#39 - Nathan Thompson
Nathan Thompson has built AI content creation systems at Copy.ai and Fullcast, figuring out how to take content production from eight blog posts a month to as many as five a day. But as he built these systems, he realized that producing more content was the easy part. The bigger challenge was making sure the content was actually worth reading, which led him to focus more on the original ideas and expertise going into the system. He tells us all about how he built these AI workflows and how his approach changed as he learned what worked. We get into why he started investing more time in interviews and conversations with experts, how bringing that source material into his content helped turn more readers into customers, and why he thinks AI could fundamentally change the role of marketers by freeing them up to focus on better ideas and conversations.
Hi all,
This week, I sat down with Nathan Thompson, who led content at Copy.ai and now runs his own shop, Systems-Led Growth.
He was hired at Copy.ai in early 2023 to write 8 blog posts a month. A year later, he had automated most of the process and was publishing 5 blogs a day.
But there was something missing. It wasn’t until he fixed his inputs that the real pipeline value emerged.
Act 1: Creating the content machine
When Nathan began at Copy.ai, he was writing blogs by hand. His output was 8 per month.
Around the same time, Copy.ai's product team had released a feature called Workflows, which runs a chain of AI tasks where each task works from the last one's result. Nathan began using the Workflows to automate his blog writing process.
He built a workflow that took in a keyword and put out an edited blog post. Along the way, it pulled the top-ranking articles for that keyword, worked out what those articles were missing, turned the gaps into a brief, wrote the draft section by section from the brief, and edited the draft against Copy.ai's style guide.
It was an impressive machine.
"I got drunk with power."
Each post now took about 30 minutes end to end. Nathan ramped up blog output to 5 per day. He was spending under 3 hours per day on content, and had plenty of time to do other things.
The posts ranked well, and Copy.ai’s site was pulling about 350,000 organic visits a month.
But even with all that traffic, pipeline only nudged up. Traffic was easy to get. Buyers weren’t.
Act 2: What the machine couldn’t do
In January 2024, Copy.ai brought in a new CMO, Kyle Coleman.
Kyle had a clear mandate: I don't care about traffic if it's not generating pipeline.
Pipeline is quite a different target. You earn impressions with compelling content, effective packaging, and strategic placement. To earn pipeline, you need to do all that and build trust.
Given Kyle’s mandate, Nathan analyzed his blog output. He quickly saw the limits of his machine.
Every step of his beautifully automated blog post machine ran mostly on publicly available third-party material, i.e. the LLM + internet. The output was a piece that any competitor could make, and not something that would accrue reader trust to Copy.ai. The machine could make posts faster, but it couldn't make them stand out. It was commoditized.
So he set out to find what Copy.ai could bring to every article that nobody else could.
Act 3: First-party inputs
Nathan realized that to fix the output he needed to shift focus to his inputs. He needed first-party material.
He went back to his boss Kyle and asked that they start producing a podcast. He also asked that they start doing webinars. He set up recurring 20-minute calls with subject matter experts across the company to mine their expertise.
“Any piece of recorded information that I could have in the form of a transcript,” he said, “I felt lethal with.”
He’d gone from writer to editor to, now, interviewer. He calls it “a really fun shift.”
So what was different between a post from the first system and one from the second? “Not nearly as much as you’d think,” Nathan said. The strategies could be the same ones a competitor covered in their content. What set Copy.ai’s content apart was a person’s perspective: a block quote from the conversation, the video or podcast episode embedded in the post, the transcript in the page’s metadata, the stories that only come out when someone talks through what they did.
Nathan’s explanation is trust. A reader could learn the topic from the article either way, but once they saw Kyle on video walking through it, “you just build a different layer of trust.”
“There is a lot of content in B2B that is written to rank and to sound really good, but wasn’t actually lived by anyone. There’s no receipts for it.”
Once the podcasts, webinars, and SME interviews were in place as inputs, pipeline started to really go up. People who had found Copy.ai through the articles were booking demos at a much higher rate.
Along the way, Nathan’s success benchmark changed from the output metric of 8 blog posts per month to the input metric of 10 really interesting conversations per month.
From outputs to inputs
Today’s savviest content teams are responding to the deluge and noise of AI by shifting focus from outputs to inputs.
They take inventory of their unique, proprietary input material and invest appropriately to make it more productive.
As a thought-starter, here’s a sample list of input material for a B2B content team:
- Internal interviews with subject matter experts
- Webinars
- Podcasts (both owned and earned)
- Sales call transcripts
- Customer call transcripts or support tickets
- Aggregated and anonymized user behavior data from your product
- Commissioned reports/surveys
- Internal analyses or experiments
- News or trending content from your industry
If it feels like your content isn’t as effective as it could be, try shifting your focus to your inputs. Take stock of what you have. Note which inputs are uniquely strong relative to your competition. Shift your goals and OKRs to producing the right inputs. And watch your outputs come alive.
.jpg)
.jpg)
.jpg)
%20(1).gif)
.gif)
.gif)
%20(1).gif)
