I’ve read most of the AI-and-fundraising books, white papers and blogs published in the last two years. A depressing number of them are brochures with page numbers — breathless, vendor-adjacent, promising that a chatbot will rescue a donor file nobody bothered to clean. They sell the miracle and skip the work.

Steve MacLaughlin’s just-published AI Driven Nonprofits does the opposite. It refuses the miracle and names the work. That refusal is the whole reason it matters.

A disclosure before I go further: I wrote the foreword to this book. Discount me accordingly. But understand why I agreed — I don’t lend my name to hype, and I’ve spent fifty years watching this sector fall for it, wave after wave. Direct mail. Databases. Online giving. Analytics. Each time a small cohort moved early and pulled away, a larger cohort waited and paid more to catch up, and a stubborn tail refused until refusing stopped being an option. AI is that wave, arriving faster and with more force than any of the others.

Ten Years Ago Steve Wrote the Foundation. This Is the Building.

In 2016 MacLaughlin published Data Driven Nonprofits. I called it a new classic the day it landed, and I meant it. Some organizations read it and built genuine data cultures; others nodded, admired the cover, and went back to guessing.

That first book was the foundation. This one is the building — and the building is on fire.

The fire is AI, and Steve’s argument is unfashionably plain: artificial intelligence is a tool in service of mission. Not a strategy. Not a substitute for judgment. Not a rescue party for problems rooted in bad data and worse culture. Feed a machine duplicate records, dead or erroneous giving histories, and tribal knowledge nobody ever wrote down, and it will hand you wrong answers with absolute, unearned confidence. It will be wrong at scale, at speed, and with an authority that makes the errors harder to catch.

Importance of a  Walled Garden Is the Whole Argument

If you take one idea from this book, take what I call the walled garden.

Steve’s point is that accurate, proprietary, well-tended data is the only real defense against a machine that hallucinates. An LLM will write you a flawless thank-you letter — perfect grammar, perfect warmth — and send it to a donor who buried a spouse last month, because the syntax was immaculate and the semantics were absent. The words were right. The knowing was missing.

He illustrates it with the 1954 Georgetown-IBM experiment, in which an early translation engine took “the spirit is willing, but the flesh is weak,” ran it into an AI Russian translator that returned “the vodka is strong, but the meat is rotten.” Syntax without semantics. Seventy years later we’ve made the syntax spectacular and left the semantics exactly where it was — buried in the data most organizations still refuse to clean.

The walled garden is not a metaphor for hoarding. It is the discipline of building a data environment good enough that the machine learns from truth instead of from a mess. Organizations that did that unglamorous work before AI arrived are ready. Those hoping AI will paper over a decade of neglect will be disappointed in ways MacLaughlin calls, with characteristic mercy, “expensive and entirely predictable.”

None of This Is New — and That’s the Point

Steve grounds the “new” era in 1905, when Charles Ward and Lyman Pierce sat down to a Washington dinner and engineered the last $80,000 of a $300,000 YMCA campaign — roughly $11 million in today’s money — with formulas, publicity clocks, and tracked metrics. They proved the art of the ask is hollow without the science of the system. [ See MacLaughlin’s How to Raise Money: A classic Guide for the Modern Fundraiser if you want to learn more about Ward Pierce.]

That’s the long arc, and it’s the most useful frame in the book. AI is not a rupture. It is the newest link in a chain that runs from Ward’s index cards through the database to the model. If that framing bores the tech vendors, good. It should.

Where the Book Leaves Some Stranded

I promised a review, not a valentine, so here’s the hole.

Steve deliberately names no software. That’s the right call for a durable book — startups consolidate or vanish faster than bacon at a conference breakfast bar — but it leaves the solo fundraiser with a messy spreadsheet and no IT department where she needed a foothold. The “starting from nothing” material reads more like a pep talk than a manual. He hands you the map and, by design, refuses to drive the car.

For a chief development officer of a large organization, that’s fine; strategy is the job. For a one-person shop, the gap is real, and pretending otherwise would be dishonest. Read this book knowing you’ll have to build your own technical literacy afterward, because no framework does that part for you.

The Divide Is Compounding, Not Linear

Here’s the line that should keep you up at night. In 2016 the paradox was that we were drowning in data and starving for insight. The new paradox is worse: AI democratizes the ability to generate insight while widening the gap between organizations that can feed it clean data and those that can’t. And that gap doesn’t grow in a straight line. It compounds, because each capability builds on the last.

The organizations on the wrong side won’t vanish next quarter. They’ll simply become a little less effective, a little less relevant, a little less fundable every year — until they’re explaining to a board why they waited. That explanation will not go well.

The Verdict

Read this book and take it seriously. Then close it and do the part Steve can’t do for you.  Build your walled garden by cleaning and updating your files.

Steve has provided an excellent map. But nobody, not Steve, not a consultant, not the smartest model on the market, is going to walk this territory in your place. The machine will not wait for you to be ready.

Roger