LH West
How We Help Our Services
AI, Honestly
Our Approach About Contact Us Free assessment
AI, honestly

The real worries about AI, answered with real numbers.

There's a lot of fear and a lot of hype around AI right now. We'd rather give you the facts. Below are the concerns we hear most from nonprofit leaders, and what the current evidence actually says. Every figure is sourced at the bottom of the page.
The worry about energy

“Isn't AI terrible for the environment?”

−
With all the news about data centers and power use, isn't every AI prompt doing real environmental harm?
The headlines about data-center growth are real, and worth taking seriously. But the cost of a single everyday question is far smaller than most people picture.
In 2025 Google measured a typical text prompt to its Gemini AI at 0.24 watt-hours, about one second of running your microwave, or under nine seconds of watching TV.3,4 Independent researchers put a typical ChatGPT text query around 0.3 Wh, and OpenAI's own estimate is about 0.34 Wh, roughly a second of an oven.5
To put that in everyday terms, charging your phone once uses about as much electricity as 75 to 80 typical AI text prompts.3,6 Each prompt also uses about a quarter-milliliter of water, roughly five drops.3
55 to 75
text prompts use about the same electricity as one hour of video streaming.3,9
One typical AI text prompt
~0.24 to 0.34 Wh3,5
One classic Google web search
~0.3 Wh, 2009 estimate5
One hour of video streaming
~18 Wh, about 55 to 75 prompts9
Charging your phone once
~19 Wh, about 75 to 80 prompts6
The honest other half
The total adds up, and we won't pretend otherwise. The International Energy Agency expects data-center electricity use to roughly double by 2030 to about 950 TWh, near 3% of world demand, and image, video, and reasoning tasks can use hundreds of times more energy than a simple text prompt.7
That's exactly why how you use AI matters: deliberate, well-chosen use keeps both the impact and the footprint small.
The worry about jobs

“Won't AI replace our staff?”

−
If we bring in AI, we'll be replacing the people who do the work.
The nonprofit world isn't overstaffed, it's stretched to breaking. You can't automate away a job you already can't fill.
Nearly three out of four nonprofits report job vacancies, and most of those openings are in the program and service-delivery roles that touch the mission directly.1 Almost a third have longer waiting lists for the people they serve, and half name stress and burnout as a driver of the shortage.1
AI doesn't shrink your team, it relieves it. It takes the repetitive, draining tasks off overloaded plates so your people can spend their hours on the relationships and judgment only they can offer. More capacity and impact, less overwork.
74.6%
of nonprofits report job vacancies they can't fill, and about 74% of those openings are in mission-facing program and service roles.1
The worry about authenticity

“Won't it make us feel fake?”

−
If we use AI, our voice and our communications will lose the human touch our community trusts.
Authenticity isn't lost to AI, it's lost to AI used carelessly: bolted on, unsupervised, standing in for human judgment.
Used with a clear plan, AI works quietly behind the scenes, drafting, organizing, summarizing, while your people keep final say over every word that carries your voice. The problem was never AI. It's AI without a plan, and most organizations don't have one: the sector has adopted AI far faster than it has governed it, with only about 15% having a written policy for responsible use.2
That's where we start. One of the first things we do together is build an AI ethics plan for your nonprofit, clear, shared rules on where AI helps, where it doesn't, and how your authentic voice stays unmistakably yours.
~15%
Only about 1 in 7 nonprofits has a written policy for responsible AI use, most are improvising without a plan. We help you build one.2
The worry about quality

“We tried it, and what we got back wasn't any good.”

−
We've used AI and the writing came back generic, off-tone, or just wrong. It's not worth the trouble.
Most disappointing results don't come from the tool, they come from how it was asked. A one-line request with no context gives you one-size-fits-all writing. Good output comes from a good brief.
The skill is in telling the tool who you are, who you're speaking to, the voice you use, and what a strong result looks like, then reading the draft critically and refining it. It's a learnable skill, and most teams simply haven't been taught it yet: 96% of nonprofits feel they understand AI, but 76% have no AI strategy or training in place.8
Think of it like bringing on a sharp new volunteer. Hand them a single vague sentence and you'll get generic work back. Give them your mission, your audience, and an example of what “good” looks like, and they'll surprise you. Teaching your team to ask well, so they get genuinely useful, on-brand results instead of random ones, is a core part of what we do.
76%
of nonprofits have no AI strategy or training in place, the real reason results disappoint, far more than the tool itself.8
What turns a vague prompt into a useful one
Contextyour mission and the situation.
Audiencewho it's really for.
Voiceyour tone, with a do and a don't.
An examplea past piece that hit the mark.
A review stepa person who edits and approves before anything goes out.
The worry about waste

“Won't everyone just use it randomly?”

−
If we open the door to AI, won't it be chaotic, everyone improvising, duplicating work, and wasting effort?
It's the opposite. The waste comes from the ad hoc version, every staffer quietly trying different tools with no shared approach. That's where you get duplicated effort, inconsistent quality, hidden risk, and a lot of wasted motion, and yes, wasted energy too.
When a team learns to use AI together, shared tools, a little training, and a clear plan, the same work takes fewer prompts, fewer redos, and less guesswork. The results are more consistent, safer, and more efficient than scattered solo experiments. Right now most nonprofits are in the scattered phase, using AI with no shared plan at all.2
We help your whole organization adopt AI as one coordinated practice, so it's an asset you control, not a dozen side experiments you don't.
65%
of nonprofits using AI are doing it ad hoc, with no shared plan across the organization.2
Still have questions?
Let's talk it through.
The free AI assessment is a low-pressure way to get a clear, honest read on where AI could genuinely help your organization, and where it shouldn't. Three minutes, no commitment.
Sources
Where these numbers come from
Figures are current as of our last review in June 2026. AI energy estimates in particular are evolving quickly; we re-check them periodically.
1.
National Council of Nonprofits, 2023 Nonprofit Workforce Survey, about 1,600 U.S. nonprofits: 74.6% report vacancies; about 74% of vacancies in program and service roles; 28.1% report longer service waitlists; 50.2% cite stress and burnout. councilofnonprofits.org
2.
Candid, AI equity for nonprofits, 2025: roughly 15% of nonprofits report having a policy for responsible AI use, and most AI use remains ad hoc rather than planned. candid.org
3.
Google, Measuring the environmental impact of AI inference, August 2025: median Gemini text prompt about 0.24 Wh, about 0.03 g CO₂e, about 0.26 mL water. cloud.google.com
4.
MIT Technology Review, August 21, 2025, reporting on Google's figures, including the “under nine seconds of TV” comparison. technologyreview.com
5.
Epoch AI, February 2025, How much energy does ChatGPT use?: a typical GPT-4o text query about 0.3 Wh; references OpenAI's 0.34 Wh estimate and the 2009 Google search figure of about 0.3 Wh. epoch.ai
6.
EnergySage, average smartphone full charge about 19 Wh, used here for the phone-charge comparison. energysage.com
7.
International Energy Agency, Energy and AI, 2025: global data-center electricity use projected to roughly double to about 950 TWh by 2030, about 3% of demand; image, video, and reasoning tasks can use far more per query than text. iea.org
8.
TechSoup and Tapp Network, 2025 AI Benchmark Report: 96% of nonprofits report at least a basic understanding of AI, yet 76% have no AI strategy in place; lack of knowledge and training is a leading barrier, especially for smaller organizations. techsoup.org
9.
Carbon Trust, Carbon impact of video streaming, 2021: about 18 Wh of electricity per hour of streaming in Europe, excluding the viewing device. Streaming on a large TV uses considerably more, on a phone considerably less. carbontrust.com
© 2026 LH West. Nonprofit consultants, Salt Lake City, Utah.