Investigative journalism is hard, expensive, time-consuming, and often dangerous. I’ve been sued, jailed in Cuba for spying, and even kidnapped over my 40-year career.

But AI is turning the tables and putting new powers in my hands. I’d go so far as to say it’s democratising investigative journalism by giving those powers to anyone.

It’s a big claim, so let me break it down in a literal field report and reveal how I’m doing the accountability reporting of an entire newsroom with a laptop and a $30-a-month subscription. AI has been crucial in how I was able to build a 19,000-strong newsletter audience in 99 countries, which led to me being invited to address the UN and advise multiple governments… in less than three years.

Let’s begin by busting a myth. This isn’t getting the AI to write 1,700 versions of the same article and blasting it across the socials. Nope, this is the opposite. Using the AI to do the grunt work, freeing me up to do the rest.

Used the right way, AI shifts the asymmetry in publishers’ favour. For decades, Big Tech has sent armies of lawyers, comms teams, and lobbyists to control the narrative. Journalists have been left to fight with notebooks and deadline pressure. The information gap was a moat, and tech knew it.

But now, it’s draining. A reporter with the right prompts can now process documents at a speed that used to require billion-dollar resources. Journalists can now do the digging they were trained for, and use the tech to turn it into hard-hitting reporting.

Now a single journalist can hold Google or Meta, Iran or Russia, history or political doublespeak to account—and still have time for lunch.

Finding the needle

I have sources like journalists always have, but many of mine are no longer human. I have alerts set up in search, and notifications on court papers and SEC filings. My tipoffs come in as a steady stream 24 hours a day. Many are nothing, but then one is a trigger. It happened the other week.

The judge in charge of breaking Google’s search monopoly ruled that a technical committee must be established to do the job. Everyone wants to know who they are.

An alert popped up on my phone from an obscure automated court reporting AI that the three had been named. The link gave me the court papers. I was off.

I dropped the committee members a line on LinkedIn, used AI to research their careers and found years-old articles that hinted at their personalities. Within an hour, I profiled them and broke the story. Then I sat down to write it. Boom. Job done.

An antitrust lawyer whose name had been linked with the role rang me from New York saying he’d scoured the court papers and couldn’t find the names. They were buried, I told him. Deep. Even with all his resources he couldn’t find it with his team. Now we’re connected too.

Financial forensics

Apples and pears. That’s journalists and accountants. Journos do words and geeks do maths. Only I love both, because you need to follow the money to find the truth.

Every quarter, the tech firms report their earnings to the US Securities and Exchange Commission. They are interminable—dull and data-laden, but full of gold. Best of all, the tech titans who love nothing more than privacy have to put these numbers out in the open to satisfy their obligations to shareholders.

It’s a goddam turkey shoot. I have uploaded years of financial filings, shareholder updates, Wall Street earnings calls into my own small language model. I’ve trained it on my historic reporting, so it knows what I am looking for. For example, is the 90.5%* of advertising Google sends to its own search and YouTube rising or falling? Spoiler: It’s always rising.

That’s why the open web is in danger of collapsing. It’s why publishers have no money to fund newsrooms. It’s also why experienced single operators like me can strike.

Apple’s numbers told me it’s reliant on China amid a tariff war. Meta’s told me 97.6% of its income relies on ads. Snap showed it relies on selling ads to the youngest teens.

These data points that justify headlines are often buried in footnotes, YOY comparisons that used to take weeks and geeks to reveal—but not now.

The fact-checking machine

AI’s a brilliant sub-editor after you’ve taught it your style.

My SLM—I call it RoboRicky—then reads the draft and alerts me if I’ve forgotten a relevant fact in a previous article. It even suggests charts and images. My newsletter contains more than three million words of investigative journalism now, but there’s more than 10 million words of source material in RoboRicky.

It checks every word against the source material to confirm it’s accurate and flags anything that it thinks is wrong.

I’ve also had fun using Google’s Gemini to punch holes in its CEO’s fibs, and had OpenAI run an analysis on whether its deals to buy content were fair. (They weren’t.)

RoboRicky + my brain + my instincts + a superhumanly unwise amount of coffee now power my one-man newsroom. No team. No budget. Me, a laptop, and my killer AI pal.

*Footnote: RoboRicky corrected an error. I’d said Google sent 89% of its ads to search and YouTube. The actual number from the Q4 SEC filing last month (and not reported but calculable by doing the complex maths) was even more, 90.5%.

The post AI is making the one-man newsroom a reality appeared first on The Media Copilot.

Kevin Schaul, a data journalist at the Washington Post, used Anthropic’s Claude Code to automate a classic data journalism challenge: consolidating messy government data spread across dozens of agencies.

The result was a story published Feb. 9 outlining how the federal government uses AI, based on inventory spreadsheets that each agency is required to publish. The problem: each agency posts files in different locations, formats and column structures.

How he did it

Schaul used Claude Code (running Opus 4.5) to search for each agency’s AI inventory page, download the files and write a Python script to consolidate them into a single dataset.

His initial prompt was ambitious: search for all agencies, find their inventory files, download them and consolidate into one CSV. That hit usage limits after 10 minutes of web searches without completing the full task.

“I should have had it fill out a spreadsheet as it went,” Schaul wrote in a blog post about the process. “Save all incremental progress to file.”

He switched to breaking the work into discrete steps: first compile a list of agencies and their inventory URLs, then download files, then consolidate. That worked better.

The consolidation step is where AI provided the biggest time savings. Claude Code iteratively wrote and refined a Python script to merge files with inconsistent column names and data formats.

“That would have been horribly tedious to write by hand,” Schaul wrote.

The trust question

Schaul emphasized he read all the generated code.

“When you’re doing data journalism, vibes are not enough,” he wrote. “I have been told ‘You’re absolutely right!’ far too many times by these tools to trust them.”

He distinguishes between having an LLM directly interpret data (which he wouldn’t trust) and having it write auditable code that can be reviewed and rerun.

“Having AI write and execute code that can be audited? I’m quite comfortable with that,” he said.

What it means

This workflow illustrates a practical middle ground for AI in newsrooms: using AI to automate tedious technical tasks while maintaining human oversight of the journalism itself.

The approach requires:

• Breaking complex tasks into discrete steps
• Saving incremental progress
• Reading and auditing all generated code
• Maintaining reproducibility through scripts

For data journalists who regularly wrangle messy datasets, this kind of AI assistance can turn days of work into hours — as long as you verify the code does what you think it does.

The post Washington Post data journalist uses Claude Code to consolidate federal AI inventory appeared first on The Media Copilot.