When Sophie Steffen asked Tom Cox of Semrush Academy and Brian da Silva of Attio where AI genuinely helps and where it quietly creates new problems, they landed on the same answer from opposite ends of the content world. The bottleneck is not the writing. It is the context that has to exist before any writing – human or machine – is worth doing.
The Messy Middle
Tom is refreshingly honest that nobody has this figured out. He describes the current moment as a messy middle, and he has watched teams build entire content machines around it – pulling Semrush data through an MCP and waking up to a finished draft in the inbox, which he readily calls magic. He also understands exactly where that magic comes from: pressure. Every content team has been asked to execute more and faster because the technology is there. His pushback is not anti-AI; it is anti-uniform. The machine that works for one type of content fails badly at another.
The failure shows up most clearly in product-specific content – the onboarding courses and step-by-step resources Tom builds at Semrush Academy. You can hand AI documents and screenshots, but you still have to walk it through the product with so much precision that something snaps.
“After a while, you realize that it’s actually faster to just do it yourself.”
The deeper error he owns is more interesting than the tooling complaint. For a while he leaned so heavily on AI for drafting that he started to neglect the stage before drafting entirely – the discovery and ideation phase. That, he says plainly, is a very clear error, because it is the stage that determines whether the draft is worth writing at all.
Context Lives in People’s Heads
The reason that stage cannot be automated is almost philosophical. The context you need does not sit in a tidy document waiting to be retrieved. For course content, it means sitting with a product team for two hours and asking all the annoying, intricate questions a customer might ask themselves. For blog or social teams, it might mean listening in on sales calls to hear what actually resonates with buyers. The information is human, distributed, and often unspoken.
“Most of this really important context lives in people’s heads. It’s not in some random document or in a notepad. You have to be a bit like a journalist to dig it out.”
That journalist framing is the crux. The work AI cannot touch is the interviewing, the prodding, the patient extraction of detail from people who have not written any of it down. Slow down and do that, Tom argues, and then – only then – can you draft with AI productively.
The Social View: Same Problem, Different Owner
Brian agreed and added the part that makes it actionable. A cold start is doomed by definition.
“AI only works if it has the context necessary to write a piece that maybe resembles what you wanna get out.”
The difference on the social side is who carries the burden. Brian is often lucky enough not to sit in every product and sales meeting himself, because that context is handed to him – but only because someone else, a product marketer or a customer-education person, did the digging first and showed him how the specific feature actually works. The context job does not disappear on social; it just changes hands.
He also drew a useful line through the type of content. For a product launch, you are not explaining every step – you are focused on a single use case and making a user feel something. But the moment a piece becomes a walkthrough or a use-case showcase, you need the full step-by-step context, including the things users actually trip over when they try the feature. Either way, the conclusion both of them reached is identical and unfashionable: a human has to do the context work, and AI is only as good as the context it is handed.
For the full interview breakdown, see our complete Expert Insight with Tom Cox and Brian da Silva.
Tools Mentioned in the Interview
The following tools and platforms were referenced during this conversation.
Related Content
Content Strategy in the AI Era: Expert Insights with Tom Cox and Brian da Silva
Tom Cox of Semrush Academy and Brian da Silva of Attio go behind the scenes on the tools, workflows, and friction of content and social in the AI era -- and why the bar for quality has never been higher.
The Content and Social Metrics That Tell the Real Story
Page views, course feedback scores, high-value saves and shares, and video watch time. Tom Cox and Brian da Silva on why no single number ever tells the whole story.
The Behind-the-Scenes Tool Stack for Content and Social
Tom Cox runs on Google Docs, Claude, monday.com, and Skilljar. Brian da Silva runs on Whisperflow, Claude, and Notion. Two specialists, almost no overlap with the fourteen-tool fantasy.
