The Case for Semi-Autonomous Nonprofits

I have helped build a few nonprofits. They are a crucial part of the economy. They address gaps in the market by doing the things that are not scalable and not profitable.

Because they do unscalable, unprofitable things, they also end up underfunded with sclerotic systems with underpaid employees with high turnover, which further slows their momentum. Since there is no singular profit motive to cut down inefficiencies, there are often imposed external accountability systems that then create more inefficiencies.

A nonprofit employee spends nearly half their time on compliance reporting. Writing grant applications. Tracking outcomes in spreadsheets. Updating donor databases. Filing reports that repeat the same information in slightly different formats for different funders. A single federal grant proposal can take over a hundred hours to write.

This kind of bureaucracy is what most of the work looks like at a typical nonprofit. This limits the time and attention for the actually impactful work.

AI can fundamentally change this dynamic.

You can think of nonprofits as two parts:

• Programs, the actual impactful activity towards the mission, e.g. running training programs, distributing supplies to people, doing research
• Fundraising & Admin, the things necessary for the nonprofit to run, e.g. grants, compliance, donor relationships, accounting, HR, IT

AI can automate the fundraising & admin. Grant applications follow patterns. An AI system that has access to your programs, your outcomes data, and your previous applications can draft a grant proposal in minutes. It can track deadlines across 30 different funders, each with their own format and reporting cycle. It can generate impact reports by pulling from the data it already has about what happened in your programs. The RAND Corporation found that compliance reporting alone consumed 11% of one nonprofit's annual budget. That is money that could go to the mission.

Donor management is similar. Follow-up emails, thank-you notes, progress updates, annual reports. These are repetitive, pattern-based tasks that AI handles well. The relationship with the donor matters. The formatting of the quarterly update does not.

Then, there is the interesting work that can be enabled on the program side.

Consider beneficiary interviewing. Nonprofits need to understand the people they serve. They can interview applicants before programs, check in with participants during, follow up with graduates after, and ideally talk to people years later to understand long-term impact. Most nonprofits do very little of this because it takes enormous time. An AI system can conduct structured interviews at scale, in multiple languages, at any hour. It can talk to a hundred program graduates in a week and synthesize the patterns across all of them. Not replacing the human relationship, but collecting the qualitative information that humans never had time to collect in the first place.

The same applies to employees. Program staff have qualitative impressions about what is working and what is not. They notice things. But there is no system to capture those observations systematically. An AI that listens to team debriefs, reads internal communications, and asks the right follow-up questions can turn scattered impressions into structured intelligence.

And then there is the data itself. Nobody needs to update a CRM. Nobody needs to log activities into a tracking system. Nobody needs to fill out a form after every interaction. The AI can be present in the conversations, the emails, the meetings. It can capture what happened and structures it automatically. The data can be invisibly taken care of.

Right now, nonprofits that use AI tend to use it as a collection of separate tools. ChatGPT to draft an email. A transcription service for meetings. Maybe an analytics dashboard for outcomes data. These tools help at the margins.

The real transformation happens when the system is integrated. When the AI that writes your grant report knows what happened in your programs, because it was there for the conversations. When the impact data in your funder report comes from the same source as the beneficiary interviews. When the donor update is generated from the same underlying information as the internal team debrief.

Often, trying to find something for external stakeholders to do feels like a waste of time in between existing work. Board meetings require lots of preparation. An underlying AI system can also give the board, and all other stakeholders, way more transparency, which in return can get more people involved in the problemsolving process.

This is what "semi-autonomous" means. A nonprofit where every human's work counts for three to five times more, because the operational overhead is handled by a system that understands the full context. The person talking to beneficiaries does not separately log, report, and track. The system does it. The person does the work that matters.

There is a further benefit that is easy to miss. When all of a nonprofit's information flows through an integrated AI system, it becomes possible to follow the latest research relevant to the mission, analyze your data, and make evidence-based decisions.

I remember staying late one day to work on an analytics project. I had just found out a tool to geographically map out data. I was curious how it would look like if I mapped our program applicants and participants by zipcode. The map was drastic. I sent the screenshot of the map and, the next day, our outreach strategy was revised.

Most nonprofits make decisions based on intuition, because they do not have time to such analytics. AI changes that. It can process information at a scale that no team of humans can, and surface the patterns that matter.

With rich data capture, full context, and agentic processes, you can virtually automate 50-70% of the current work, in addition to increased effectiveness due to automated qualitative and quantitative analytics.

What nonprofits end up doing, once the operational work is automated, is more of the thing they exist to do: in-person work. Talking to people. Being present for the communities they serve. Building relationships with donors. Building partnerships with other organizations. The automation makes the human work possible.

That's what people want to do at a nonprofit. They want to help others. They want to actively work towards the mission. They don't want to have stressful grant reporting days, they don't want to fill out a CRM for hours, or sit in process-oriented meetings.

When people talk about AI automation, they almost always talk about businesses. This is understandable. Businesses have money to spend on AI tools, and the ROI is straightforward. But nonprofits are actually easier to automate.

For-profit companies are extraordinarily complex. They have hundreds of internal systems, layers of departments, regulatory requirements that vary by industry and geography, supply chains, vendor relationships, legacy software that nobody understands anymore. Automating a business means integrating with all of that. It is a massive systems problem.

Nonprofits are often not as large. The work doesn't change as fast. The workflows are more repetitive, the systems are fewer, and the core activity is more uniform. A job training nonprofit and a food bank and a youth mentoring program all face similar operational challenges: apply for grants, manage beneficiaries, track outcomes, report to funders. The playbook is more transferable than in business.

An off-the-shelf nonprofit system can handle 50% of the infrastructure for majority of nonprofits.

When you automate a business, the value first accrues to the owners. Efficiency gains become profit margins. Workers may benefit through higher wages and the output can become cheaper over time, but the primary beneficiary of automation in a for-profit context is capital. This is one reason people are anxious about AI.

When you automate a nonprofit, the value accrues to the people the nonprofit serves.

There are no shareholders. There is no profit margin to capture. Every dollar saved on overhead is a dollar that goes to programs. Every hour freed from grant reporting is an hour a caseworker spends with a family. Every bit of qualitative data collected at scale is an insight that improves how services are delivered.

This is the rare case where automation is unambiguously good. The incentive structure is aligned. The more efficient the nonprofit becomes, the more impact it creates. And the more impact donors can see, the more likely they are to give again. It is a virtuous cycle.

AI is going to increase wealth inequality. The IMF projects that while AI may reduce wage gaps in some areas, it will substantially increase wealth concentration, because the people who own AI systems capture disproportionate returns. In advanced economies, up to 60% of jobs are exposed to AI disruption.

That means more wealth will need to be redistributed. Through taxes, through philanthropy, through mechanisms we have not invented yet. The efficiency of that transfer matters enormously.

If donors give money and cannot see results, the money will stop. And if the money stops, the alternative is darker. People with resources will focus on protecting themselves rather than helping others. The cost of helping people should get cheaper alongside everything else that AI makes cheaper. We see how some things become a commodity (e.g. food, smartphones, TVs), the things that don't get as efficient become even more expensive comparatively (e.g. housing, education, healthcare). There is a moral imperative to make nonprofits efficient.

Governments should incentivize nonprofit automation the way they incentivize business technology adoption. The return on investment is not profit. It is a society that can handle the displacement AI creates. Nonprofits are the mechanism through which redistributed wealth reaches people who need it. Making that mechanism efficient is not optional.

As AI is automating everything and creating social anxiety, we need to shift our focus on automating organizations that are trying to do good with limited funding.

Building a semi-autonomous nonprofit infrastructure is one of the most impactful things one can work on right now.

X/Twitter · LinkedIn · Email · Home