I invited people to come to one session if they could not make both. Nobody was compelled into the room, which mattered given how the conversation about the technology needed to go.

 

Starting by taking the objection seriously

I opened the first session with an explicit statement: this is not a pitch for AI. Nobody was being asked to commit to anything, and the organization — not the facilitator — would decide if, where, and how AI fit into its work.

 

I followed that immediately by naming the environmental concern out loud, before anyone in the room had to raise it. I acknowledged it as legitimate, unresolved, and entirely appropriate to hold at the same time as exploring the technology.

 

This matters more than it sounds. When a facilitator arrives with a tool and treats the room's objection as resistance to be managed, the room stops talking. Naming the tension first is what made the rest of the conversation honest.

 

Building shared literacy without condescension

I opened with a poll. About half the board used AI occasionally; the rest were split between never having touched it and using it regularly and intentionally. That spread is the actual challenge in most small organizations. Pitch to the middle and you lose both ends.

 

I covered what generative AI does well, where it fails, and why. The framing I used throughout: treat it like a bright, fast-learning, extremely well-read intern with no life experience and no judgment. It produces impressive first drafts. It can be confidently and completely wrong. It does not know what context you failed to give it. It needs a human reviewer, every time.

 

I explained hallucination mechanically rather than offering it as a caveat — why a predictive system generates plausible fabrications, and what that means for deciding when output can be trusted and when it has to be checked.

 

Showing the difference instead of describing it

Rather than explain prompt quality, I demonstrated it live, using the organization's own problem.

 

My first prompt was a single sentence: identify grants World Salmon Council could apply for in environmental education. The result looked reasonable — a broad list of federal and foundation opportunities.

 

My second prompt included background documents board members had reviewed in advance, the organization's most recent annual report, scoring criteria for evaluating fit, context on staff size and budget, an explanation that grants were becoming a smaller share of a shifting funding strategy, and a request to rank the results and flag its own assumptions.

 

The output differed in kind rather than degree: six opportunities instead of dozens, scored and ranked, with the top recommendation a regional funder the generic search had missed entirely.

 

The demonstration landed because the board could see exactly what changed for them. It also carried the more important lesson: you do not need to write a perfect prompt. You need to keep talking to it.

 

Experimentation with a deadline and a reporting requirement

Every participant left Session 1 with the same assignment: try one AI-assisted task connected to the organization's work, try it in more than one tool if possible, and come back prepared to say what surprised you. I framed it deliberately — this was not about getting it right, it was about finding out how it feels.

 

Between sessions, I held a one-on-one working session with the executive director focused on her actual workflow. The most useful moment was diagnostic. She had approached AI with a rigid set of instructions about how to build a budget model, and was frustrated with what came back. I suggested she bring it the problem rather than the prescribed solution: describe the outcome, ask it to ask questions, work it through collaboratively.

 

Making the decision an explicit agenda item

I didn’t assume the answer in Session 2. I put four questions to the group:

 

• Does using AI feel like the right thing for this organization?

• What would need to be true for it to be genuinely aligned with the mission?

• Is anything enough of a barrier to stop, and what do we need to keep watching?

• What would it look like to say yes responsibly — and what would a no look like?

 

The executive director singled out the last one. She wanted to say yes. She wanted to be responsible about it. She did not yet know what that meant, and said so out loud.

 

Before that discussion, I delivered on a commitment I made in Session 1: independent research on reducing the environmental footprint of AI use. The findings were practical and specific. Concise prompting reduces compute. Smaller models are sufficient for simpler tasks. Batching related requests is more efficient than sequential ones. The energy mix powering data centers shifts through the day, so scheduling non-urgent AI work for the afternoon draws on more solar. I compared providers on emissions transparency and infrastructure efficiency. I flagged carbon offsets as a line item the organization could write into grant budgets.

 

That research did more than answer a question. It demonstrated that the objection had been heard as a real constraint to design around rather than an obstacle to talk past.

 

Reframing the values tension as a reason to engage

The harder half of the answer was strategic. An environmental organization that abstains from AI does not slow AI down. It removes its own perspective from the rooms where the technology's rules are being written — the same dynamic that has left other populations underrepresented in technologies designed without them.

 

There was also an immediate, local version of this. Oregon is actively debating data center siting, including projects with direct implications for river temperature and salmon runs. An executive director who arrives at that conversation as an informed, active user of AI arguing for responsible siting is far harder to dismiss than one who arrives opposed to the technology outright. Handled well, the mission tension becomes a source of credibility rather than a liability.

 

Guardrails defined by the board, not the consultant

The board did not vote yes and move on. It named what would have to be true first. Working through the discussion, four guardrail categories emerged:

 

• Environmental impact and energy use

• Cognitive dependence — the risk of outsourcing thinking rather than supporting it

• Security, privacy, and whether organizational data trains a vendor's model

• A quarterly review cadence, on the explicit principle that this is not a set-and-forget decision

 

Ownership was assigned deliberately. The executive director owns the organization's AI position and reports on it to the board. Two board members were paired as her advisors: the board's most active AI user, and a member who had deliberately not rushed to adopt. The counterbalance was the point. Someone deep in the technology is not well positioned to notice what it may be costing.

 

An ownership model built for a working board

I closed with an exercise that addressed the structural gap underneath everything else: one staff member cannot carry five priorities alone, and a volunteer board cannot help if nobody knows who to ask.

 

Five work streams were defined, each paired with named board members: AI strategy and guardrails; the shift to a recurring donor strategy; grant research and applications; budgeting and reporting; and volunteer engagement.

 

One board member hesitated to sign up, saying he had only just used AI for the first time and did not know what he was doing. I clarified that he was not being asked to be an AI expert. He was being asked to be an experienced human being the executive director could turn to for judgment. He signed up.

 

The final step was left deliberately open. The executive director and each advisor would define together what involvement actually means. Some would want standing working sessions. Some would want to field an occasional call. Both are legitimate. What is not workable is leaving the mismatch between need and willingness undiscovered.

 

One reordering that changed the plan

Across both sessions, grant writing dominated the conversation. It was the obvious use case and the most immediately appealing.

 

Late in Session 2, the board member closest to the organization's finances redirected the group. Grants are necessary and ongoing, she said, but the thing that had been stuck since 2021 was donor diversification and recurring giving. Without solving that, the organization stays flat.

 

The roadmap was reordered on the spot: document the AI position and guardrails first, then build a strategic roadmap oriented toward the shift to recurring donors. Grant work continues, but it stopped being the headline.

 

An organization can spend a productive year applying AI to the wrong priority. Part of my job as facilitator is making sure the loudest use case and the most important one are the same thing — or noticing out loud when they are not.

 

 

Outcomes

 

The board reached a decision it made itself: yes to AI, conditional on guardrails it defined, with a quarterly review commitment and named ownership.

 

The executive director moved from unfamiliarity to active daily use, and more usefully, to a considered position on it. Her own framing was that she wanted to use AI as a tool without letting it stunt her development as a leader. She began asking it to show its reasoning rather than just its answers — treating it as a class rather than a shortcut.

 

Board members with no prior exposure completed hands-on tasks, compared tools, and arrived at the second session with reasoned preferences and specific observations about how different tools handled the same request.

 

Priorities were reordered around the organization's real strategic need rather than its most visible one.

 

And the board asked for more. Following the workshops, the board requested an ongoing advisory engagement to coach the executive director on AI adoption and on leadership more broadly. That engagement is written into the organization's budget and will begin as funding allows.

 

 

What This Work Demonstrates

 

A values objection is a design constraint, not resistance. The fastest way to lose a mission-driven room is to treat its ethical concerns as an obstacle to overcome. Naming the tension first, and then doing actual research on it, is what earned the rest of the conversation.

 

Fluency spreads are a real problem in small organizations. A room containing both a daily AI user and someone who has never opened one of these tools cannot be served by a single pitch. The work is reaching both ends without patronizing either.

 

Alignment without ownership is just a nice conversation. The board's stated fear — another engagement producing assets nobody implements — was well founded. The most valuable stretch of the workshop was also the least exciting: naming who owns what, and asking people to be honest about their actual capacity.

 

A new leader needs advisors, not just tools. Handing a first-time executive director powerful technology without a structure for gut-checking her judgment produces confident output nobody has vetted. Pairing her with named board members per work stream addressed a leadership gap that AI would otherwise have widened.

 

Design the engagement around the constraint the client already told you about. A lean team that says it cannot afford to experiment is not going to be served by a four-hour block that ends in principles. Splitting the format and building the experiment into it cost nothing and changed what the sessions could produce.

 

The loudest use case is not always the right one. Grant writing was the obvious answer. Donor diversification was the correct one. A neutral third party in the room is often the reason that distinction gets made out loud.

 

 

What This Means for Leaders Navigating Similar Challenges

 

If you lead or govern a small mission-driven organization, the question in front of you is probably not whether AI could help. It is whether you can take it on without breaking a team that is already stretched, and whether using it sits comfortably with what your organization stands for.

 

Those are the right questions. They are also not answered by a tool demonstration or a downloadable policy template. They are answered by getting the people who make decisions into a room together, working through the hard parts honestly, and leaving with commitments specific enough to act on.

 

That is what Trailhead Communications does. Not just for organizations with big budgets and mature AI programs, but for the ones who need to move carefully but still need to move.

 

Return to case studies page →