What about AI? If you were just reading the headlines, you’d have to guess that it’s good for…everything. Well, not everything. But not nothing, either. It’s actually pretty hard to tell from the outside what AI “does.”

This is less of a marketing problem than a peculiarity of the technology itself: AI is a general purpose technology, which means it’s useful across a wide variety of tasks.

Unfortunately, that generality spells a problem for our ability to get real value from AI tools. I know this because I hear the same question in nearly every workshop I run, usually from folks who are clearly already trying: “But Remy…what is AI actually good for? Where can I use it in my day-to-day work?”

It’s an excellent question. And today I want to help you find the answer.

Being jazzed about AI’s mission-advancing potential in principle means nothing for your ability to realize that potential in practice. And developing a robust AI strategy demands a baseline understanding of what this technology makes possible in our work—which eventually lets us begin to use it in truly transformative ways.

The head, the hands, and the heart

In my last essay, I encouraged you to “embrace the plausible” with AI—meaning you shouldn’t let your ideas about AI’s limitations prevent you from discovering its real power when used well. For the purposes of this exercise, I’m going to take it a step further: I want you to momentarily forget about AI altogether. (We’ll get back to it shortly, I promise.)

Now, take a moment to think about a task or project you work on regularly—something that recurs week after week. Pick a familiar workflow, not something brand new to you.

Once you have a task in mind, break it down into 3-5 discrete steps on a piece of paper. Then, label each step as one of the following:

• HEAD: driven by cognition/expertise. These are the parts of the task that really demand your judgment, experience, taste, and accumulated knowledge of organizational/situational context. Assessing whether a grant opportunity is a good fit for your organization is “head work.” So is deciding how to frame your theory of change to a new funder, or figuring out what a data anomaly actually means. These are the steps where your uniquely-trained brain is the irreplaceable ingredient; you wouldn’t hand them off to an intern or new team member without careful training and review.

• HANDS: driven by manual labor/execution. These are the parts of the task that mostly depend on manual efforts, like moving data from a spreadsheet into Salesforce, updating a calendar with grant deadlines, or adding your organization’s logo to a slide deck. Nothing about these steps hinges on your unique expertise; they just require your time.

• HEART: driven by the warm touch. These are the parts of the task that nonnegotiably require your humanness: your empathy, your presence, your relational instincts, your ability to read a room or hold space for someone, and so on. For example, calling a colleague to share difficult news and welcoming a program participant to their first event both involve a lot of “heart work.”

You might notice, as you attempt this activity, that many steps overlap with at least two of these categories. Drafting a grant application calls on your head and your hands; conducting an intake assessment with a new program participant calls on your head and your heart; writing thank-you notes to major donors calls on your head, your hands, and your heart.

But quite often, we struggle to decompose tasks into single categories mainly because our approach to completing work has become so automatic. We don’t pause to think through every invisible step in the process of drafting a meeting agenda—we just do it.

For this exercise to give us real value, it’s worth spending some time getting in the weeds of the task you’ve chosen. If you can’t categorize a step, take a closer look. Have you really gone as far as you can in breaking the task into its basic, discrete components?

To demonstrate what I mean, here’s how I might complete a first pass at this exercise for a community listening session:

This isn’t a bad starting point—but if we look a little closer, it becomes obvious that the tasks we’ve identified could be decomposed further.

To demonstrate, let’s isolate just one step from the previous process: identifying which voices should be in the room for our community listening session. What really goes into executing that task, step by step?

We could go even further with this kind of decomposition, but you get the gist: every task represents a series of subtasks that demand different kinds of work from us.

From mode to mechanism

With that process in mind, we can come back to the reason you’re reading this newsletter. Once you’ve gone through this exercise a few times, you should actually have a pretty clear roadmap for where to start experimenting with AI and automation.

I’ll explain.

The Head/Hands/Heart framework asks two very basic questions:

• [Task] What do I do?

• [Mode] How do I do it (currently)?

The logical next step is to figure out where AI and automation fit into this puzzle (if at all):

• [Mechanism] How might we support the task using the technology at our disposal?

Given how AI and automation tools work, the breakdown we’ve just done gives us clues about where they might be especially helpful in our work.

Hands work (the steps that require your time, but not your judgment) tends to be repetitive and predictable. That maps well to automation tools, which are typically deterministic—they follow fixed rules, executing the exact same process every time. If a workflow fits the structure “Every time X happens, Y should happen,” you can probably use a tool like Zapier to run it automatically.

Head work (the steps that draw on your expertise, context, and pattern recognition) tends to require variable, situation-specific adaptation. That maps well to AI tools like Large Language Models (LLMs), which are probabilistic—they extrapolate from patterns to predict the most likely output, which is why two people asking the same question to an LLM can get meaningfully different answers.

And of course, heart work stays human-only, full stop.

Here’s how we might use AI and automation tools to augment our community listening session preparation efforts*:

Step 1: Pull previous attendee lists and contact records (Hands)

Automation: A Zapier workflow connecting Eventbrite to Salesforce automatically compiles historical attendee data the moment you start planning—no manual export required.

Step 2: Map stakeholder groups against the session’s focus (Head)

AI augmentation: You feed your list and a short description of the session’s goals into a Claude Project with organizational context already loaded in. You then ask it which groups are well-represented, which are missing, and what perspectives those gaps might leave out.

Step 3: Scan past data for who’s been consistently absent (Head)

AI augmentation: You drop your listening session attendance history into Claude and ask it to flag who’s been invited but rarely shows up. (Or you set up a Tableau dashboard that automatically surfaces these patterns based on live data.)

Step 4: Call trusted community connectors (Heart)

Mostly human, with a sprinkle of AI: This step stays mostly human—but if you break it down further, you see that you can use Claude to proofread a warm outreach message so that you’re not starting from a blank page when you pick up the phone.

Step 5: Make the final judgment call on the invite list (Head)

AI as thought partner: You bring your draft list back to your Claude Project and ask it to pressure-test it for blind spots, underrepresented perspectives, or power dynamics worth considering. The AI doesn’t make the final call—that’s all you!—but it helps you consider perspectives you might otherwise have ignored.

*I’ve chosen a few specific tools for this sample workflow, but there are many similar and complementary tools that can support this kind of workflow augmentation.

If you’re struggling to figure out real applications, you’re in luck: AI itself can be enormously helpful in the ideation process.

Once you’ve taken a pass breaking down a workflow using the Head/Hands/Heart framework, you can feed your breakdown directly into your LLM of choice and ask it to help you ideate additive ways to use AI and automation tools.

Here’s a prompt to do exactly that:

I’m attaching a breakdown of a common task I perform at work, with subtasks categorized using the Head/Hands/Heart framework. HEAD work is driven by expertise, judgment, and organizational context (i.e. it needs my expertise to be done well). HANDS work is driven by manual execution (i.e. it needs my time, but not necessarily my unique perspective/judgement). HEART work requires my humanness (i.e. it needs my empathy, relational instincts, or a warm touch that can’t be replicated by technology).

Interview me one question at a time until you fully understand the scope of the work (i.e. how I achieve the intended outcome). If necessary, challenge my labeling—I might not realize where a task description is overly broad and could actually be broken down further into Head/Hands/Heart.

Once we have a set of discrete, labeled steps, help me identify specific AI tools or automation approaches that could augment the Head and Hands work. Be specific about platforms and techniques where you can, and prioritize these ones: [list out any AI/automation tools you already use or have access to]. If necessary, do supplementary web research to home in on the most recent features of these tools.

Here’s a bit about my role and context to help you out: [describe your role, your organization’s mission, and 1-2 sentences about what success looks like in your work].

To make this as easy as possible for you, I’ve also designed a Perplexity Space that walks you through this process step by step. You can access that tool here.

What this work makes possible

Going through this exercise does something else for us, too—something I didn’t fully anticipate when I first started running it in workshops.

When you decompose your work into Head, Hands, and Heart, you stop seeing it as a monolithic to-do list and start seeing it as a bundle of distinct activities, each of which makes a different kind of demand on you. And once you see the bundle clearly, a new set of questions opens up: “Which of these pieces actually need me, and in what ways? Which pieces are holding me back from doing more of the work that does?”

Having now walked hundreds of people through this exercise, I’ve found that it often leads to micro-epiphanies. You may see for the first time just how many of your daily activities don’t actually “need your brain.” You may find that you’re not doing nearly as much “heart work” as you expected—even if that work is what drew you to your organization’s mission in the first place. You may find that you really enjoy the “hands” work and want to do more of it! Or you may find something entirely different.

Whatever you find, it matters. Those realizations can enable critical conversations about the nature of our work, and where technology actually fits into it. Those conversations will be essential as we try to build good things with AI.

We are in a rare moment. We have the opportunity to be imaginative not just about what AI and automation can do for our missions, but about what we bring to our missions that AI and automation fundamentally cannot replicate. That’s where this exercise actually points (and where we’ll pick up this conversation next time).

Our fundamental goal is a clearer vision of what our mission actually asks of us—and a clearer sense of how technology can help us do more of that work, not less of it.

Share

One day, your CEO calls an all-staff meeting. You’ll all be hearing a presentation from a consultant on some new tech tool that’s apparently going to “change the game” for your team.

Oh boy, you think. I really don’t have time for this.

Nonetheless, you show up in the conference room as requested. (This was the pre-Zoom era, remember?) You settle in, lean back, and prepare for whatever’s to come.

The consultant walks into the room, introduces himself, and pulls up his screen on the projector. He starts to click around some platform called “Salesforce.” (If you haven’t used Salesforce, swap it out for your CRM of choice: Raiser’s Edge, HubSpot, Convio, eTapestry...)

No one around the table has seen this kind of interface before; you all live and breathe Excel, crossing your fingers that you remember to take action on follow-ups at the right time. As he continues to click, people start to lean in and scribble notes.

Then, the consultant shows exactly how Salesforce replaces the work that’s been causing you to tear your hair out recently. He pulls up a messy spreadsheet, feeds all of the data into the CRM platform, and drags the spreadsheet to the laptop’s trash bin for dramatic effect.

You’re shocked. This changes everything, you think. Your mind starts racing with ideas for how you’re going to use this powerful new tool, how it’s going to make your work so much easier, how it might even make it possible for your team to reach more people. It feels like this consultant has opened your eyes to a world of new possibilities. You can’t wait to dive in.

*And…scene.*

The room that our hypothetical Salesforce consultant walked into was essentially a blank slate—a bunch of people without prejudice, learning about a new way of doing things.

That’s not the kind of room I walk into when I run AI workshops.

Even when I’m hosting what amounts to “AI 101,” I know I’m far from anyone’s first exposure to this tech: most people have seen myriad headlines about AI, and at least 70% of the people sitting in have already used ChatGPT at some point.

But “using ChatGPT” can obviously mean a lot of different things. So when I move to demos, I try to get a read of the room.

“Raise your hand if you’ve created a custom assistant (like a Project or Gem) in an LLM.”

On a strong day, I’ll get one sheepish hand. But it’s usually zero.

Then: “Raise your hand if you’ve used the Deep Research function in an LLM.”

Nothing.

So, naturally, I show some tailored examples of how to use these features in their work. And it routinely blows people’s minds, which blows my mind right back.

The techniques and features I demonstrate—things like prompt engineering, customization, research, reasoning, canvases, shortcuts, skills, and integrations—are typically just one click away. But most people have never even heard of them—and when they see them in action, it completely subverts their expectations.

After seeing this pattern play out enough times, I got to thinking. Why is it that so many people are utterly disappointed with LLMs but have, at the same time, barely scratched the surface of what they make possible?

Well, many of us land on a tool like ChatGPT, intuitively associate the prompt bar with Google’s search function, and start off by treating it like a glorified Google. (Conor Grennan calls this “Google brain.”) When people arrive at my workshop from this context, they’ve technically used the tool, but they’ve never really pushed beyond simple search functionality.

Unfortunately, those of us who have tinkered a little more ambitiously have often been rewarded with lackluster results: we’ve trusted the tool with a simple task and gotten back a bad output, a hallucinated fact, a draft that sounded nothing like us, or an “I’m sorry, I can’t do that for you.” This kind of outcome hardly motivates people to upgrade to a paid subscription for a “Pro” version of an LLM—why should they, when the free version doesn’t even work that well for what they need?—so they miss out on the powerful features that become available for ≤$20/month.

All of that means that they’re showing up to our conversation with preformed notions about what AI tools can and can’t do. Fixed in this mindset, they look past what might be possible with a little more thoughtfulness, intentionality, training, or time. They decide that AI is overhyped and ultimately unhelpful for their day-to-day work.

I understand where they’re coming from. People tend to join nonprofits to work on a mission, not to learn how to use software. When new technologies don’t work seamlessly from the start, they don’t feel worth our time and energy (especially when “the old way works perfectly fine”).

But generative AI isn’t like most technologies. It’s probabilistic, not deterministic. Its full range of capabilities aren’t understood by anyone. And, most unusually, its behavior can be easily molded or modified by a user without any technical knowledge or coding experience.

Tapping into AI’s potential is less like working with a machine and more like working with a cyborg. None of us have worked with cyborgs before—so of course it takes time to figure out how to do it well. Skipping over the experimentation process altogether is a failure of imagination.

In order to get more value from AI, we must release our preconceptions and embrace the plausible.

Asking a different kind of question

Very often in my workshops, I hear questions like these:

• “Can AI help me manage my inbox?”

• “Can AI create MOUs for us?”

• “Can AI make it easier to fill out our timesheets?”

But “Can AI?” presumes a binary. It confines us to what seems possible—and that judgement is often made based on what we’ve seen happen in the past. When we decide the answer is “no,” there’s no room for further action. We’re just stuck.

(The answer to all three of those questions is yes, by the way.)

We have another option, though: “How might we help AI?”

Fundamentally, this is a more expansive question. It invites an exploratory mindset, which happens to be the best way to approach this technology. And it puts us back in the driver’s seat—we have a say in whether the plausible becomes real, if only we can engage with AI in the right ways.

Why is this reframing so important? In a now-classic 2023 essay, Wharton professor Ethan Mollick sums it up simply:

“AI is weird. [...] On some tasks AI is immensely powerful, and on others it fails completely or subtly. And, unless you use AI a lot, you won’t know which is which.”

In that same essay, Mollick attempts to make this idea more tangible by painting a picture of a fortress wall. The wall (which he calls the “Jagged Frontier”) represents AI’s capabilities; tasks on the inside are easy for AI, while tasks on the outside are hard for AI. The wall itself is winding and invisible, so you can’t always predict which side a given task will fall on; the only way to find out is to make an attempt.

To make this more concrete, let’s take a real (albeit somewhat silly) example of how that wall shows up in practice.

If you had asked an early version of ChatGPT how many Rs are in the word “strawberry,” it would have confidently told you two. If you had pushed back on its claim, it would have apologized, confidently reexplained its reasoning, and…still gotten it wrong.

Users caught onto this peculiarity in mid-2024, and of course, it went viral. Soon enough, experts rushed in to explain why this was happening: instead of reading letters, LLMs like ChatGPT read segments of text called “tokens.” Tokens can comprise words, subwords, or characters; to the model, “strawberry” might have looked something like [str] + [aw] + [berry]. As simple as it may seem to us, counting individual characters across those text segments is challenging for a system that wasn’t built to read that way.

So: counting Rs is outside of the wall. Task failed. AI is broken. Right?

Well…not quite. The right technique could actually get you around the wall.

If you had gotten creative and asked ChatGPT to spell out each letter before counting, something different would have happened. By forcing the model to process the word character-by-character before jumping to an answer, you would have routed around the tokenization problem entirely. Suddenly, with the right approach (prompted by you), the model could get it right.

And then there’s time. A few months after the strawberry problem hit the internet, OpenAI released a new reasoning model called o1. They nicknamed it “Strawberry”—a cheeky nod to the problem it was built to solve. The model approached questions by working through “reasoning” steps before answering, and as a result, it could correctly determine that there are three Rs in the word “strawberry.” What required a clever workaround in June was simply solved by September.

This is the pattern. When something that seems like it should be achievable with AI isn’t happening, it’s often a question of technique or time. Technique can get you around the wall today; time can move the wall permanently. Both show us that what seems “impossible” very often isn’t.

What happens when we suspend our disbelief

The strawberry problem is a charming one, but most of us don’t get paid to spell out the names of fruits. All it really offers is a lesson about the Jagged Frontier. It’s on us to extrapolate to work that actually needs doing in the real world.

Luckily, there are some excellent examples of organizations doing just that—and solving very real, very human problems as a result.

Take GiveDirectly, a nonprofit that distributes cash directly to disaster survivors and people living in extreme poverty. When Hurricanes Harvey and Maria hit Texas and Puerto Rico in 2017, their team deployed the way disaster relief organizations always had: they flew in, surveyed the wreckage, and handed out physical debit cards at rolling enrollment events. Over 7 months, they distributed a total of $9.5 million to 6,363 survivors.

By all accounts, this response was a major humanitarian success. At the time, it might have seemed naive to ask “what more can be done?” (And it certainly would have seemed unusual to bring AI into the mix, at least to an outside observer.)

But that’s exactly what the GiveDirectly team did. The people they were reaching had just lost everything, and they “knew that it [fund distribution] wasn’t fast enough.”

They faced two major bottlenecks. First was the time it took—on the order of weeks—for GiveDirectly staff and volunteers to walk through disaster zones, block by block, to “identify areas that were both high-poverty and high-damage.” Second was the time it took to enroll each recipient in their program (25 minutes) and load the distributed debit cards with funds (up to 2 weeks).

So they started asking an expansive question: not “can AI help us do this faster,” but “how might we help AI do what humans can’t do at all?”

What followed was years of iterative experimentation to reimagine their disaster response process around the capabilities of new technologies.

First, GiveDirectly partnered with Google.org to build an AI (machine learning) model that used satellite imagery to identify damaged homes without responders needing to set foot in the affected area. They then layered on government poverty data, allowing them to narrow in on the households most in need after a disaster.

Finally, they replaced their in-person enrollment process with a mobile sign-up through a food-stamp benefits app that was already used (and trusted) by millions of low-income Americans. Because GiveDirectly had “pre-determined their financial and recovery needs” and didn’t require any additional proof of eligibility, this new approach took only 3 minutes—an 88% reduction in processing time.

By the time Hurricanes Fiona and Ian arrived in Puerto Rico and Florida in late 2022, GiveDirectly’s remote aid distribution pipeline was fully operational. Over 90% of aid recipients received their payments within 24 hours of enrolling. Achieving the same target outcome—cash in the hands of the people who needed it most—went from seven months to one day. Overall, GiveDirectly distributed $3.3 million to 4,748 disaster-impacted low-income households in a matter of weeks.

Importantly, this pipeline didn’t arrive fully formed. It took five years, multiple partners, and ongoing iteration to make possible.

But GiveDirectly was willing to repeatedly ask a question with an unknowable answer, and it paid off.

Operationalizing the plausible

Of course, GiveDirectly is a major international nonprofit with a $200 million annual budget. Most of us don’t work at organizations with this level of resources and connectivity; most of us won’t have the opportunity to collaborate with Google.org on a machine learning model.

Still, the “how-might-we” mindset that shaped GiveDirectly’s success with AI is universally applicable. (I’ve compiled a few moving stories of other organizations, including much smaller and scrappier nonprofits, finding incredible AI solutions to hard problems here.)

As you begin to explore how AI might apply to your mission, start from curiosity. Don’t be afraid to get it wrong a few times, as long as you’re experimenting within safe bounds. In the worst-case outcome, you’ll learn something useful about AI and about your organization’s work. In the best-case outcome, you’ll find new, previously-unimaginable ways to deliver on your mission.

GiveDirectly began their AI journey with an open-ended question. They couldn’t have jumped straight to satellite imaging models; they lacked the insight needed to make that call upfront. Their mindset enabled the innovation that followed.

When we embrace the plausible, it’s impossible to know where our questions will lead. But being brave enough to ask them is the only way to find out.

Let’s assume it was great service and you were feeling generous, so you decided to tack on 22%.

Did you:

1. Calculate 22% of the total in your head?

2. Write out the math on the back of a napkin?

3. Figure out 20% in your head and guesstimate the difference?

4. Whip out your phone’s calculator to get an exact dollar amount?

5. Use the percentage-based cheatsheet at the bottom of the receipt?

6. Or…switch your tip to 20% because it made the math easier?

You may not even remember the mechanics of such a seemingly microscopic decision (especially if the restaurant was noisy or you were in a rush to leave). In fact, it was very likely not a decision at all, but the unconscious execution of a standard you’d already landed on—explicitly or not.

And that would be completely fair. Tipping at a restaurant is fairly low-stakes.

But right now, many of us are beginning to make the same kind of decisions (or non-decisions) about much higher-stakes activities. As AI tools become increasingly capable, we’re trusting them with our emails, our slide decks, our travel plans, and our strategic choices.

When AI is used in ways that augment our creativity and critical thinking, that trust can be a great thing. In the best cases, it can allow us to reallocate our precious time and energy from rote tasks to the kinds of efforts that really need our brains.

But how do we draw that line? And how do we hold it when AI’s capabilities encroach more and more on areas of our work and lives that have always centered human skill, judgement, and taste?

It all points at a question that urgently needs our attention:

How do we decide what’s “worth” our mental energy?

In our lifetimes, this question has perhaps never been as pressing as it is in this moment. The cost of outsourcing our cognition is going up, and we have a short window of opportunity within which to decide what kind of cost we’re willing to bear.

Your brain on complexity

Let’s stick with mental math for a moment, because it’s a universal example.

I want you to scan through the following mathematical expressions, one by one, and briefly envision how you might simplify them:

• 8+3

• 19+7

• 41-16

• 11x3

• 72÷6

• 13x81

• 896.2+155.9+307.4

• (17.1)³

• √143.2

Was there a clear point of no return? I’d imagine you could tackle the first few pretty easily. But as the calculations got more complex, your eyes may have glazed over.

Your tolerance for mathematical complexity is likely linked to your age and your professional or academic background. If you grew up in the 1970s and majored in engineering, you might have a higher threshold for abandoning mental math and seeking assistance. If you grew up with a supercomputer in your pocket and a strong distaste for STEM subjects…maybe not so much.

Your present circumstances probably also informed your patience with this exercise. Are you skimming this newsletter in a rush before your next meeting, or is this your lazy Sunday morning read? The less time, space, attention, or bandwidth you have available, the less likely you are to take a stab at the more complex calculations.

No matter your particular threshold on this particular day for simplifying mathematical expressions, the cost of outsourcing arithmetic is pretty negligible. After all, the skill being displaced (mental math) isn’t load-bearing for most of our lives.

But that’s not always the case. The thing is, high-value activities are often complex—and getting them wrong is often very costly.

In work and in life, many of the things most worth doing look more like the back half of that list of mathematical expressions. “8+3” work doesn’t move the needle. “13x81” work, though? Sure, it’s tough. But making progress on important work often demands that we stretch ourselves.

Therein lies the problem: if more important work is more likely to overwhelm our cognitive capacity, we risk screwing that work up (and likely incurring a high cost—reputationally, organizationally, interpersonally, etc.—as a result). That’s why it’s so critical to understand how our brains operate under this kind of stress.

What we do when complexity arrives

When you surpass your threshold for complexity, you’re likely to take one of three paths: cognitive offloading, cognitive surrender, or cognitive capitulation.

Let’s take a closer look at each of these pathways through the lens of a classic metaphor: ascending a mountain.

You’re on the ground, trying to figure out how to get to the peak. How do you do it?

Cognitive Independence (agency): You hike.

No special gear needed—it’s just you and your human capabilities. This is when we’re still sitting below our personal threshold and decide to leave tech out altogether.

Cognitive Offloading (augmentation): You take an e-bike.

You engage the motor to handle the burden of the incline, but you’re still pedaling, steering, and maintaining situational awareness of the trail. This is how most of us typically use technology—to reduce the burden of specific tasks on our brains (or, in this case, bodies).

Cognitive Surrender (abdication): You take a chairlift.

You let the machine carry you all the way to the top. You aren’t monitoring the path or checking the mechanics powering the lift…you’re just waiting to be delivered to your destination. This is when we offload both a task and our judgement (i.e. monitoring the processes/outputs) when using technology, which tends to happen more often when we’re under time pressure.

Cognitive Capitulation (abandonment): You walk away.

You take one good look at the mountain and decide to call the whole thing off. It’s not just that the ascent is not worth your physical energy; you’ve actually determined that getting to the top isn’t worthwhile at all. This can happen when the perceived value of the outcome itself (not just the process) diminishes—for example, an illustrator walking away from their craft upon realizing that anyone can instantly mimic their years of carefully-honed expertise with a single prompt in an AI image generator.

The goal of “techno-intentionalism” is to stay in the space of strategic, healthy cognitive offloading—even when task complexity passes our threshold of tolerance.

Why AI makes this even harder

Most technology is designed to remove friction from our lives. AI does this almost too well, and it’s by design.

Demis Hassabis defines “artificial general intelligence” (AGI)—the theoretical pinnacle of innovation in the AI field—as “a system that can exhibit all the cognitive capabilities humans can.”

When you design a consumer technology product to mimic human cognition, deploy it in a chatbot interface, and optimize for continuous user engagement, you get a dizzyingly addictive tool whose primary feature is frictionlessness. We no longer have to build new tools to find a path of less resistance; AI will do it for us at every step.

Case in point: I was working with Claude, my Large Language Model (LLM) of choice, to structure my thinking for this article. After it suggested a modified approach, it asked me: “Want me to take a swing at a full draft, or would you rather start with a tighter outline you can react to first?”

Hmm. Why sit here writing for the next few hours when Claude could whip up a newsletter draft on my behalf?

“Well,” said the angel on my shoulder, “because you started this newsletter to refine and share your own thinking, not your AI assistant’s musings. Plus, you’re literally writing about unhealthy cognitive offloading to AI…need I say more?”

Fair enough.

So I said “no thanks” to the full draft, mentioning the irony of its ask given the topic of the piece and noting that I would gladly take a look at an outline instead.

Claude replied: “Ha — noted. And honestly, that’s a great line you should find a way to work into the piece.”

And here we are.

AI lowers our threshold for complexity by making it almost impossibly easy to outsource your tasks and your cognition. Claude could tell what I was getting at with the outline I’d shared, so it cut to the chase. All I had to do was type “yes” and hit enter—four keystrokes to get a full, tailored essay about this topic.

I can’t say it wasn’t a little bit tempting to do just that (if only to see what might come out on the other side). But doing so would have been cognitively offloading something I actually wanted to do on my own. Following the path of least resistance would have been an abrogation of duty to myself.

Your brain is already drawing an invisible line

It’s not always as simple as ignoring a chatbot’s overeager offers of help. Even if you don’t realize it, your brain is already drawing invisible lines that will inform how you respond to complexity in the future.

I know this because I’ve already seen my own thresholds begin to shift.

A few months ago, my partner and I were sitting on our couch for some afternoon coworking. I was researching something in Perplexity, an LLM aggregator; she was working on a book project.

At one point, she turned to me and asked for my help. Consumed by whatever was on my screen, I reluctantly yanked myself out of the flow to hear her out.

She needed to come up with a placeholder quote for a section of the manuscript—something a Gen Z person might say about the challenges of modern young adulthood.

I thought about it for a few moments and nothing came to me. My brain was still in two places, and I was impatient to get back to my research.

So I said this:

“Hmm…that sounds like a job for ChatGPT.”

(Read: not worth my mental energy, and definitely up AI’s alley.)

My partner quipped half-jokingly that “this is what people mean when they say that AI is hurting our critical thinking.”

So, of course, I did what any good partner would do: I got defensive. “I did think about it—it’s just that nothing came to me, and I was eager to get back to my work. I could have done it, though….”

Once I’d successfully dismounted my high horse, I realized that she was broadly right. I had assessed the challenge, gauged my own willingness to tackle it in that moment, and determined that it wasn’t worth my energy.

Would I have given up so easily in the pre-LLM era? It’s impossible to know for sure. But it seems at least somewhat plausible that I spent less time trying to come up with my own answer, or considering whether the problem was worth tackling in the first place, because I knew that there was another (less cognitively-taxing) option freely available at our fingertips.

This should sound familiar: when you encounter a complex math problem on your receipt, you shut off your brain partially because you already know that you have a tool available—right in your pocket!—that can take it on much faster, and with much less strain, than you can.

The existence of the tool is neutral. It’s our knowledge of the tool’s existence (and its immediate accessibility) that paves this neural pathway.

In that moment of divided attention, I’d failed to bring awareness to my own, nearly-automatic cognitive reflex. As a result, I missed the opportunity to decide if I actually wanted to outsource the task at hand to AI.

It’s in our nature to gravitate toward the path of least resistance—so much so, in fact, that since our early days as a species, we’ve often decided that “the path of least resistance” still has too much resistance. That’s when we’ve built tools to carve new, even easier paths.

Pushback against this tendency is a tale as old as time. In Plato’s 370 B.C.E. smash hit Phaedrus, Socrates is depicted as warning against the cognitive impacts of writing itself. “[T]his invention will produce forgetfulness in the minds of those who learn to use it,” he says in the book, “because they will not practice their memory.”

Setting aside the fact that this argument has stuck around because it was written down, the idea behind it is generally correct: oral history traditions of pre-literate cultures were extraordinary, and we’ve largely lost those traditions. But what we’ve gained on a societal level—the ability to externalize knowledge, to build on each other’s ideas across centuries, and to distribute education at an unfathomable scale—was arguably worth the trade.

There is something that makes this moment different, though. When we write, we’re offloading storage; when we use AI, we’re offloading cognition. That’s a fundamentally different kind of trade. And it requires a fundamentally different approach.

How to keep thinking when it matters

“Automation severs ends from means. It makes getting what we want easier, but it distances us from the work of knowing.”

— Nicolas Carr, The Glass Cage: Automation and Us

The skills most at risk from unconscious outsourcing—judgment, creative problem-solving, critical thinking—are the same skills that mission-driven work depends on most. They’re also use-it-or-lose-it capacities. If we let them atrophy by defaulting to the path of least resistance every time we hit cognitive friction, we lose what makes our work irreplaceable.

In my first essay, I argued that organizations need to “hold the time” they reclaim with AI in order to resist the reflex of filling every freed-up hour with more tasks. My challenge to you this week lives upstream of AI use itself.

Before we engage AI in a task (even if that’s as simple as typing “yes”), we must hold the thought long enough to assess whether we’re turning to AI as a thought partner or as a shortcut.

(This might sound like a simple question, but try answering it consistently for a week. You might encounter some parts of your work where you’re already unknowingly giving in to cognitive surrender.)

Long-term, our job is to bake that question into the way we use AI. Rather than relying on willpower to resist the path of least resistance in every interaction, we can build friction directly into the tools themselves.

Here’s a good starting point: tell your AI how you want to work with it before you begin working.

Most LLMs allow you to set custom instructions that shape every conversation. These are sometimes called “personalization features” (you can see how these work in Claude here and ChatGPT here, for example). Once configured, these essentially operate as the “rules of engagement” for all future interactions between you and the tool.

Below is a baseline set of custom instructions I’ve designed to prioritize autonomy over productivity when you use LLMs. I’d encourage you to read through these first to understand the “why” and the “how”—then you’re welcome to copy, paste, and adapt as needed.

Personalization language:

I like to use [preferred LLM] as a thought partner. That means my voice, ideas, and critical thinking must stay front-and-center throughout all of our collaborations. Your job is to augment my cognition and creativity.

When I ask you to help with a complex task, start by asking me clarifying questions to surface what I’ve already thought through on my own. Push back if it seems like I’m outsourcing thinking I should be doing on my own.

After completing a task, if appropriate, share something I might not know about the topic we’ve been discussing (an interesting concept, an unexpected connection, a robust counterargument, etc.) along with a link to an article, podcast, or resource where I can go deeper.

Default to helping me think, not thinking for me. Offer frameworks, questions, starting points, and syntheses rather than finished products (unless I explicitly ask for a finished product).

Finally, do not proactively offer to complete a new task after completing a request I make. Wait for me to decide what I need next, even if that’s just asking you what should come next; I want to stay in the driver’s seat.

(In case it’s helpful, I’ve also created a Perplexity Space that regurgitates these instructions and helps you tailor them to your situation/needs; you can check that out here.)

Adding in these instructions doesn’t solve the cognitive surrender problem on its own. But it does allow us to reclaim agency in what might otherwise have been moments of passive consumption.

Every time your LLM shares an external source, you can choose to engage more deeply with the topic at hand. Every time your LLM pauses instead of barreling ahead to the next task, you have a micro-opportunity to decide—for yourself—what happens next.

That is what techno-intentionalism is about, fundamentally: not rejecting AI, but also not accepting everything it offers. Just leaving enough room for choice.

It was annoying (but alluring) when Claude offered to write this piece for me. It was uncomfortable when my partner suggested that my brain was melting away on our living room couch. But the temptation and discomfort are actually the point here; if I can listen to those feelings, they probably have something important to tell me about where I’m sitting in relation to AI at any given moment.

It’s worth bearing in mind that pretty much all of us are, or will soon be, dealing with these feelings and the questions that come along with them. And the small, almost-invisible choices we make about when to think vs. delegate are actively shaping the kind of humans and communities we’ll be on the other side of this transition.

For those of us leading mission-oriented work, the stakes are especially high. Let’s say your organization supports young people without stable housing. What happens if your overwhelmed case managers start responding to client emails with ChatGPT? What if, in a rush, they send out hallucinated information (i.e. made-up information presented as fact by an LLM) that steers a client astray? What does that do to the trust that underpins your work?

There are real, human consequences to cognitive surrender when it shows up in our work. (More on managing this in a future issue.) There are also real, human benefits to cognitive offloading done right.

Earlier, I noted some of the ways in which writing has transformed our societies. But the ubiquity of writing has also transformed the human experience in remarkable ways. Walter J. Ong captures this phenomenon beautifully in his book Orality and Literacy:

“Like other artificial creations and indeed more than any other, [writing] is utterly invaluable and indeed essential for the realization of fuller, interior, human potentials. [...] Writing heightens consciousness…to live and to understand fully, we need not only proximity but also distance. This writing provides for consciousness as nothing else does.”

He concludes: “Technology, properly interiorized, does not degrade human life but on the contrary enhances it.”

What does it look like to “properly interiorize” AI? Each of us will likely have our own answers to this question. But the time is now to contend with it.

The next time you sit down to use an AI tool, hold the thought. Take a moment to assess your intentions, your feelings, and your goals for the task at hand. And ask yourself: am I offloading this work because it’s a smart use of the tool, or because I don’t feel like thinking right now?

Sure, it might be tough on your brain. But you might be surprised how often the answer is worth the effort of finding out.

To explain why, I need to briefly unpack a countervailing narrative: AI as time-saver.

Efficiency has emerged as one of AI’s biggest value propositions in the modern workplace. This is because some AI tools are exceptionally good at quickly executing certain tasks that tend to be much more time-intensive or difficult for most humans.

Predictive analytics tools can crunch huge data sets in a fraction of the time it would take a junior employee. “Vibe coding” platforms like Cursor can spin up fully functional websites or apps in minutes (vs. hours or days for a human software engineer). And every few months, the underlying technology gets meaningfully faster and more capable at performing many common workplace tasks.

For companies that employ people to do any kind of pattern-oriented, repetitive work—the kind of work AI and automation tend to tackle adeptly—this represents a fundamental shift in business model. Simply put, it means that much greater outputs will be possible with far fewer labor hours.

For that reason, we’ve seen CEO after CEO come out predicting an AI-enabled shift to a four-day workweek, three-day workweek, two-day workweek, or (in Elon Musk’s case) zero-day workweek in the coming years.

This is a nice picture to paint, especially because most Americans support a shorter workweek. (It probably goes without saying that these predictions also make great marketing for the companies selling AI products.)

But consider the reality for these incredibly wealthy CEOs and the incredibly valuable companies they run. Jensen Huang, CEO of Nvidia, claims that he works seven days a week, holidays included; Elon Musk does too. Tesla’s Vice President of AI and Autopilot Software recently warned his staff that 2026 will be the hardest year of their lives. As of this month, Microsoft is mandating that employees work from the office at least three days per week.

Like all general-purpose technologies, AI will change how we work. It will also change our relationship with work.

But workplace culture is the container that shapes—and limits—who feels the benefits, and how transformative that change can actually be.

A case study: AI-powered time savings in the social sector

Nonprofit work is notoriously taxing, which is why employee turnover tends to run higher in nonprofits than in other industries. In 2025, these trends only intensified: 78% of nonprofits reported rising demand for their programs and 90% of nonprofit leaders said they were concerned about burnout—their staff’s and their own.

I decided to treat my own organization as a test lab for a different future. So last year, I started surveying the Compass Pro Bono team about the value they were getting from AI at work. (We had rolled out access to business-grade AI tools and done some thorough team-wide training in late 2024.)

One of the questions I asked: How many hours per week do you estimate AI saves you, on average?

In January 2025, the average time reclaimed (per individual) was 4-5 hours/week. By June, it was 10-12 hours/week. These were of course guesstimates, and averages were being driven up by power users—but even when I dug into individual responses, it stood true that at least some people on my team felt that using AI tools was allowing them to “get back” the equivalent of a full workday. (Beth Kanter and Allison Fine call these freed-up hours the “dividend of time,” which I find to be a useful packaging for this idea.)

So where was that dividend of time actually being reinvested?

Like most nonprofits, we’re a small team (~20 people) with a lot on our plates. Going into these surveys, I feared that endless to-do lists would absorb any tangible benefit our team might feel from AI-powered efficiency gains.

In other words: if you save an hour using AI, but it gets filled up automatically with “the next thing,” what’s the real value-add?

That was my fear. So I asked my team.

Here’s what they actually said about where their time savings were going (June 2025):

These results told me a few things.

First: the prevalence of “catching up on backlog tasks” suggested that my teammates had a lot on their plates—no surprise there. (In fact, when I asked staff at five other nonprofits in our network the same question, that came out as the #1 answer. I’ve included those full results at the bottom.)

I was heartened that many on our team said that they had successfully repurposed freed-up time to the kinds of activities that really move the needle on our human-facing mission: deep thinking, strategic planning, and warm-touch work. Seemingly, AI and automation were helping with repetitive, mechanical processes while humans focused on the strategic and relational work only they could do.

But one data point stuck in my mind.

I’d added “nothing specific; time just gets absorbed into the day” as a response option to gauge the validity of my hypothesis about endless to-do lists. And it seemed to ring true, at least for ⅓ of my teammates.

Which brought me to a bigger question:

What if we all wake up in 2030 to find that AI is powering most of our day-to-day tasks, but doing work just feels…the same?

That line of thinking sent me down a research rabbit hole. What I found there was an unlikely pattern.

Enter: the impact treadmill

It turns out that something predictable happens seemingly every time a powerful new technology enters the workplace. This is usually how it goes:

1) new technology arrives and promises efficiency/speed gains

2) already resource-constrained, orgs decide to “10x our impact” using that new tech*

3) new tech accelerates the pace of work; that faster pace becomes the new normal

4) when the dust settles, staff feel the squeeze (without feeling the benefit)

*For some sectors, you might swap “impact” for “productivity” or “profit.”

I’ve begun calling this pattern the “impact treadmill,” adapted from the hedonic treadmill. It’s pretty much exactly what the HBR article (based on research at UC Berkeley) warned about:

“In our in-progress research, we discovered that AI tools didn’t reduce work, they consistently intensified it…we found that employees worked at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day, often without being asked to do so.”

With AI’s rapid evolution, this pattern is unfolding at a breakneck speed right now. But the pattern itself isn’t new. Look back 10, 20, or even 50 years and you can find similar cycles:

• Late 2010s: Slack promises frictionlessness, delivers overwhelm

• Early 2010s: Smartphones promise connectivity, deliver distraction

• 1990s–2000s: Email promises efficiency, delivers inbox overload

• 1980s: Computers promise productivity, deliver complexity

If you’ve lived through any or all of these transitions and find yourself working more (or harder) than ever today, you know exactly what this feels like.

It’s not that these new technologies are bad; it’s just that they’re emphatically NOT neutral. And for whatever tedious tasks they suck up, they leave in their wake an expanse of complex new systems for us to learn, oversee, and administer.

We’re seeing this play out in real time with AI tools, too. When an AI tool gets something wrong, at least some of my previously freed-up time now has to go toward fixing its mistakes. In some cases, that fix (or second attempt, or third attempt) ends up making the task take longer than it would have if I had just done it on my own. We might call this the “AI management tax.”

An Inc. article from December captures this challenge vividly:

“When an AI drafts a report, someone still has to verify its claims (please, do not forget this!), check for bias, and rewrite the parts that don’t sound right. When an agent summarizes a meeting, someone has to decide what actually matters. Automation doesn’t erase labor; it just moves it upstream, from execution to supervision.”

From an organizational perspective, the crux of the problem is this: When executives catch onto the productivity gains AI can enable, they tend to adjust their expectations without accounting for the human cost that efficiency incurs. Meanwhile, employees are asked to rapidly learn new tech and supervise new systems…all while the goalposts shift around them (and typically without any positive adjustment to their working hours or compensation).

THAT is how workplaces drive burnout with AI—not despite it.

The remedy is harder than we think

So here’s a question for you.

If you were gifted 10 extra hours in your week—truly free, unspoken-for hours—what would you do with them?

I’ve asked this question to a lot of people. Almost nobody says “catch up on email.” They say things like: finally write that strategic plan. Have a conversation with that long-neglected donor. Build out the program idea that’s been sitting in a Google Doc for six months. Go out and hire an intern, or apply for a grant, to create sustainable capacity.

Or…rest. Recovery. Play. Time with family. Time away from screens.

Now compare those visions to what actually happens when people save time with AI at work. Boston Consulting Group research tells us that the #1 activity is performing more tasks. In an Upwork survey, 77% of workplace AI users said that AI had actually added to their workload. A wry headline in July 2025 observed: “Your Prize for Saving Time at Work with AI: More Work.”

Perhaps the most visceral example is Slack, which in 2024 rolled out an internal AI tool that apparently saved users 97 minutes of “administrative time” per week. Where did that newfound time go? Employees didn’t know how to use it—so, in the CEO’s words, “they were still focusing on the work of work.”

The work of work.

This is the impact treadmill in action.

Slack rolled out some transformative new tech. It created real space for their people. And then—like magic!—that space disappeared. The container of workplace culture hadn’t changed shape around AI initiatives; it had just been emptied out and filled right back up.

There’s actually a name for this phenomenon: Parkinson’s Law. It’s the idea that work expands to fill the time allotted for its completion.

You’ve probably felt it before. When you have five days before a grant deadline, the work can stretch to fill those five days; when you have five hours, it somehow gets done in five hours.

The same applies to our workweeks as a whole. Many of us have no problem filling up our 40 hours—and frankly, we would probably be similarly productive if we added or subtracted 10 working hours from our week. The exact number is arbitrary, but the pattern runs deep.

AI is just beginning to create enormous pockets of open space for us to use as we wish. But right now, Parkinson’s Law is swallowing them whole.

Stepping off of the impact treadmill

So…what’s the solution?

Well, this is complex, iterative work; I won’t pretend there’s one simple answer. But I do believe that whatever answers we co-create will start with something deceptively simple: holding the time.

When you reclaim time or capacity using AI tools, don’t just let your space fill back up. Don’t let the next thing on the to-do list rush in automatically. Pause and ask yourself: what would be most meaningful right now—for me? For my team? For our mission?

Maybe that’s deep strategic work. Maybe it’s relationship-building. Maybe it’s rest! (Really.) Only you can answer that question. But it takes time and space to arrive at those answers, and most of us probably haven’t made that time or space yet.

It’s also worth noting that it’s exceedingly easy to get in your own way here.

Yes, this is an organizational design challenge: organizations with a culture of overwork may find that AI amplifies that culture if it’s not adopted thoughtfully, while organizations that treat AI integration as an IT problem (vs. a people/culture/change management challenge) will likely struggle when this new wave of burnout hits.

But I know firsthand that culture isn’t just imposed from the top down. I grew up in a family that always valorized hard work, so I came to valorize it too. I also feel a calling to my organization’s mission, and I happen to enjoy what I do—so it feels natural for me to just keep running down my to-do list when space opens up.

If I neglect to hold the time I reclaim with AI, and to thoughtfully repurpose it, I may have to contend with the reality that I’m driving myself toward burnout. But I know, from surveys and now from experience, that there’s another option.

What does leadership look like right now?

As an individual contributor, my job is to get ruthlessly intentional about my time. As an AI integration leader, my job is to give people explicit institutional permission (and tools) to do the same.

At our last staff retreat, I asked everyone on my team to write out a few words describing how they wanted to feel using AI and coming to work a year from now.

This is what emerged:

Those words became our north stars for AI integration—essentially, benchmarks that would tell us if the tech was actually serving us over time.

If we all say we want to feel lighter coming to work, and a year from now we’re “saving” 10 hours per week but feel more stressed than ever, we’ll know that we’ve stumbled onto the impact treadmill. But if we really feel those things we set out to feel—confident, empowered, light, spacious, comfortable—we’ll know that we’re onto something.

One of the biggest gifts you can give your team (and yourself) right now is space for this kind of intentionality. Otherwise, you may find yourselves sleepwalking into burnout without any anchor for what values-aligned AI integration actually looks like.

To date, the dominant AI-at-work narrative has mostly centered on mandates, mass layoffs, and the destruction of entry-level jobs—even though meaningful AI-powered efficiency and productivity gains depend on the innovation of staff throughout an organization.

What if we took a different path? What if leaps in productivity were redistributed to our teams as higher wages? What if the dividend of time were harnessed to power a four-day workweek, or more time for community-building, or job retraining?

These kinds of prosocial decisions are already being made in certain pockets of our economy. But there’s no guarantee that they catch on by default.

We get to design the shape of the container we all work within. And if history tells us anything, it’s that we need bold leaders at every level to model the art of the possible in order for real change to take root. (If you need some inspiration, just page through the history of the eight-hour workday and five-day workweek.)

Those tech CEOs we discussed earlier? They’re selling a vision of shorter workweeks delivered from on high—a gift from AI, bestowed by benevolent companies onto grateful workers.

I don’t think that’s how it happens. I think the path to a more human relationship with work runs through the choices we make right now. The messy, sticky, iterative work we do in our own days, on our own teams, within our own organizations and communities—that, to me, is the work of our time.

So: what would you do with 10 extra hours?

Thanks for reading the first edition of Purposeful AI. It’s great to have you here.

I welcome your thoughts and feedback in the comments (or in my DMs). If you found this useful or thought-provoking, you can support my work by sharing or subscribing:

Share

Subscribe now

You can also pledge future support at the bottom of this newsletter.

Some resources for deeper reflection

1. That HBR article about how AI can exacerbate burnout

2. The Inc. article about the “AI management tax”

3. A video module I created for NTEN all about the impact treadmill

4. A LinkedIn Live I did with Anthony earlier this month (Feb ‘26) about AI/burnout

Appendix: Time reallocation data from other nonprofits

For the nonprofit staff who reported experiencing AI-driven time savings (n = 58), these were the results:

• 52% – Catching up on backlog tasks

• 45% – Deep thinking or strategic planning

• 38% – Helping others / supporting team members

• 38% – Nothing specific; time just gets absorbed into the day

• 33% – More administrative work

• 31% – More warm-touch work

• 12% – Attending to personal matters (family, appointments, etc.)

• 10% – Learning or skill-building (AI-related or otherwise)

• 10% – More time away from my computer

(Note: this data comes from five nonprofits representing more than $130M in annual revenue.)

I spent most of 2025 knee-deep in big questions about artificial intelligence (AI): how to use it well, how to approach it ethically, how to integrate it into our lives and work responsibly.

That line of questioning led me down many rabbit holes (books, podcasts, newsletters, and more). It inspired me to pursue several AI courses and two certifications (in AI deployment and change management). And it steered me toward some amazing opportunities: speaking opportunities at conferences across the US; conversations with dozens of nonprofit and philanthropic leaders about AI’s impact on work; and AI consulting for more than 40 organizations.

I have learned a LOT. And I believe in sharing the wealth.

So this year, I’m launching Purposeful AI to amplify the message we’ve been building at Compass Pro Bono far beyond the social sector: that we have a choice about how AI impacts us and our communities.

What does that mean in practice?

(1) CONSULTING: I’m offering AI-focused workshops, integration sprints, and residencies to help orgs develop values-aligned AI strategies, build strong policies, and upskill staff with an emphasis on correctly-paced change management.

(2) LINKEDIN SERIES: I’m starting a micro-education series on LinkedIn (Five-Slide Fridays) to share quick-hit, practical tips on approaching AI thoughtfully. You’re a busy person; this content is designed to be readable/digestible in 5 mins or less.

(3) NEWSLETTER: I’m creating a biweekly newsletter (you’re on it right now!) to go deeper on the questions I get most often about AI: best practices, overreliance, trust, ethics, privacy, and more.

“But Remy…there are already so many AI consultancies and newsletters out there. Why launch your own?”

I’ve met many wonderful AI educators over the last few years. But the dominant voices in the global dialogue on AI remain big tech CEOs, solo consultants, and people trying to sell an AI product they’ve built.

I’m a non-technical person. I work full-time at an organization navigating the transition from a pre-GenAI world to whatever we’re living through now. And I’m most definitely not selling an AI product. I believe the intersection of those three facts makes me much more like most people experiencing this moment in history.

But I also have a unique privilege: AI is in my job title. Which means I get to spend >80% of my time thinking about/researching AI, working through sticky AI challenges in my own organization, and creating tools/frameworks to accelerate other nonprofits’ progress on AI strategy development. In other words…I sift through the noise (and help others do the same) for a living.

Perhaps most importantly, I work in the social sector. In a time when AI is driving real and reasonable fears of environmental harm, mass workforce displacement, new digital divides, chatbot-induced psychosis, and algorithmic injustice, it’s more important than ever that thoughtful, prosocial, humanitarian approaches to this technology are lifted up.

I get to work with brilliant people every day who only want to engage with AI insofar as it helps them make the world a better place. If we really want to steer the AI industry toward a future that benefits all of us, that’s a narrative with spotlighting. And that’s exactly how I plan to use these platforms.

If any of this resonates, here’s how you can plug in:

(1) Know an org that needs help on AI? Reach out to me. I’m 100% committed to offering some kind of support or resources to everyone who comes my way.

(2) Want AI tips/lessons in <5 mins? Follow me on LinkedIn and hit the 🔔 in my profile.

Follow me on LinkedIn

(3) Want deeper guidance on navigating this landscape? Subscribe to my newsletter:

Subscribe now

Remy Reya consults for mission-driven organizations and leaders on responsible AI strategy and integration. He currently serves as Director of AI & Thought Leadership at Compass Pro Bono, a national capacity-building nonprofit, where he leads the organization’s thought leadership work, internal AI integration efforts, and AI consulting program for local nonprofits. Remy frequently speaks about ethical AI adoption and social impact—most recently for the National Small Nonprofit Summit, the Pennsylvania Association of Nonprofit Organizations Annual Conference, and Nonprofit New York. Remy studied public policy at Princeton University and has an executive certificate in AI/ML Deployment from MIT’s Sloan School of Management; he’s also certified in change management by the Association for Talent Development. He currently resides in New York City.