Content Strategy in the AI Era: Expert Insights with Tom Cox and Brian da Silva

Sophie Steffen: Welcome back to Let’s Talk Marketing. Today we have a virtual roundtable, joined by Tom, senior content strategist at Semrush Academy, and Brian, head of social and creator at Attio. Welcome. People usually see the finished content pieces – the courses, the social posts, the videos, the campaigns. What we don’t often see are the tools, the workflows, and the challenges behind them. That’s what we’re focusing on today: not the strategy or the theory, but the day-to-day reality of content and creator-led growth. What’s working, what isn’t, and how those roles look behind the scenes. Let’s start with a quick intro round. Tom, why don’t you start?

Tom Cox: Of course. I’ve been thinking about how to describe what I do recently, and I’ll call myself a content strategist and a writer, not necessarily in that order. It’s a role that’s evolved over time. I’ve worked across B2B SaaS and B2C companies over the last decade, helping craft their brand story through content and writing – from startups where I was part of the events function crafting messaging and positioning, to Preply, where I met Brian and where Sophie also has some connections, where I helped grow their organic search function and webinar program with experts. Then later to Gartner, where I focused on research and thought leadership content as well as lifecycle marketing. And now I’m at Semrush, where I’m part of the customer education team, building the academy and knowledge bases for the Semrush Enterprise product. Very excited to be here.

Sophie Steffen: Perfect. Brian, your intro?

Brian da Silva: Yeah, excited to be here. I’m Brian, social and creator lead at Attio, a fast-growing London-based startup. I’ve been working in social now for almost a decade, which is insane to think about. It’s changed so much since I started. I specialize in organic social specifically, but I’ve moved into managing creator programs as well over the last five years.

Sophie Steffen: Perfect. Let’s continue with you, Brian. Looking at your stack reality, what specific tools were dominating your screen yesterday?

Brian da Silva: I have a pretty simple stack, but I use it every single day. I’ll start with my favorite because it has changed the way I work personally: a tool called Whisperflow. I personally hate typing – I’m not that good at it, which is hilarious because I have a newsletter and obviously I need to type. So when I found Whisperflow, it expedited the way I work, and it lets me write in the way that I speak. And it’s made me quicker with just about everything because I’ve combo’d it with Claude. I use Claude as my secondary tool, every single day. It’s my thinking partner. It’s what I use to clean up copy, clean up my thoughts. And the last tool, that I feel like everyone uses, is Notion. I use Notion as my database. It’s like my brain inside of work, and it’s connected to Claude, which automatically updates my Notion with things like inspiration or my content calendar. Those three are in my personal tech stack that I use every single day as a social pro.

Sophie Steffen: I have not tried Whisperflow. Have you, Tom? And if not, what’s your go-to stack at Semrush Academy?

Tom Cox: Right now we’re two-thirds of the way through Q2, so it’s very much crunch time. There’s a lot of content production and writing; the strategy is out of the way. Similar to Brian, there’s nothing too strange about the tech I’m using. I have used Whisperflow and I do like it – to use voice, if you’re on a walk and you have an idea, to capture it and then play with another tool like Claude, is very effective. But really I start analog: I always have a notepad nearby, something before it becomes digital. Quite honestly, Google Docs is where I spend a lot of my time, writing, working, collaborating with freelancers. We’re using Claude a lot within Semrush – we’re big adopters of AI – and that’s where some of the drafting takes place, within a project pre-built with product messaging, positioning, and guidelines, so it works as a kind of co-writer. We use monday.com to manage projects, with a Gantt timeline and statuses. And we’re starting to work with Skilljar, building the academy content within an LMS. So those are dominating my tabs: Google Docs, Claude, monday.com, Whisperflow, and Skilljar.

Sophie Steffen: That’s quite a heavy stack. Moving into workflow friction – Tom, let’s start with yours. You mentioned Google Docs is where you write and edit. At Semrush Academy you’re producing courses and self-service resources. What’s the part of the process that cannot be sped up? And where do you see AI actually supporting your work versus slowing it down or creating new problems?

Tom Cox: Honestly, we’re still in the messy middle of figuring out how to get the best out of AI. I’ve seen teams build a whole content production machine around AI, where they no longer discover or search for topics – they pull in Semrush data via an MCP and have a ready draft in their inbox, which is like magic. I get that this comes from a place of pressure to be more efficient, because every content team has been asked to execute more and faster. But I’d push back on the idea that it works the same way across all types of content. For more product-specific content, like onboarding courses and self-service resources that teach someone a product step by step, it’s a very different challenge. You can share docs and screenshots with AI, but you still have to walk it through the product very precisely, and after a while you realize it’s actually faster to just do it yourself. For a while I thought I could use AI for so much of the drafting that I started to neglect how important the step before that is – the initial discovery and ideation phase, and that’s a very clear error. Within academy and course content, that can mean sitting with a product team for two hours, going through it step by step, asking all the annoying intricate questions customers may be asking themselves. For blog or social teams, it could mean listening into sales calls and seeing what resonates with buyers, because most of the important context lives in people’s heads, not in some random document. You have to be a bit like a journalist to dig it out. That’s not a part you can speed up. You need to slow down and interact with people, and then you can draft with AI.

Sophie Steffen: I like that it’s in people’s heads. Sometimes I assume AI has all the answers – I’ll tell it to draft me a LinkedIn post and go through so many iterations that it would have been faster and more authentic to just write it down myself. Brian, how does that compare to your world in social?

Brian da Silva: Just to add context to what Tom said, it’s really important to provide AI with as much context as possible. A cold start where you say “I want to write about X topic, write it for me” – of course the output is not going to be good, because AI only works if it has the context necessary to write a piece that resembles what you want to get out. So it’s important to do that first step, sit with the product teams and have that context before you write a script or a piece of content. On the social side, I’m lucky that I don’t always have to sit with the sales and product teams, because that context is often provided for me. But someone – the product marketer who gets all the nitty-gritty and shows you how to use the specific feature – has to do that. On social, context matters and so does what you’re publishing about. If you’re launching a product, you’re not necessarily going to explain every single step; it’s more focused on a specific use case and getting a user to feel a certain way. But if the piece is a walkthrough or a showcase of a specific use case, then you need that step-by-step context of how to actually use it and the things users might face. So number one, context is important, and someone has to do that job – whether it’s Tom, a PMM, or a customer-education person. You need that context to get going, even for social.

Sophie Steffen: The context has to be done by a human, and then you use AI to enhance it. Brian, moving into execution – you deal with creator relationships. Could you take us behind the scenes of your last major campaign? Where does the friction start: the contract, the brief, or the content delivery?

Brian da Silva: I’ll take you back to our biggest launch ever at Attio, about two months ago. It was for Ask Attio, the conversational layer we built inside of Attio – and Attio is an AI CRM where you manage your customer relationships. Ask Attio was our biggest feature launch ever, so we wanted to activate the entire social ecosystem: founders, employee advocates, our creator partners, executive leadership. Focusing on our paid creator partners – I’m a team of one at Attio, managing the creator function, the creator program, and social, with some help internally on the creator motion. I’m managing twelve creators now, and we activated all twelve for Ask Attio, which comes with challenges, because you have to review twelve different pieces of content alongside everything else for launch. I wouldn’t call it friction; the challenge is reviewing that volume of content from creators with varying backgrounds. They’re not all sales or go-to-market creators – we have founders, a go-to-market engineer, a salesperson, someone who works in VC and built her own syndicate. Their perspectives and the way they’ll use Ask Attio differ from what we’d launch with from the brand side. So the challenges were giving enough time to review content and keeping a watchful eye that it aligned with what we wanted to get across, and making sure their use case actually made sense. Worst case, a creator goes live and the use case they’re highlighting isn’t possible, and the prompt they showed doesn’t work. So it was volume, and making sure everything they went out with supported the tool, showed it in a good light, and was actually possible.

Sophie Steffen: Volume is something that can’t be sped up. You need the time. Tom, moving to something less about friction – when you produce content, including video, how do you decide which content gets a human face, an expert or internal host, versus which gets anonymized?

Tom Cox: There are a few factors, and it depends on the type of content. To touch on what Brian said about creators and use cases – that’s something we’re playing with in the academy. There are huge topics like AI search, and your company can have a belief or methodology, but sometimes these topics are so broad that you have to put your hand up and say, “We don’t have all the answers, but here are some experts we speak to who have solid approaches and use cases that may fit your situation.” That’s one way to bring a human element in. In organic search, adding experts is essentially information gain – what new content your page brings into the conversation compared to competitors. That uniqueness helps you rank. And we can’t forget E-E-A-T: experience, expertise, authority, trust – who authored or edited the content. That’s important from a search perspective but also a human one. If you saw an article on WebMD, you’d want to know it was at least vetted by someone with a medical background. Search is usually a reflection of what’s good for the user. So those are a few ways we decide whether to make content human-led or just written.

Sophie Steffen: There’s a pattern: the human aspect is becoming evident as the way to stand out – the uniqueness, the vetting, the expert part – because AI is this soulless technology. Moving into success metrics, one of my favorite topics. Tom, how do you measure if a content piece did its job? Which stats and KPIs tell you content and product are converging?

Tom Cox: For the customer education team at Semrush, we look at page views. I know that’s basic, but it shows we’re getting the word out and distributing, which is important. A lot of content teams spend time in creation limbo, constantly pumping out material without getting the word out internally and externally. So we set goals for that. But we’re also looking at the experiential side, especially for courses. One way our team does this is adding a feedback form at the end of each course – a quick check on satisfaction – which helps us see whether people find the content clear and useful. Did you learn something? Is it clear? I’d love to tie all of this neatly back to retention, but we’re not quite there with a robust enough model. It’s nice to attribute, but it’s tricky. As we build this academy, that’s the long-term goal, and hopefully Skilljar helps us do that. If we can prove people are sticking around and leaving feedback, that feels important – it’s a good sign we’re actually helping people.

Sophie Steffen: It’s mainly the attribution of content reflected in retention – translating the impact of individual pieces. That’s difficult because there are so many data points, and attribution is never first or last touch. Brian, what’s your take on success metrics?

Brian da Silva: From the social side, success comes down to two things: the purpose of social is to drive awareness for the brand, and to drive traffic to the site or specific landing pages. For awareness, I look at total engagements – but there’s a conversation of high-value versus low-value engagement. I want to drive higher-value engagement: saves, sends, comments. Anything that requires the user to take an extra step. If you’re saving a piece of content on LinkedIn – and they make it hard to save – it’s because you really found it useful. So that’s a signal that it resonated. A comment is high-value because you’re taking time to consume the content and then drop a comment. I also look at impressions and reach as an indicator. Traffic is simple – we track everything through UTMs and make sure we’re playing our role in driving traffic to the site. Another way of measuring, which Tom brought up, is having a robust data model. At my previous company we had a mixed marketing model where we could see the impact of social, because the team was more robust and we had a legacy YouTube program driving a ton of signups, so we were contributing dollars back to the business. At Attio our social program isn’t super robust yet – I’d say we’re at level one of what social could be. So we also look at the onboarding survey: very clear parameters like “How did you hear about us?” with social as an option, then we filter by channel. And we get granular with video – I’d define video success differently than a static piece. For video I look at watch time, average time watched, and drop-off, to see where users stop and where we need to refine. We’re in a world where everyone wants everything to be human-led, and video seems to be the thing people turn to. But if you make a one-minute video and people watch for an average of fifteen seconds, they’re missing the entire point. So I refine and look at those things on the back end, because if people only watch ten seconds of a two-minute video, how much value am I actually providing?

Sophie Steffen: The biggest challenge in video is the attention span – TikTok has reduced it to five, ten, thirty seconds. Brian, do you have one single success metric, or is it the mix?

Brian da Silva: It’s a mix, because it adds a more complete story. If I focus on one thing, it doesn’t tell the larger story. And social isn’t just organic – it’s the creator program, paid creators, employee advocates. All of that tells a more cohesive story than me saying “we had 25,000 impressions last month.” That’s not impressive, and it doesn’t tell the story I want. So I give as much context as I can across social as a function, so the business understands the impact social has.

Sophie Steffen: With one metric you usually don’t get the full picture. Let’s tackle the daily reality of tools again – not the stack, but the workarounds. Tom, every vendor promises a seamless experience. Is there a tool driving you crazy, and do you have a hack around it?

Tom Cox: I’m not sure it’s driving me completely crazy, but I’d love a closer tie-in between the Google Docs we work on and the project management software. It’s a small thing, but in the project management software you can’t just drop a link in like you would in a spreadsheet – you have to click “Add New File” and choose the name. Too many steps. We’re used to Notion or Google Sheets where you drop a link in and everything is intuitive. I also feel like I haven’t taken advantage of the built-in Gemini feature within Google Docs during drafting and editing – using the chatbot alongside the content. That’s probably an action point for me after this call. It doesn’t rob me of hours, but when you have twenty-five line items and need to add everything manually, that’s a lot of clicking I don’t want to do.

Sophie Steffen: Brian, any issues with tools, or friction in your day-to-day managing social content?

Brian da Silva: I’ll start with the tool. This is specific to my workflow, but it drives me insane. We use Frame.io to review video content. When you’re leaving feedback on a specific part, you have to put the timeline cursor on the specific second and then drag it along the timeline to leave feedback on a section, instead of being able to stop it and pick another part where you want it to end. Half the time it’s not precise, and if you accidentally click out, it stops and you have to do it all over again. Two, I’m not very good at Figma – I love it, but I’m not well-versed in design, so when our design team tells me to go do something, I don’t always know how. That’s more my lack of ability than the tool. And the last one is specific to Claude on the iPad: if you have a long response and try to copy the whole thing, you can’t – you have to go paragraph by paragraph. It drives me nuts, and I like to work on my iPad for things outside of work.

Tom Cox: Totally with you on Frame.io. I’ve been in that exact situation – it’s finicky, and you can’t zoom into the timeline to be precise about the seconds. And with Claude on mobile, you can’t access a project, and if you have a long list of chats you don’t know which to reference.

Brian da Silva: Tom, you’ve got to pin the chat. It’ll solve your problem on mobile, because it pins across all your devices.

Tom Cox: Ah, okay. We’re learning. Thank you, let me give that a go.

Brian da Silva: I can answer your friction question now. It’s funny, because the meme you always see on social is “it took legal forever to approve this and we missed the boat.” I’ve actually never had a problem with legal from the social side. When I worked at Microsoft, a huge company, I was too low-level to even interact with legal, and that would be the only company that might stop me from publishing. But at a startup, the level of approval is one or two people max, and it’s not hard to get things out – you’re welcome to test. So never really a problem with legal. There has been friction with PMMs, sales folks, and other parts of the business that aren’t even tied to social but have an opinion on it – because everyone uses social every day, so everyone has an opinion. But managing it for a business is different; it’s a unique skill set. The friction with PMMs is them coming to the table having done half of the work I’m supposed to do: “I’ve put together the plan for social already, I just need you to review it,” or “I’ve already written initial copy, try not to change too much.” So I navigate that relationship, especially at startups that aren’t as structured as the bigger, more established company I came from – where Joe has three jobs in marketing and social was one of them, and now I’m here and he’s still going to do this role. It’s about saying, “I’m here to own this function. How can I support you from the beginning without you doing all the work, come in as an expert, and make sure we get the most out of social?” And from the sales and enablement side, I see a ton of friction. When you work in marketing you play some role in employee advocacy – I put together social guides for bigger launches that enable the entire company, especially sales who are active on a channel like LinkedIn, to publish. The friction with enablement is that you write the copy options, give the context, the asset, the instructions, and people still publish the wrong thing, or copy that isn’t approved. A real example: someone screenshotted the brand asset from LinkedIn – with the LinkedIn UI border still visible – and published that on their personal brand as the asset. I messaged them: “What went wrong here?” There’s a folder with six different options, and this person decided to go with that. So you can give someone everything, every single thing, with clear instructions and free help, and people still get it wrong. The challenge is how I navigate that conversation, and how I get them to publish that last thing correctly when I’ve done ninety percent of the work. Those would be the two main friction points over the last five years across the startups I’ve worked at.

Sophie Steffen: That’s common in a lot of companies. One friction is also that advocates either ignore the request to share entirely, or share something completely different even when you’ve prepared everything. Tom, did you have any of those stories?

Tom Cox: I didn’t have the exact example of someone screenshotting the LinkedIn UI – that’s a unique case. But that’s part of the challenge: getting a consistent through-line from product marketing to the different channels, the blog team, social advocacy. A clear message that isn’t jumbled. And in the end, AI platforms need a consistent message to understand a brand, and that includes across LinkedIn now – it’s being pulled as a cited source. If someone is modifying how they talk about a brand, that can impact how the brand is perceived in online search and AI search. So we’re doing our best to control the message, but it can still be tampered with. And advocacy programs are sometimes hard to get off the ground, so it’s a seesaw: they’re sharing, fantastic, but they’re also sharing something that needs tweaking. No specific stories, but I feel your pain, Brian.

Sophie Steffen: Tom, since you’re at Semrush, well known for ranking and positioning – with the rise of AI and how content production has changed, what content strategies actually work to gain positioning for SEO, GEO, citations? What’s actually influencing positioning versus what just sounds good on a LinkedIn post?

Tom Cox: There are so many counter-narratives right now it makes you dizzy. Someone on my team shared a case from an event: one speaker said, “For AI search, listicles are perfect, we recommend creating listicles and appearing on listicles.” The very next speaker said, “Listicles – scrap them, get rid of them, they don’t work.” So it’s conflicting and confusing. What I love is that the Semrush customer education team has worked with in-house and industry experts to build a methodology for how to think about this – an approach we’re calling the AI operating system, which you can actually search for now. It’s not the final be-all strategy, but the crux is that it all comes down to – ironically, because everyone said SEO is dead – strong SEO foundations. You need a technically healthy site with access to high-quality content that offers something new, information gain, E-E-A-T. Those are the foundations. But it also means a new alignment across teams to all speak the same message. The role of the SEO manager, who previously just looked at keyword research and strategy, has evolved, because now they’re also asking how to get product marketing, brand, and social in the room to talk about the same message everywhere, to help with clarity. Clarity about how AI search speaks about our brand, and visibility – and that involves social, third-party sites, review sites. It needs to be consistent, because you could search for your tool in Perplexity and it will say something different than ChatGPT, something different than Claude, and none of it is what you want to be known for. So it’s not really a tactic or a hack – it’s the long game. There’s no one way; there are hundreds. The most important thing is that alignment: auditing how AI talks about your brand, how consistently you’re mentioned compared to competitors and for which topics, then showing that to your team to get buy-in and do something about it collectively across channels.

Sophie Steffen: Also how you appear on review platforms, especially if you’re a SaaS product, and where you get cited. The story you tell across your channels has to be coherent so people don’t get confused – and it’s ever-changing, so you have to be flexible.

Tom Cox: It’s SEO. It’s the long game. It’s what we always say with SEO – it’s not something that gets results next week, but you need the right strategy in place to see those results later down the line.

Sophie Steffen: Brian, anything you’d add on AI and trends, or challenges marketing teams might be facing? How is content strategy changing?

Brian da Silva: AI has lowered the barrier to publishing. Before AI, publishing a piece of content was harder to do. Now everyone and their moms can publish stuff on social – they can go onto Claude or ChatGPT, say “I want to write about this thing,” and get something that’s poorly written, sounds like AI, and publish it. So AI has lowered the barrier to being active with content. But now we live in a world where people want to consume quality content – human craft, content that’s super specific, helpful, and built to be shared. This rise in AI slop has made us crave something actually worth consuming. So now that the barrier is lower, the bar for content is super high. You can see it on LinkedIn, which is moving to an algorithm that isn’t follower-based – it’s an interest-based graph, based on what you’re consuming and what LinkedIn thinks you’ll consume. My impressions have gone down over the last two weeks because LinkedIn is figuring out what’s quality and worth serving. The battle between AI slop and quality content is full-on right now, and the reason is that the bar was lowered. So content teams are going to be very busy over the next decade. Content is a tough job right now, because the bar is so high for quality.

Sophie Steffen: It’s counterintuitive – you might think content jobs are at risk, but it’s the other way around. You have to invest more quality time and effort to stand out. Moving into the last part: lessons learned. Brian, what are the big lessons you learned the hard way, or advice for someone starting your role on Monday?

Brian da Silva: I have three pieces of advice. Number one: hold a really high quality bar for everything you do. That means keeping your audience and who you’re speaking to at the center of everything. It’s easy to chase trends – Spotify changed their logo to a disco ball and every brand on LinkedIn jumped on it, which is fine, but the pieces that matter always have your customer at the center. And take your time with every piece, make it intentional, whether it’s the copy, the format, or the asset. The biggest issue I see in social is people leading with formats – “I want to make a video” – when they don’t even know the idea or the script yet. You need to understand the concept and why you’re making it before you decide on the format. Two: publish and refine based on data, not just vibes. Don’t make decisions off personal bias or “I think this is going to be cool” without anything to back it up. It’s okay to let something spark an idea, but root your decisions in real feedback, data, things you’ve tested, or a conversation with the creative team. And the single most important one: be really easy to work with. Social is extremely cross-functional – you’re working across product, a creative team, sales, employer brand, other marketing folks. You have to be easy to work with, because if you’re not, no one will want to work with you and getting things out becomes difficult. Early in my career I wasn’t the easiest person to work with; now I can look back and see why it was annoying. If you have a bad relationship with your creative team, they’ll make your life difficult, especially if you depend on them for assets. In one of my first roles I’d be headstrong – “This is the idea, don’t change it” – and the creative team didn’t want to work with me. Now I focus on building those relationships first. It’s give and take; sometimes you don’t get what you want, but a healthy collaborative relationship makes your life a thousand times easier. So those are the three pieces of advice I’d give someone starting in social.

Sophie Steffen: Don’t be a grumpy cat – it’s easier when you’re a nice, kind person. Tom, any recommendation for someone taking your job tomorrow?

Tom Cox: Taking my job? I’ve got to watch out, especially with all this AI stuff. I’ll go with a trio as well. First, learn about this AI stuff – learn how it can cut things that took painstaking hours, but study where it goes wrong, what it repeats, what it sounds like, what is obviously AI. And most importantly, try not to use it to take away the things you enjoy doing at work. Don’t let it take away the ideation process, that early stage that requires analog input, or even the polishing stage where you think through whether something carries the message you want – because that trains your critical thinking, and if you don’t have that, you’re basically dead. If you do that, you’ll be all right in content and as a writer, and hopefully you’ll have fun, because we’re all here to have fun. Two: be aware that content is a game of quality and speed. It’s a seesaw. A lot of the time you need to push back against executing something very quickly, but you’re also playing with the time and love and insight you put into it, which is quality. So you’re always playing against this seesaw – try to get somewhere in the middle, and be ready to navigate conversations about speed and timelines while being prepared for what you’ll do about quality and helping the brand stand out. Usually that’s content that’s accurate, clear, and useful. And three: have a side project, where you’re just letting loose and creating something. It’s a training mechanism, it’s good long-term, and it makes you think about what you truly believe is important. It puts you through moments where you research deeply, which is great for content too. Someone recently told me that some of the most interesting people they know have a ton of different projects and are just enjoying themselves. So don’t worry if you have a bunch of side projects – enjoy it, try new things, test, break stuff, launch stuff.

Sophie Steffen: I love both your trios. I’d second the quality and the enjoyment, because if you care about something, you’re willing to take that extra step, and that increases the bar. But you also have to balance timelines and deadlines, and enjoy yourself – otherwise, what’s it for? Where can people find more about your work, Tom, and then Brian?

Tom Cox: Thanks for hosting this, Sophie. You can follow me on LinkedIn – I don’t post all the time, but every now and again I’ll drop some posts. And you can check out my website, huntingthemuse.net, where I publish articles on content, creativity, and the writing life. It’s a space where I write about the things I’m genuinely curious about.

Brian da Silva: Feel free to connect with me on LinkedIn, just Brian da Silva – I think I’m on the first page if you search that. And I have a newsletter on Substack called Organic Social Club, focused on social and content in general. If you want to learn more about organic social programs, it might be a good one to check out.

Sophie Steffen: Thank you so much, both of you. I learned so much in this chat. To our listeners, thank you for listening, and follow us. See you next time.