Everyone asks which tool got us there. It was not the tool.

The 7% problem.

The number comes from Virtuous’s 2026 Nonprofit AI Adoption Report. 92% of nonprofits report using AI. 7% report real impact from it. The gap between those two numbers is the story.

The obvious question is what is different about the 7%. The answer most people reach for is that they have better prompts, better tools, more technical staff, bigger budgets. None of that is what we found.

I believe what is different about the 7% is what was already there before AI arrived.

AI did not make us effective. AI gave us a way to multiply the effectiveness we had already spent three years building. The orgs that get real value from AI are the orgs that are already serious about how they spend every hour.

What we built before AI was even an option.

In 2022, as a nimble startup with no need for fancy automations and time saving tools, I built the unsexy infrastructure first.

A coach prospect would fill out an application form. Zapier piped that submission straight into Salesforce, where the contact was created automatically and slotted into the right stage of our coach pipeline. Mailchimp picked them up from there and ran them through a drip sequence designed to warm them up to what coaching at Level Up actually looks like. Calendly sat at the bottom of every email, so a prospect could self-schedule an intro call without anyone on our team coordinating a single calendar.

The Salesforce pipeline tracked every prospect from application through active coach. Stages, dates, drop-off points, all visible at a glance.

All that automation, probably wasn’t necessary but the stack cost us almost nothing - and we were planning on growth. ChatGPT was not public yet. Claude did not exist. None of this needed AI, but all of it needed thinking.

The principle I kept coming back to: how do we do this without touching it as much as possible? Every manual step we eliminated was a step we did not have to staff up to do later. That mattered when it was just me. It matters more now.

The proof showed up over time. Since 2022, more than 40 coaches have come through that pipeline. Roughly 80 prospects in total. 82% of them self-scheduled their own intro call. The 20% who required manual outreach were mostly no longer interested by the time we got to them. The pipeline was qualifying as it routed.

The system itself was doing the filtering.

Nobody has chased an applicant. Nobody manually entered applicant data. Over four years, it’s part of a staffing line we never had to add.

The accelerant we layered on top.

A year later, AI showed up. We did not throw the pipeline out. We sharpened every layer of it.

We used NotebookLM to generate the first episode of a series called Level Up Cincinnati: Inside the Program. The episode, The Impact of Coaching is hosted on Spotify and embedded as the first thing every new prospect sees in their welcome email.

A coach prospect can now hear the program described, in conversation, before they ever talk to a human.

The drip sequence itself got a rewrite. I worked with Claude to tighten the copy, sharpen the call to action in each email, and tune the cadence based on what we were seeing in open rates and click-throughs.

The analytics layer is the one that surprised me most. I had Claude write a lightweight script that pulls Mailchimp engagement data, summarizes click-through patterns, and lands a weekly report in my inbox. Used to be a thing I checked when I remembered to. Now it checks itself.

Each one of these only worked because the pipeline beneath it already existed. The podcast is useless if no one is getting the welcome email. The tighter copy is useless if Zapier is not delivering prospects into Mailchimp in the first place. The weekly analytics report is noise if there is no funnel underneath to measure.

AI did not give us the pipeline. AI gave the pipeline a podcast, a sharper script, and a weekly checkup.

Last week, we did it again.

For four years that pipeline did one job extremely well: get coaches in the door. But the moment a coach said yes, the automation stopped and the manual work started.

Onboarding meant a chain of links and Adobe Sign forms going out by email. Someone had to send them. Someone had to chase the people who did not complete them. Someone had to file the signed copies into the right Google Drive folder. Someone had to know where every coach and every scholar stood in the process.

Last week, I pushed the same system downstream to fix that, for coaches and scholars alike.

That work is now gone.

A coach or scholar logs into the Level Up app and gets prompted to complete onboarding. They click through each document and sign it in the app. Webhooks fire as each signature lands. The signed document files itself into the correct Google Drive folder automatically. Their status updates in real time on an admin dashboard, so our Director of Student Success can see at a glance exactly where everyone stands. No hand-holding. No chasing.

The follow-up layer is where AI does the work. If someone falls behind, a single click generates a personalized email with how much time is left before Match Day, exactly which forms they still need to complete, and where they are in the process. Then it sends.

Every piece of that workflow used to be manual. Every piece of it is now automated.

Same playbook I used in 2022, four years later and one stage further down the journey: audit the workflow, design out every manual step, layer in AI where it earns its place. The foundation did not just hold. It compounded.

A multiplier, not an upgrade.

The Virtuous report frames the gap as an AI problem. The gap is an operations problem. The orgs in the 7% did not get there by buying smarter software. They got there by doing the boring work first. Mapping their processes. Cleaning their data. Building the foundation that an automation layer could sit on.

Most of the 92% are running pilots on the latest tool, building chatbots on top of broken intake forms, layering generative output on top of processes nobody has audited in five years. The tool is fine. The foundation under it is the problem.

AI is not an upgrade. It is a multiplier. And zero times anything is still zero.

Start with the workflow, not the tool.

The audit I did in 2022 is the same audit I would do today. I walked the coach intake process end to end and wrote down every place a human had to touch it. That list was the map. Every item on it became a manual step I designed out before I built anything else. The pipeline this post just described is what came out of that exercise.

That work is the work. It does not require AI. It requires honesty about where your time is actually going.

A broken workflow with AI on top is a broken workflow that runs faster. A working workflow with AI on top is leverage.

The orgs in the 7% were already serious about how they spent every hour before the tools arrived. That is not a credential you buy. It is a habit you build.

We are not an AI nonprofit

We are an operational nonprofit. AI is the latest layer we built on top of work we have been doing for four years.

The 92% have the same tools we do. What they do not have is the four years we spent auditing every workflow and automating every manual step before the tools arrived.

The toolbox.

The whole thing runs in two connected stages.

Getting coaches in the door (2022):

• Salesforce: pipeline stages, contact data, source of truth

• Zapier: form-to-CRM connection, no manual data entry

• Mailchimp: prospect drip sequences

• Calendly: self-scheduled intro calls

• NotebookLM + Spotify: an AI-generated welcome podcast, embedded in email

• Claude: drip copy rewrites and a weekly analytics report

Onboarding coaches and scholars (last week):

• Salesforce: still the source of truth; flags when someone enters onboarding

• Firebase: the engine that moves each person through in real time

• Dropbox Sign: in-app document signing for commitments, waivers, and releases

• Google Forms: pre-program and pre-match questionnaires

• The Level Up app: where coaches and scholars complete onboarding

• Admin dashboard on Vercel: live status for our Director of Student Success

• Postmark: reliable, branded email

• Claude: drafts the personalized nudge when someone falls behind

Most of it runs on free tiers and nonprofit pricing. The rest is a few modest paid tools that earn their keep.

What comes next.

This is the last post in The $0 Department on a regular cadence. Nine posts in, I think the case is made: a 1.5-person nonprofit can turn what used to be overhead into student impact by getting serious about how we work.

I am not done building. The cadence just shifts. From here on, I will share what I build when there is something worth sharing, not on a schedule.

Stay subscribed if you want to follow that. If you only signed up for the series, thank you for showing up.

---

This is the last regular post in The $0 Department: What Happens When a Nonprofit Stops Hiring and Starts Building. Subscribe to catch what comes next.

Level Up Cincinnati empowers first-generation college students through mentorship, scholarships, and professional development. 100% of donations go directly to student programming. Learn more at https://levelupcincinnati.org

In February, we ran a comprehensive board governance audit. Thirty minutes. The thing I had assumed required a board member’s expertise turned out to be the kind of work AI does in one sitting.

What that did to my mental model of when small nonprofits actually need legal help is the part worth writing about. The audit was the proof. What came out of it was a rule we now use every time AI gets close to anything legal.

What most small nonprofits do about legal.

Legal counsel for nonprofits of our size is largely unaffordable. We are fortunate to work with our friends at Pro Bono Partnership of Ohio who provides pro bono legal counsel, but often that lies behind weeks of lead time.

Most of the ED’s I talk with fall into one of two camps. They ignore the governance work until something forces it. Or they try to recruit a lawyer onto the board to get advice in the room.

The question is not whether you can afford a lawyer. The question is whether the work you are about to do actually needs one.

For a most work that small nonprofits agonize over, the honest answer is no. We just had no way to know that until AI gave us a way to test it.

Thirty minutes, all of it forensic.

I pointed Claude at the Google Drive folders where every governance document we owned actually lived.

Meeting minutes. Bylaws. Conflict of interest statements. Articles of Incorporation. Board composition rosters. Financials. The Q4 2025 email update that quietly replaced a meeting.

Then I asked it to do the part that takes a human a week. Cross-reference every document against every other document. Find what was missing. Find what was inconsistent. Find what was technically there but stored in the wrong place.

What came back was a categorized list, and most of it was not what I expected.

Some of the surfaced gaps were filing problems. Documents that existed but lived in the wrong folders, invisible to anyone who looked. Some were inconsistencies, governance titles used three different ways across documents that referred to the same role. Some were genuinely missing. Policies that recognized governance frameworks recommend for organizations at our stage, but we did not have on file.

The thirty minutes produced the inventory. My board director and I closed out the real gaps the following week.

Our legal partner did not see this audit. They did not need to.

The most useful thing AI did was not generate new work. It made our existing work findable.

What AI did not do.

AI did not write a binding document.

It did not give a legal opinion.

It did not stand in for our legal partner.

The output was an internal artifact that helped us see ourselves clearly. That is the distinction that gets blurred whenever people talk about AI and legal work.

Blurring it is what gets small nonprofits in trouble in both directions. Treating AI as a universal substitute creates real risk. Treating AI as off-limits keeps you paying for work that no longer needs paying for.

The rule we use lives between those two mistakes.

The three-tier rule.

Here is what we use now.

Tier 1

The work is analytical. The output stays inside the organization. Governance audits. Document inventories. Policy gap analysis. Board materials. First drafts of internal-use documents. AI does the heavy lifting. A human reviews the result before it informs any decision. No lawyer required, because nothing the AI produces is binding or relied on by anyone outside the org. The board audit at the top of this post is what Tier 1 looks like in practice.

Tier 2

These documents will be used by or relied on by external parties, but the form is well-established. Liability waivers. Website terms and conditions. Privacy policies. Basic vendor agreements. AI can produce a clean draft because the templates are stable. A lawyer signs off because external reliance and enforceability require professional judgment. The cost equation flips. You pay a professional to review substance instead of drafting language. The time required is generally hours, not days. Our liability waivers and website terms live in this tier.

Tier 3

The stakes are high, the facts are unique, the jurisdiction matters, or the organization needs an advocate. Foundational legal setup, the kind of work that defines what the organization is. Personnel matters with real exposure. Negotiated contracts with significant downside. AI cannot exercise judgment, cannot represent the organization, and cannot account for the local nuance that makes these situations turn out one way or the other. AI assists. A lawyer leads. Behind the human, never in front of them. Our 501(c)(3) formation work lived here. A lawyer led it then. A lawyer would lead it today, even with AI in the room.

Many small nonprofits do not have a tier system. They have one stance for everything. That is exactly the problem.

The single-stance approach costs you on both ends. If your stance is “AI for everything,” you walk into Tier 3 work with a tool built for Tier 1. If your stance is “AI for nothing legal,” you keep paying for Tier 1 work that AI now does better and faster than the people you used to hire for it.

The tier system is the discipline that lets you use AI seriously without using it stupidly.

AI changed who we need to recruit.

For many small nonprofits, “we should get a lawyer on the board” is a perceived gap that never quite gets solved.

The audit demolished that thinking for us.

Tier 1 work is much of the document-review work a legal-credentialed board volunteer would handle on a given quarter. Document review. Policy gap analysis. Cross-referencing what is on file against what should be on file. AI does that in thirty minutes, with no calendar coordination required.

Which means the real gap is not legal expertise. It is whatever skill mix actually matches where the org is going next.

AI did not replace lawyers in our work. It changed who we need to recruit to govern it.

My honest take.

AI has been right almost every time we have used it for this work. Right on the audit. Right on first drafts of harder documents. Right on the cross-referencing.

That pattern creates a temptation, and the temptation is the actual risk.

Not that AI gets things wrong. That humans, watching it be right repeatedly, stop checking.

The next document you do not bother to send to legal review. The next policy you skip the second pass on. The next moment you say “Claude was right last time, we are fine.” That is where the failure lives.

The three-tier rule is the discipline that keeps the temptation from becoming the failure. It is not interesting. It is not flashy. It is the thing we do anyway, because the alternative is being lucky until we are not.

Being right almost every time is exactly when discipline matters most.

What to do on Monday.

Two questions for any nonprofit ED reading this.

First, where does your governance documentation actually live, and could AI cross-reference it tomorrow morning? If the honest answer is “across three Google Drive folders and a filing cabinet,” that is the first job. You do not need new tools. You need the tools you have to be able to see your own work.

Second, do you have a rule for what AI gets to touch alone and what gets a human stamp? If you do not, the rule comes before the next AI experiment, not after.

We do not need a lawyer on the board. We need clarity. Then we know exactly when to call one.

This is post seven of The $0 Department: What Happens When a Nonprofit Stops Hiring and Starts Building. Subscribe to follow the full series.

Subscribe now

Level Up Cincinnati empowers first-generation college students through mentorship, scholarships, and professional development. 100% of donations go directly to student programming. Learn more at levelupcincinnati.org

Not because it was broken. Because we had new data points about what coaching at Level Up actually looks like, a stronger story about why it matters, and I wanted to tell it. The old page was fine. It just wasn’t us anymore.

It took me about twenty minutes.

Six months earlier, that same update would have taken a full day. Maybe two. And honestly? I probably would have skipped it. Left the old page up. Moved on to something more urgent. Not because I didn’t care about how we presented coaching to the world. Because the effort required to change one page on our website made it not worth the fight.

The cost of a hard-to-update website isn’t what your site looks like. It’s what never makes it onto your site at all.

The updates that didn’t happen.

Before I get into what changed, I want to sit with that idea for a second.

How many updates did I skip over the years because the friction wasn’t worth it? Event landing pages that stayed as basic placeholders. An impact report that lived as a PDF for two years because building it as a web experience would have taken weeks of design work. Metrics that sat in a spreadsheet instead of on the homepage because updating a single number meant rearranging an entire section.

None of these were emergencies. The site still worked. It loaded. It had our information on it.

But it wasn’t keeping up with the organization. Level Up was growing, evolving, getting sharper about who we are and what we offer. The website was a snapshot of the last time someone had the bandwidth to deal with it.

That someone was me. And I had a lot of other things on my plate.

Drag, drop, settle.

We’ve been on Squarespace since the beginning. And to be fair, Squarespace is a solid platform. It lets you build a decent-looking website without knowing how to code. Millions of organizations use it for exactly that reason.

But here’s the reality. Every page was a project. Click and drag this block here. Adjust the spacing. Check how it looks on mobile. Realize the template doesn’t quite do what you need. Work around it. Settle.

Hours per page. Sometimes a full day for something that should have been simple.

And the output, even after all that work, looked like every other Squarespace site built on the same template. You could customize it within the boundaries the platform gave you, but those boundaries were tight. The animations, the interactive elements, the design touches that make a site feel like yours and not a template. Those weren’t options.

So you ended up with a website that was functional and generic. It worked. But “works” and “represents who you are” aren’t the same thing.

Section by section, then all at once.

As I started working more with AI, I noticed it could produce solid HTML. Clean, responsive sections with the right colors and fonts if you told it what you needed. So I started there. One section at a time. I’d describe what I wanted, provide Level Up’s brand colors and typography, and get back something that looked better than what I could drag and drop in Squarespace.

That was the first unlock. But every new section meant re-explaining the brand from scratch. Colors, fonts, spacing preferences, the overall feel. It worked, but it was repetitive.

Then the tools improved. I was able to build skills in Claude where it already knew our brand guidelines, our design aesthetic, our website patterns. No more re-explaining. I’d say “build a section for coach testimonials” and it already knew what Level Up looks like.

That changed the economics entirely. I went from building individual sections to producing full pages. Then I took the biggest step. I created a VS Code project on my desktop that maps every page on our website in HTML. The entire site, structured as code, living in one place.

Now when I need to make an update, I tell Claude which page needs work, describe what I want, and it makes the change. I copy the HTML and paste it into Squarespace. What used to be a day of dragging blocks is now a conversation.

And here’s where it’s going next. We’ll eventually move away from Squarespace entirely to something like Vercel, where Claude can push changes directly. No copying and pasting. I’ll describe the update, probably through voice using Wispr, and it deploys. The only thing keeping us on Squarespace right now is a few MailChimp form integrations we haven’t migrated yet. That’s a small project we’ll get to.

The trajectory is clear. What started as “help me build one section” is becoming “I talk, the site updates.”

The impact report that became a website.

The best example of what this unlocks is our annual impact report.

For years, impact reports meant the same thing for every nonprofit. Design a PDF. Spend days getting the layout right. Export it. Post a download link on your website. Hope people actually open it.

This year we did something different. We built the entire impact report as a live website. Interactive. Animated. Updated in real time as our numbers change. It lives on our site at impact.levelupcincinnati.org and it’s a fundamentally different experience than scrolling through a static PDF.

Donors can explore our metrics. Prospective coaches can see the program’s reach. Board members can share a link instead of attaching a file. And when our data changes, I update it. Not in a week. Not after a redesign cycle. That day.

This project, built as a standalone site by a developer or design agency, would have been a separate budget line. Probably several thousand dollars. We built it ourselves, and it’s more engaging than anything we would have gotten from a template.

That’s the thing about lowering the barrier. You don’t just do the same things faster. You start doing things you never would have attempted.

A hundred dollars a month.

Here’s the direct version of this.

If you’re paying a web developer or an agency to maintain your organization’s website, I can’t think of a structured argument for why that still makes sense. Not for a standard site with content pages, event information, forms, and basic interactivity.

A Claude subscription is a hundred dollars a month. The design skills that used to require a professional are now conversational. The technical barriers people cite, hosting, repositories, deployment, are solved by the same tool that writes the site. You don’t need to know how to code. You need to know what you want and be able to describe it.

The learning curve is light. And it gets easier every month because the tools keep improving. Things that were clunky weeks ago are smooth now. Designs that required multiple rounds of iteration six months ago land closer to right on the first try.

If your website is genuinely complex, a technical web application with custom backends and integrations, that’s different. But most organizational websites aren’t that. They’re content, forms, and a clear presentation of who you are. You can do that yourself now.

The person on your team who “doesn’t have the skills to build a website” has the skills to describe what they want. That’s the same thing now.

Showing up in a nice suit.

There’s something we believe deeply at Level Up. How you show up matters. First impressions are formed in seconds. The confidence that comes from knowing you’re presenting your best is real. It opens doors that don’t open otherwise.

Your website is your organization’s suit.

A prospective coach hears about you and looks you up. A donor checks your site before deciding whether to give. A university partner is evaluating whether you’re credible enough to work with.

Every one of those moments is a first impression. And if your site is outdated, generic, or clearly untouched for months, you’re starting behind before anyone reads a word.

For a small nonprofit, a polished and current web presence signals that you operate at a level your size wouldn’t suggest. Donors see a professional organization. Coaches see a program worth their time. Partners see someone they can trust.

The key word is current. Not polished once and left to age. Reflecting who you are right now, not who you were the last time someone had a free afternoon.

Good enough doesn’t matter anymore.

The toolbox.

Here’s what I use to build and maintain levelupcincinnati.org:

• Claude: AI partner for writing HTML, designing sections, and iterating on pages

• VS Code: project workspace mapping every page on the site

• Squarespace: current hosting platform (migrating to Vercel)

• Wispr Flow: voice-to-text for describing updates naturally

• Level Up brand skill: Claude remembers our colors, fonts, and design patterns

This is the sixth post in The $0 Department: What Happens When a Nonprofit Stops Hiring and Starts Building. Subscribe to follow the full series.

Subscribe now

Level Up Cincinnati empowers first-generation college students through mentorship, scholarships, and professional development. 100% of donations go directly to student programming. Learn more at https://levelupcincinnati.org

Four astronauts just flew around the moon for the first time in over 50 years. But here’s a detail most people don’t know about the program that got us there the first time.

NASA still has the blueprints for the Saturn V. Every schematic, every technical drawing of the F-1 engines that powered it. All archived. But when a team of young engineers disassembled old F-1 engines to study them, they found something the blueprints couldn’t tell them.

Every engine had been handcrafted. Skilled welders spent full days on single welds. Engineers solved problems on the fly and scribbled fixes on scraps of paper that were never filed. The knowledge of how to actually build the thing lived in the hands and heads of the people who made it.

By 1973, when Apollo ended, that team scattered. Retired, reassigned, moved to the private sector. Their knowledge, unwritten and irreplaceable, went with them.

The engineers studying those old engines decades later had never met the people who built them. The blueprints told them what. Nobody was left to tell them how.

That’s a story about a rocket. But it’s really a story about every organization that depends on what’s in someone’s head.

What keeps me up at night.

Level Up Cincinnati has 1.5 employees. I run virtually every operational function. Our Director of Student Success, JeRod, is part-time.

Here’s the honest question I ask myself more than I probably should: what happens if I’m no longer able to lead Level Up?

Not morbid. Practical.

Who knows how coach matching works? Who knows the financial reporting cadence? Who knows why the Salesforce fields are structured the way they are, or how the scholarship disbursement process runs, or what our grant writing system actually looks like under the hood?

The answer, right now, is mostly me.

Our board knows this. They’ve flagged it as one of our biggest organizational risks. And they’re right. We built something special at Level Up. Our scholars depend on it. Our donors trust us with their dollars. That trust has to survive any leadership transition.

This is the daily reality of a lean nonprofit. And if you’re running a small organization, I’d bet it’s yours too.

The mentor who wouldn’t let it go.

When we started Level Up, I got introduced to Tony Aloise.

Tony spent 20 years at Procter & Gamble building processes and procedures across global business units. When he retired, he turned that same discipline toward his real passion: mentorship. He now works with nonprofits across Cincinnati for free, helping them build the infrastructure to grow. The Mentor Impact award in Cincinnati carries his name.

Tony became a true mentor for me. Over multiple sessions, he pushed me to think critically about how Level Up would grow. And every single time, he brought it back to the same place.

What are your policies and procedures? How do they live on? How do you set yourself up for growth and make sure everything is documented?

Every time we’d meet, he’d re-ground me on this. And every time, my response was the same.

I agreed with the principle. I just hated the execution.

The problem with policies and procedures.

Tony was right. He was always right. Policies and procedures are the lifeblood of how an organization runs. They’re what make your best thinking repeatable and your operations transferable.

But here’s what I kept bumping up against: traditional policies and procedures become stagnant documents that sit around and nobody pays attention to.

Think about every organization you’ve ever worked at. There’s a Google Drive folder somewhere with documents that were written with good intentions. Maybe a policy manual. Maybe an employee handbook. Maybe a set of SOPs that someone spent weeks creating.

Nobody reads them. Nobody updates them. They collect dust. And when a new person comes in and needs an answer, they don’t dig through a folder. They ask someone.

Which brings you right back to the problem. The knowledge lives in someone’s head.

I believed in what Tony was preaching. I just needed a better way to execute it.

From static manual to living knowledge center.

I couldn’t solve that puzzle until recently.

We’re currently building a Knowledge Center for Level Up Cincinnati. Not a Google Drive folder full of SOPs. A living wiki where every policy, procedure, and piece of organizational knowledge is structured, searchable, linked together, and easily updated.

Don’t think of it as a document. Think of it as a brain.

Here’s how it works. I open up Claude, review which sections we’re working on that day, and then I just talk. I dictate how a process actually works. Claude asks follow-up questions about anything it doesn’t have context for. When it has a complete picture, it writes the page and automatically links it to every related section across the wiki.

It publishes live, immediately. No formatting. No layout work. No bottleneck.

The entire knowledge center will be complete in under 30 days. 53 pages covering every part of how Level Up operates. Mission and culture. Fundraising. Finance. Operations. Board governance. Partnerships. Marketing. Events. Student curriculum. Coach management. Technology. All of it.

And here’s what makes it fundamentally different from a policy manual: it’s alive. When something changes, I tell Claude what changed. The wiki is current within minutes. No version control headache. No “wait, is this the latest document?” No rewriting a 40-page manual because one process shifted.

It’s not just a reference. It’s a co-worker.

The next layer is what gets me most excited.

Someone joins Level Up tomorrow. They need to onboard a coach. Instead of sitting with me for an hour while I walk them through it, they ask an AI agent: “How do I onboard a coach?”

It pulls from the wiki, walks them through the steps, and points them to the resources they need. It’s essentially me, available at all times, without me being there.

That’s forever. That’s not just while I’m around.

Here’s a real example from yesterday. JeRod came to me and said our coaches would love a monthly email dedicated to them. Content specific to their mentoring, what’s happening with their students, things they should be thinking about.

Before, that might have felt daunting. Create the content. Draft an email. Design it. Do this every month. Hours of effort.

Here’s what it actually looks like now. I’ll build a template with Claude. Design it once through iteration until it’s right. Then each month, JeRod and I sit down, dictate what we want coaches to know, and Claude produces a finished email I drop into Mailchimp. Twenty minutes, once a month. And it’ll get better over time.

We’re calling it the Coaching Compass.

And here’s the part that ties it all together: this new process will be documented in the wiki too. When someone joins our organization, they’ll know that monthly we send the Coaching Compass, here’s the template, here’s what they look like, here’s how to produce the next one. All linked. All current.

We don’t just have a new process. We have a new process and full documentation of it. At the same time.

That’s the loop. That’s what makes the knowledge compound instead of evaporate.

The scraps of paper.

For years I carried two things that didn’t fit together. The belief that Tony was right about process and procedure. And the certainty that the way organizations actually do it is broken.

I don’t carry that tension anymore.

Those Apollo engineers scribbled their fixes on scraps of paper because documenting it properly would have slowed them down. They were building something that mattered, under pressure, with limited time.

Sound familiar?

That’s every nonprofit leader I know. You’re moving too fast to stop and write the manual. And by the time you could, half of it’s already changed.

The tools caught up. What used to take weeks of technical writing now takes a conversation. What used to become a static document nobody reads now becomes a living system that grows with your organization. What used to be the most fragile part of any organization, the knowledge in someone’s head, can now be the most durable.

The barrier isn’t time. The barrier isn’t cost. The barrier is deciding to start.

NASA lost fifty years because no one captured what the engineers knew. Your organization doesn’t have to lose a single day.

The toolbox.

Here’s what we used to build the Knowledge Center:

• Claude Code: AI partner for writing wiki content, creating links, and structuring pages

• MkDocs + Material theme: wiki framework

• GitHub: version control and content storage

• Vercel: hosting with automatic deploys on every push

• Wispr Flow: voice-to-text for dictating processes

• Obsidian: organizational knowledge base that feeds the wiki

Less than $125 in monthly cost.

---

This is the fifth post in The $0 Department: What Happens When a Nonprofit Stops Hiring and Starts Building. Subscribe to follow the full series.

Level Up Cincinnati empowers first-generation college students through mentorship, scholarships, and professional development. 100% of donations go directly to student programming. Learn more at https://levelupcincinnati.org

That’s the question. Not “what’s Jada’s GPA?” or “did Jada fill out her survey?” The whole picture. Is she thriving? Is something off?

Is her coach relationship strong, or is there a gap we’re not seeing?

For years, answering that question meant jumping into multiple spreadsheets, pulling transcript data from one system, survey results from another, and getting on a call with a coach to ask how things felt.

We were already doing the work. Touching base with every student. But there was no single place to go and see it. No way to check in at a glance.

We always knew this is where we needed to get to.

The foundation we didn’t build alone.

When we first adopted Salesforce, we needed help. Real help.

Salesforce is free for nonprofits. That’s one of the best kept secrets in the sector. But free software and a functional CRM are two very different things. Setting it up in a way that actually maps to how your organization works - that takes expertise.

Paul Delaney, president of AMEND Consulting and a Level Up board member, was quick to offer their team’s services. AMEND is a Cincinnati-based consulting firm specializing in business operations, enterprise technology, financial planning, and M&A integration. They also happen to be on the forefront of AI adoption for businesses in the region.

As a corporate partner, AMEND didn’t just write us a check. They rolled up their sleeves.

They worked closely with us to understand our workflow and then built the structure. A coach recruitment pipeline. Integration with Givebutter, our donation platform. Grant tracking. Sales funnel pipelines for corporate partners, and donors.

That work was essential. Most nonprofits don’t have an AMEND in their corner or a consultancy willing to give that kind of time and expertise. We were fortunate. I don’t take that for granted.

The bones AMEND built are still the backbone of our operations. We still reach out to them when we need an answer or a quick hand. That kind of ongoing partnership is invaluable.

Here’s what changed.

The landscape shifted. Fast.

Three years ago, I couldn’t have done what AMEND did. I didn’t have the Salesforce knowledge, and the tools didn’t exist to bridge that gap.

Today they do.

In the last three months, AI has gotten to the point where I can work directly inside Salesforce, creating custom objects, building fields, designing dashboards, and deploying code. Not by following tutorials. Not by learning Apex from scratch. By describing what I need and having AI handle the engineering.

Updates that required a specialist three years ago, the tools can now handle. That changes the equation for every nonprofit, especially the ones who don’t have an AMEND on their side.

That shift happened faster than I expected. And it’s still accelerating.

Four days. Zero lines of code I wrote.

We wanted dashboards. Not just data in fields, but visual indicators that told a story at a glance.

Here’s what we built using Claude Code as an AI development partner:

• A Scholar Dashboard on each student’s contact record: relationship health, GPA trends, satisfaction ratings, coach pairings, employment status, and milestones

• A board-ready Student Health Dashboard showing program-wide retention, GPA, employment, graduation, satisfaction, and scholarship totals

• A data pipeline connecting Google Forms to Salesforce through Zapier. When a coach or scholar fills out a semester check-in survey, the data flows automatically into the right fields. No manual entry.

8 custom components, 118 commits. I didn’t write the code. I described what we needed, made the architecture decisions, and reviewed the results. AI wrote the Apex controllers, the Lightning Web Components, the test coverage - all of it.

The insight that made it worth it.

This is the part that made the whole project click.

Our semester surveys ask both coaches and scholars to rate their relationship on a scale.

It’s a five-stage framework: Initiating, Experimenting, Intensifying, Integrating, Bonding. Both sides rate independently.

Now imagine a coach rates their relationship as “Intensifying.” They feel good. Things are moving.

The scholar rates it as “Initiating.” They’re still figuring things out.

That gap used to be invisible. We might catch it eventually through a conversation, or we might not. Now it’s on a dashboard. We can see it the moment the surveys come in. And we can talk to the coach about it before it becomes a retention problem.

We didn’t build a better CRM. We built a window into how every scholar is actually doing. The data was always there, scattered across surveys and calls and gut feelings. Now it’s organized, visible, and actionable before anyone has to ask.

Where AI hit the wall.

It wasn’t seamless. Salesforce wasn’t built for AI to navigate.

The admin interface buries its buttons and fields behind layers of protection. Normal automation can’t reach them. We had to write custom workarounds just to break through.

A couple of examples:

• Salesforce flat-out refused to let the AI create a new field. “Insufficient Privileges.” I did that one myself.

• The same type of button had different code behind it depending on the page. No consistency. No pattern to follow.

None of it was a dealbreaker. But it’s worth naming because this is how AI development actually works right now.

You’re the architect. AI is the builder.

When the builder hits something it can’t solve, you don’t scrap the project. You grab a wrench and keep going.

Everything we built in four days sits on top of the foundation AMEND put in place.

The AI didn’t replace the expert work. It made it compound.

Every organization has a gap between the tools they pay for and the way they wish they actually worked. Closing it used to require a specialist.

Today it requires a question: what do I wish this thing did? Whatever your answer, the tools to build it are within reach.

How is Jada doing?

At a quick glance, I can tell you:

Jada is working part-time while pursuing her master’s degree in nursing. She and her coach Becca list their relationship in the Intensifying phase - they’re building something real.

Five stars across the board for Level Up support metrics. She’s crushing it at school; GPA continues to trend well.

Her dreams are ahead of her. And we’ll be watching, ready to step in if she ever needs us.

That’s what this platform was built for.

Not dashboards. Not data.

Jada.

---

This is part four of The $0 Department: What Happens When a Nonprofit Stops Hiring and Starts Building. Subscribe to follow the full series.

Subscribe now

Level Up Cincinnati empowers first-generation college students through mentorship, scholarships, and professional development. 100% of donations go directly to student programming. Learn more at https://levelupcincinnati.org

No computer science degree, no bootcamp, no side projects on GitHub. Before Level Up, I couldn’t tell you the difference between a frontend and a backend.

The app our scholars use every day? I built it. Using AI. And if your organization has a problem that an app could solve, you could build one too.

What we were working with before.

Before the app, everything ran through email chains, group texts, and a patchwork of tools that sort of worked together if you squinted.

Event details went out via email. RSVPs came back via text. The coach directory lived in a spreadsheet. Coaches and scholars didn’t have access to the full Level Up network - they were isolated.

When we started it was manageable. Messy, but manageable. At 30 scholars and 30 coaches, it wasn’t.

If you’re running a small nonprofit, you probably know this exact feeling. The tools you started with stop scaling long before the mission does.

I didn’t set out to build an app.

I wasn’t trying to solve a crisis. I saw what AI tools could do. Vibe coding was trending and I thought: what if I just tried?

The first version was built with ChatGPT and Visual Studio Code. I’d describe what I wanted, it would generate code, and I’d paste it into the editor.

It would break. I’d paste the error back into ChatGPT. It would suggest a fix. Sometimes it worked. Sometimes it didn’t.

Clunky, time-consuming, and frustrating on the bad days. But I kept going, because every small win taught me something I didn’t know before.

Then the tools got better. And so did I.

The AI tools improved dramatically, even in the span of months. What used to take hours of back-and-forth started taking minutes. Instead of building individual pages, I could work on the entire code base seamlessly.

But I was changing too. I started recognizing patterns in the code. I learned what Firebase does, how React Native works, what Expo is for. Not because I took a course - because I was building something real, and every problem forced me to learn.

The app got rebuilt multiple times. Each version was better than the last. The first version was a PWA (that’s a progressive web app) - meaning available online, not the kind you find in the app store. Basic, functional, got the job done.

Today we’re on native iOS and Android apps built with Expo, powered by Firebase, deployed and live in both app stores - it’s slick.

Getting into the App Store was its own education.

Submitting to Apple and Google Play without a mobile developer is exactly as intimidating as it sounds.

We got rejected. Multiple times.

Apple’s review process is thorough - they flag accessibility gaps, design inconsistencies, and anything that doesn’t meet their human interface guidelines. Google has its own set of requirements around permissions, content policies, and performance standards. Every rejection came with specific feedback, and every round of feedback forced us to improve.

The first submission wasn’t close. By the third or fourth round, we had better accessibility, cleaner navigation, a more polished experience across the board. The app that finally got approved was unrecognizable compared to what we originally submitted.

That process made the product better. But it also proved something I didn’t expect: you don’t need a mobile developer to meet Apple and Google’s standards. You need patience, attention to detail, and the willingness to treat every rejection as a build list.

Worth noting - the Apple Developer Program is free for nonprofits. The door is more open than most people realize.

The result? It’s ours.

If we need a new feature, I build it. No waiting on a contractor’s timeline; no $5,000 change order; no scope creep negotiations. And that’s true for any organization that goes this route - once you build it, you own it. Updates happen on your timeline, not a vendor’s.

What you could build.

Our app handles events and RSVPs, a community directory, a resource library with everything from financial literacy guides to mental wellness exercises, and push notifications. That’s what made sense for us.

Your version would look different. Maybe it’s a client intake tool. Maybe it’s a volunteer coordination app. Maybe it’s a resource hub for the families you serve. The point is, building whatever you need is now within reach.

A couple years ago, this kind of project required a developer or a firm. Today it requires willingness to learn and about $100/month in AI subscriptions.

The learning curve is real.

Building an app with AI is not “type a prompt and get an app.” It’s problem-solving. It’s debugging errors you don’t fully understand. It’s learning concepts you never planned to learn.

There were days I got frustrated - where a simple app feature quickly ate hours of my time. But every one of those hard days made the next build faster. The patterns start repeating. What felt impossible in month one felt routine by month six.

If you’re considering this, know that the hard days are part of the process, not a sign you’re doing it wrong.

The barrier is lower than you think.

The AI coding tools available today are dramatically better than what I started with a year ago. I started with ChatGPT; today I build with Claude Code and GitHub - it’s the best $100/month I spend on our business. The framework is React Native with Expo for builds and Firebase for the backend, all free.

But AI alone isn’t enough - I leaned on YouTube tutorials, developer communities, and free documentation constantly. The resources exist. The entry point is being willing to sit with something you don’t understand until you do. If you get stuck - ask AI - it can guide you.

Let’s be honest - we were never going to hire a development firm. We would have just gone without, and our scholars would still be checking email chains for event details. AI didn’t save us $50,000 - it gave us something we couldn’t afford in the first place. That’s the real opportunity here, not just for us, but for any small organization that’s been going without because the budget never had room.

What our scholars actually use most.

We track content usage across our resource library (that’s a best practice for another day) so we can see what scholars are engaging with most. Among the most-clicked items in the entire app?

Short audio mental wellness exercises.

The most popular is a two-minute walking mindfulness track - something a scholar can pull up between classes when they’re stressed or overwhelmed.

The idea that our scholars have that in their pocket, a couple clicks away - and that having access to it might positively impact their success - is the most rewarding part of any of this.

Here’s what I used to build and ship the Level Up app:

- Claude Code, ChatGPT, GitHub, GSD Repo: AI coding and debugging

- Wispr Flow: voice-to-text for faster input

- Visual Studio Code: code editor

- React / React Native: frontend framework

- Node.js / npm: runtime and package management

- Firebase: database, authentication, hosting, push notifications

- Expo / EAS Build: iOS and Android app builds

- TestFlight: Apple beta testing

- Xcode / iOS Simulator, Android Emulator: local testing

- GitHub: version control and code repository

- Vercel: web hosting

- Google Cloud Console: cloud project management

- Apple Developer / Google Play Console: app store submissions

All free or close to it. (>$300 in subscriptions)

This is the third post in The $0 Department: What Happens When a Nonprofit Stops Hiring and Starts Building. Subscribe to follow the full series.

Subscribe now

Level Up Cincinnati empowers first-generation college students through mentorship, scholarships, and professional development. 100% of donations go directly to student programming. Learn more at https://levelupcincinnati.org

$15,000 a year.

That’s what we paid a grant writing consultant. And for a small nonprofit, that’s not a line item - that’s three scholarships.

Three first-generation college students who get tuition support, professional development, and a network of people who believe in them.

So when I tell you we deleted that line item, I want you to understand what it meant.

It wasn’t a cost-cutting decision. It was a stewardship decision - and the results changed how we operate.

You’re already the expert. That’s the problem.

When we started Level Up Cincinnati, hiring a grant writing group made sense. We were new to the landscape. We needed help navigating it.

The way it worked: we’d meet monthly, sometimes bi-weekly during busy grant cycles. They’d ask me questions. I’d pull information. They’d write a draft. I’d review it. I’d flag corrections. They’d revise. I’d pull more files. More meetings. More revisions.

And here’s what I didn’t realize for a long time: I was still doing most of the thinking.

The consultant needed me to explain our mission, our outcomes, our students, our financials. Every cycle. I was the subject matter expert. They were the middleman. And the middleman wasn’t saving me time. They were adding steps.

This isn’t a knock on grant writing consultants. If you’re just getting started, outside help can make sense.

The real issue is bigger than who writes the grant. It’s about what happens to the knowledge.

The moment that changed my math.

Going into our second year with the consultant, I tried something different. I pulled up a grant application and started answering the questions myself, using AI to help me draft, refine, and proofread.

The first time I did it, the process was simple. Write a draft response, check it, let AI help with the language and framing, refine it again.

And I realized something that surprised me.

The time I spent explaining to someone else how to write it was more than the time it took to just do it myself.

That one realization is the principle behind everything that followed. If you’re already the person with the knowledge, the fastest path isn’t finding a better writer. It’s finding a better process for getting what you know onto the page.

The principle that changed everything: institutional knowledge compounds.

Here’s the idea that matters most in this post.

Think about how most organizations handle grant writing, whether it’s a consultant, a freelancer, or even an internal staff member. The knowledge about your organization lives in different places.

Your program data is in one system. Your financials are in another. The narrative about your mission lives in someone’s head, or scattered across last year’s reports, or buried in a Google Drive folder no one has opened since the last application cycle.

Every time a grant comes up, someone has to reassemble all of it. If it’s a consultant, they ask you for everything from scratch. If it’s an internal grant writer, they’re bouncing between departments. If it’s you, you’re digging through your own files trying to remember how you framed something last time.

The result is the same regardless of who’s doing it: you’re rebuilding context from zero every cycle.

When you build a system where that knowledge is centralized and compounds, the opposite happens.

I started keeping previous grants organized (via Obsidian) in project folders. Every application, every narrative, every budget justification. When a new grant came up, AI could draw on all of it. The language, the framing, the data points, the specific way we talk about our coaching model. It all carried forward.

Every grant I wrote made the next one easier.

That’s compounding. And it’s the single biggest advantage of building this kind of system. The fact that your organization gets smarter every time you go through the process.

It doesn’t matter whether you’re replacing a consultant or improving how your internal team works. If your institutional knowledge isn’t organized in a way that builds on itself, you’re doing some version of starting over every time. And that’s where the real inefficiency lives.

Research before you write. That’s the other unlock.

Our grant success rate is close to 90%. We stopped applying to everything and started researching first.

Before we write a single word, AI helps us research the funder. Their values, their giving history, the size of grants they typically award to organizations like ours, what they’ve funded recently. We know before we start writing whether it’s a genuine fit and what a reasonable ask looks like.

This is a principle any nonprofit can apply today, regardless of what tools you use: spend more time deciding where to apply and less time writing applications that don’t match.

When you only apply where you’re a genuine fit, your success rate goes up and your wasted effort goes down. That’s not an AI insight. That’s a strategy insight.

The numbers behind the shift.

Using the grant writing service, we brought in $65,000 in foundation grants.

Since doing it ourselves, we’ve brought in $465,500:

- Year one: $95,000

- Year two: $143,500

- Year three: $227,000

Year over year growth of 27%, 51%, and 58%.

Notice the growth curve. It’s not flat. It’s accelerating. That’s the compounding at work. The process gets better because the knowledge base gets deeper.

Total time per application now: 1 to 2 days. Maybe 2 to 4 hours of actual work. Down from 30 to 45 days.

What this doesn’t solve for you.

You own the narrative. AI will give you a draft. You need to read it and make sure it sounds like your organization, not a template. The voice has to be yours. Back to the idea of compounding, the more you write - the more your future drafts will hit the mark on first pass.

You own the numbers. Financial figures, student metrics, impact data. You need those ready and accurate before AI ever touches them. AI has gotten better at working with data, but verifying it is still on you. Always.

The preparation is the work. At the end of each year, we build a comprehensive program summary. All the wins, all the challenges, student metrics, financial data, key projects. That takes real effort. It’s not glamorous. But it’s what makes everything else fast. The speed you see in a 2-hour grant application is built on a foundation of organized, accurate information.

If you skip the preparation, AI will just help you write mediocre applications faster.

The tools will change. The principles won’t.

Here’s what I used this year: Claude, Obsidian, ChatGPT, Wispr Flow, NotebookLM, Google Drive, Salesforce, foundation directories for research. All free or close to it.

But I want to be direct about something. By the time you read this, some of these tools will have changed. New ones will exist. Some of the ones I’m using now might not even be the best option anymore.

That’s how fast this space moves.

The principles are what transfer:

• Centralize your institutional knowledge so it compounds.

  Keep your past work organized so it feeds your future work. Don’t let it live in someone’s head, walk out the door with a vendor, or scatter across departments where no one can find it.

• If you’re already the expert, build a process that lets you get it down directly.

  The fastest path to a great grant application isn’t finding a better writer. It’s giving the person with the knowledge a better system.

• Research before you write.

  Time spent on the front end deciding where to apply is worth more than time spent writing applications that don’t fit.

Those three ideas eliminated a $15,000 annual expense and helped us raise $465,500 in grant revenue.

They’ll work whether you have a consultant, a grant writer on staff, or you’re the ED doing it yourself at 11pm. And they’ll work whether you’re using the tools I listed above or something that doesn’t exist yet.

Three scholarships.

I keep coming back to this.

$15,000 was three scholarships.

Three first-generation college students who now get the support system that their peers with college-educated parents take for granted.

That money doesn’t go to a consultant anymore. It goes to students. And the grants we write ourselves bring in seven times more on top of it.

Every dollar we don’t spend on overhead is a dollar that goes directly to a student. That’s the math that matters.

This is the second post in The $0 Department: What Happens When a Nonprofit Stops Hiring and Starts Building. Subscribe to follow the full series.

Subscribe now

Level Up Cincinnati empowers first-generation college students through mentorship, scholarships, and professional development. 100% of donations go directly to student programming. Learn more at https://levelupcincinnati.org

That math shouldn’t add up. But it does.

Most nonprofits grow by adding people, more staff, more layers, more cost.

We made a different bet.

We invested in systems instead of headcount.

Three years ago, Level Up had 10 scholars. Today we have 30. Revenue is up 38%.

We scaled the mission without scaling the overhead.

And we did it with AI.

This is a stewardship story.

At Level Up Cincinnati, we support first-generation college students from low income households. We match them with dedicated coaches. We run events, student programming, write grants, build technology, manage donor relations, handle communications, produce financial reports, and publish an impact report our donors actually read.

With a team of one and a half.

Not because we’re grinding ourselves into dust. Because we built systems that handle the work that used to eat our time, so we can spend that time on the work that actually matters. The relationships, the students, the mission.

Here's what that looks like in practice.

I sat down recently and mapped out everywhere AI touches our operations. The list came to 40 use cases.

Not hypothetical. Not “we’re planning to.” Already running.

Things we built once that would have cost over $117,000 the traditional way: A custom mobile app, full website, Salesforce CRM setup (with custom dashboards), legal documents, brand guidelines, onboarding automations, coach training materials, and strategic plans.

Things that save us over $53,000 every year: Grant writing, financial reporting, email marketing, content across every channel, meeting notes, donor prep, coach matching, quality control on every communication that leaves the organization.

Year 1 total value: over $170,000.

That’s not money we pocketed. That’s money we didn’t spend on overhead.

Which means it went exactly where our donors want it to go: to scholarships, programming, and the relationships that change our students’ lives.

The habit that changed everything.

I didn’t wake up one day and decide to become an AI-powered nonprofit.

It started small - drafting an email, cleaning up a document, summarizing meeting notes. Nothing revolutionary.

But I made a habit of it. Every time a new task hit my desk, the first question became: Can AI help with this? Most of the time, the answer was yes.

And here’s the thing about habits: they compound.

What started as saving 20 minutes on an email turned into building an entire mobile app. What started as cleaning up a grant draft turned into replacing a $15,000 annual consulting contract.

The progression wasn’t dramatic. It was consistent. Consistency is the part most people miss when they think about AI.

"But what about the human element?"

Fair question. Here’s my honest answer: it’s a non-factor.

AI handles the backend - the systems, the drafting, the formatting, the data crunching, the process automation. The stuff that used to bury us.

We handle the relationships. The coaching conversations. The check-ins with students at 10pm, when they’re stressed about midterms. The donor meetings. The moments that actually change someone’s trajectory.

We don’t get caught in the weeds anymore - we’re well above them. We see the whole vision, and we win because of that.

When you’re not drowning in admin work, you actually have time to show up for people.

Here’s what the rest of this series looks like.

We used to pay $15,000 a year for a grant writing consultant. We don’t anymore.

The app our scholars use every day was built with AI tools. It cost us nothing to develop.

Salesforce is free for nonprofits. Configuring and maintaining it is where most organizations spend $20,000 - or give up entirely. We took control of ours.

Our annual impact report used to take weeks. Now it takes hours. It’s also the best one we’ve ever produced.

Our entire onboarding pipeline, from application to first coaching session, runs on automations we built ourselves.

Every post in this series will include real numbers, real workflows, and honest reflections on what worked and what didn’t.

Why I’m writing this.

I believe every small nonprofit in this country could be doing some version of what we’re doing. The tools are free or close to it. The learning curve is real but manageable. The payoff is measurable.

But most organizations haven’t started. Some think it’s too complicated, some think it’s not for them, some are just too busy surviving to try something new.

I get it. We were there too.

So if you’re running a nonprofit, and you’re curious where to start, go back to the habit I mentioned earlier. One question, applied consistently: Can AI help with this?

I’m opening up the entire playbook. Follow along if you want to see how it works.

This is the first post in The $0 Department: What Happens When a Nonprofit Stops Hiring and Starts Building. Subscribe to follow the full series.

Level Up Cincinnati empowers first-generation college students through mentorship, scholarships, and professional development. 100% of donations go directly to student programming. Learn more at levelupcincinnati.org.