• LinkedIn
• Twitter
• Instagram
• YouTube
• RSS Feed

Apple •Spotify• Pocket Casts •Youtube •Overcast •RSS

What’s up everyone, today we have the pleasure of sitting down with Olga Andrienko, Former VP of Marketing Ops at Semrush.

Summary: Olga thought she was ahead of the AI curve, but a weekend course on autonomous systems showed her she was thinking too small. She pitched a shared internal AI stack at Semrush, built systems off APIs, skipped procurement by using already-approved tools, and tracked hours saved instead of promising vague ROI. She started with the work she already knew, made it faster, and used that time to build better systems. Now she’s looking ahead, watching work blur into participation, prepping for a wild future, and making sure ops teams aren’t caught off guard while the rest of the company is still testing prompts.

In this Episode…

• How AI Agents Reshape Marketing Ops Roles
• How To Beat AI Imposter Syndrome And Start Using Custom GPTs
• AI Use Cases in Marketing
• AI-First Thinking While Tackling Compliance and Orchestration 
• AI’s Endgame: Play-to-Earn and Mandatory Human Quotas
• What Happens When You Optimize Your Body Like a Martech Stack

About Olga

Olga Andrienko spent nearly 12 years at Semrush, where she helped build one of the strongest B2B marketing brands in tech. She started by leading social media, then expanded into global marketing, eventually becoming VP of Brand and later VP of Marketing Operations. She helped guide the company through its IPO, launched brand campaigns that drove massive reach, and scaled AI systems that saved her teams hundreds of hours. 

Most recently, she built out a marketing and AI ops function from scratch, automating reporting, content feedback, and influencer analytics across the org. Recently, Olga announced she was leaving Semrush to go out on her own. She’s now building a marketing SaaS product while advising companies on how to use AI agents to rethink marketing operations from the inside out.

How AI Agents Reshape Marketing Ops Roles

Olga had already logged countless hours with Claude and ChatGPT. She was building chatbots, fine-tuning prompts, and staying sharp on every update. Then she joined a weekend course on agent-based AI. At first, it felt like overkill. By the end of day two, she had completely changed direction. That course forced her to realize she had been spending time in the shallow end. Agent AI wasn’t just a smarter assistant. It was a structural overhaul. It changed what could be automated and who was needed to do it.

Agent AI builds systems instead of just responding to inputs. Olga described a clean divide between tools that help you finish tasks faster and agents that actually run the tasks for you. 

How agent AI differs from task-level tools:

• Traditional tools require manual input for each use
• Agent systems operate autonomously and initiate actions
• Tools accelerate individual work
• Agents orchestrate end-to-end processes
• Tools help you move faster
• Agents help you step away entirely

She saw use cases stacking up that didn’t fit inside marketing’s current playbook. Systems could now operate without manual checkpoints. Processes that once relied on operators could be built into fully autonomous loops.

“I went into panic mode. Even with our tech stack at Semrush, I realized we were behind. Every company is behind.”

The realization came with a cost model. Internal adoption of Claude and ChatGPT was rising fast. Olga noticed growing subscription bills across teams, with everyone spinning up individual accounts. She ran the numbers and saw the future expense curve. Giving each person their own sandbox didn’t scale. What made sense was building shared tools through APIs, designed to solve repeatable tasks. That way you can maintain quality, cut costs, and still give everyone access to powerful AI systems.

Timing mattered. Olga was coming off a quarter where she had high visibility, internal trust, and a direct line to leadership. Instead of waiting for AI priorities to come down from the top, she used that leverage to move. She pitched a new team and made the case for shifting from brand to ops. She had technical interest, political capital, and an urgent belief that velocity mattered more than perfection.

Key takeaway: Marketing ops leaders are uniquely positioned to build agent-level systems that scale across teams. Instead of waiting for strategy teams to greenlight AI plans, use cost data to make the case for shared infrastructure. Build with APIs, not individual tool access. Push for automation at the system level, not just task-level assistance. If you understand the workflows, know the tools, and already have trust inside the org, you are the one who should be building what comes next.

Back to the top ⬆️

How To Beat AI Imposter Syndrome And Start Using Custom GPTs

AI imposter syndrome shows up fast. It tells you the developers will handle it, the data team will figure it out, and you should stick to writing copy or launching campaigns. Olga ignored that voice. She opened up ChatGPT, looked at the most repetitive task on her plate, and started testing. No credentials. No roadmap. Just frustration, curiosity, and a weekend.

“Anybody who says they have figured AI out or that they’re on top of this, they’re lying to you.”

She did not wait for a manager to assign her an AI project. She looked for work she already understood. Rewriting vague marketing text. Fixing formatting issues. Translating copy into other languages without sounding robotic. These were not moonshot experiments. They were annoyances. She built a custom GPT for each one.

That work gave her traction. It also gave her time back. She found herself reclaiming an hour a day just by handing off the small, repeatable parts of her job. That time opened up new space to build more. The learning came naturally because it was grounded in daily tasks she already owned.

“If we look at this like a Maslow pyramid, the repetitive tasks are the base layer. That’s where you start.”

Confidence grows when the work starts to feel useful. That shift does not come from reading whitepapers or watching LinkedIn demos. It comes from applying the tool to one thing you do every week and watching it cut your time in half. That is how you build fluency. Not all at once. One custom GPT at a time.

Key takeaway: Choose a task you already know well and automate it with a custom GPT. Keep the instructions specific and tied to your current workflow. Run it repeatedly until it saves you real time. Then build another. Confidence in AI tools comes from using them to solve work you already understand, not from waiting until you feel qualified.

Back to the top ⬆️

AI Use Cases in Marketing

AI Agents Creating Drafts from Context That Humans Perfect

AI content agents are getting better, but they are not off the leash. Olga built two systems to test how far automation can go without turning content into generic filler. One starts with human writers. The other starts with a structured form. Both rely on real performance data, brand knowledge, and experienced editors.

The first system runs inside Google Docs. Writers draft copy. The AI overlay scores it using past campaign performance, conversion data, and hand-labeled examples of strong and weak copy. It flags weak headlines, vague CTAs, bloated structure. Then it explains why. Olga’s team noticed that when the starting draft is weak, AI only smooths the surface. It cleans up the grammar but keeps the same empty thinking.

“If you give it shit, it’ll give you a slightly better version of shit,” Olga said.

They stopped asking the model to rewrite content and started using it to train people instead. It became a feedback engine that teaches writers how to think more critically about quality. The system doesn’t remove effort. It shifts it toward judgment and learning.

The second system handles content generation. It begins with an n8n form and pulls from a structured internal database that includes:

• Knowledge base articles
• Competitor analysis
• Customer call transcripts
• USP frameworks
• Jobs-to-be-done docs
• Past blog content

From there, the AI drafts content using brand voice guidelines and performance references. Once complete, the system pushes the draft into a private Slack channel where experienced editors review it. These are heads of growth and marketing, not junior copywriters. They decide what stays, what goes, and when it’s ready. Olga is training the model on their edits to gradually improve generation quality.

Olga expects near-autonomous agents within a year. She also expects a growing divide. Seniors who understand how to get value from AI will speed ahead. Juniors who never get a chance to train will fall further behind. Nobody wants to invest in their development, and companies are leaning hard on tools instead of mentorship. She predicts that millions of people will build new industries around work that the AI-first world leaves behind.

Key takeaway: AI agents can generate useful content if you feed them detailed internal context and pair them with senior editors who know what good looks like. Start with real inputs: past performance data, brand-specific voice rules, customer insights, and competitor comparisons. Let AI draft. Let humans edit. That way you can scale content without scaling mediocrity, and build internal systems that raise the bar for everyone involved.

Back to the top ⬆️

How to Use a Risk and Reward Grid to Prioritize AI Projects

Olga built her AI risk and reward matrix to solve a simple problem. Everyone had ideas, but not everyone understood the cost of getting them through procurement, legal, and security. The matrix helped her choose where to start by making tradeoffs visible. She ranked AI use cases on two axes: how risky they were and how much time or value they could realistically return.

She started with the most boring, repetitive work. Her influencer team had been spending hours each month manually tracking brand and competitor mentions from 40 influencers across 7 competitors. That job involved copy-pasting from spreadsheets and doing basic math no one wanted to touch. Olga hired an AI workflow architect, and in the first month, they automated that task. The result was two hours saved per team member per month. No pitch decks. Just fewer clicks.

“This is just like copy-pasting, calculating the number of mentions. Not enjoyable,” she said.

Other wins came fast. In analytics, her team couldn’t send any sensitive data to OpenAI because Semrush is a public company. So Olga used Vertex AI in the Google Cloud suite to build an internal chatbot that pulled from spreadsheets. The tool had no context memory and a clunky UI, but it let her Head of Analytics query data directly. That saved five hours per week. Olga’s own reporting work was another target. She spent five hours building a semi-automated workflow that now saves 30 hours across the team each month. As it stabilizes, her own involvement drops to three hours per month. She sees the return clearly and so does everyone else.

The grid changed how she prioritizes. Startups can be loud, messy, and high-volume with zero brand equity to damage. Scaleups can take smart risks and have enough resources to fix what breaks. But public companies move slowly. Olga color-coded her matrix by department based on how likely each idea was to get blocked. Green for go. Red for prepare your arguments. Hyper-personalized onboarding is her dream project, and she wants to build it in time. For now, she is focusing on earning space with small wins that are hard to argue against.

Key takeaway: Use a risk and reward grid to make internal tradeoffs visible. Start with repetitive work your team already avoids. Automate what is annoying, not what is strategic. Measure time saved in hours, not in vague outcomes. Use internal tooling first if external vendors trigger compliance reviews. Prioritize based on where you can ship quickly, prove value early, and stay out of legal trouble. That way you can build a reputation for delivering AI projects that work in practice, not just in pitch decks.

Back to the top ⬆️

AI-First Thinking While Tackling Compliance and Orchestration 

How To Use Google Workspace To Skip AI Vendor Approvals

Procurement delays do not usually happen because legal says no. They happen because legal is juggling 27 other things and your shiny new AI tool is number 26. That delay kills experimentation. Olga does not waste time waiting. Instead, she shifts all her early AI adoption toward platforms that are already approved. That move alone can save weeks of negotiation and thousands in budget.

At her last role, the company ran on G Suite. That meant any new features from Google (like Gemini or Vertex AI) were already cleared. No procurement required. No legal review. No paperwork. She saw that Gemini’s deep research tool had finally improved. So she pulled in her social media director and reran a research project that would have cost $5,000 with an agency.

“We just did the whole thing in Gemini. No procurement, no wait. That’s $5,000 saved, immediately.”

Gemini is not perfect. The interface is messy. The onboarding is buried. The documentation is lacking. But it is available. Teams can use it without submitting yet another vendor request or explaining data risks to five departments. The velocity is what matters. Google is also quietly winning back credibility with moves like the Wiz acquisition, which helps their AI offerings clear security reviews faster. That builds confidence and keeps projects moving.

For use cases that do not touch sensitive data, Olga moves even faster by hiring freelancers and agencies. She focuses first on public-facing workflows like social media, where third-party data is already accessible. Her playbook is specific:

• Start with use cases that don’t trigger legal review
• Let external partners use their own AI tools
• Write a contract that gives you IP ownership
• Include a buyout clause for the tools or accounts
• That way you can start building tomorrow, not next quarter.

Key takeaway: Use your existing infrastructure to sidestep AI vendor approvals. If your company runs on G Suite, prioritize AI tools like Gemini and Vertex AI that are already cleared by legal and security. Target public-facing workflows first, where compliance risk is low. For everything else, move faster by working with external partners and using contracts to protect your IP. This structure gives your team permission to experiment without waiting for approval.

Back to the top ⬆️

How To Decide Which AI Agent to Use

Every vendor now claims their AI agent is the one to trust. HubSpot says it can personalize based on CRM behavior. Webflow claims to tailor content directly on the page. Zapier stacks workflows across tools and calls it intelligence. Your CDP quietly insists it has the cleanest data and can trigger the most relevant actions. Each one insists it can orchestrate customer engagement, but they all pull from different contexts and operate in silos.

Olga doesn’t waste time entertaining each promise. She points out that most of these AI agents are just surface-level wrappers on features that already existed. If you’re working with a major platform like HubSpot or Salesforce, she suggests sticking with it. Switching to a niche vendor for one AI capability means introducing procurement overhead, security reviews, and more brittle integrations. Most large vendors will eventually catch up and absorb whatever the niche players are doing. Sometimes slower is safer.

“If you’re already using HubSpot, and they have agentic capabilities, I wouldn’t go test ten different smaller companies.”

Instead of evaluating each AI feature individually, Olga recommends thinking about ownership and control. If you have a technical team, build your own orchestration layer. Set up a custom connector that sits between your tools, manages the logic, and routes the right data to the right place. Self-host it for security. That way you can control how agents interact, without relying on brittle integrations or vague black box logic from vendors that barely integrate with each other.

Most teams won’t have that luxury. If you lack internal engineering support, Olga says to evaluate based on practical constraints:

• Which vendor already has the most access to your data?
• Who already passed your legal and IT reviews?
• Which platform can handle 80 percent of your use cases?
• Does anyone on your team have the bandwidth to manage orchestration?

When no clear answer emerges, pick the vendor with the clearest product roadmap and the lowest switching cost. Avoid trying to do everything with every tool. Choose one as the anchor, work within its limits, and monitor how it evolves.

Key takeaway: Focus your AI agent orchestration around the tools that already own your key data and pass your security thresholds. If your team has engineering support, build a custom orchestration layer with clear rules and full control. If not, pick one core vendor, understand what it can do today, and ask direct questions about where its AI capabilities are headed. Prioritize maintainability and clarity over novelty, and make orchestration a team responsibility, not just a shiny feature checkbox.

Back to the top ⬆️

How To Build an AI-First Reflex in Marketing Ops

Every task starts the same way. Olga asks whether AI can do it faster. She does not log the idea for later. She does not build a sandbox. She solves it in real time. If the answer is unclear, she opens the tool and figures it out while the task is still fresh.

“There’s no backlog. I just have this task. How can I make it quicker?”

One day, she turned a dog walk into an emotional processing session. Claude asked her questions about relationships, values, and emotions. She answered out loud in audio mode, then asked the model to summarize it. She pasted the result and sent it directly to her partner. Another time, she needed a ceiling lamp. She took a photo, asked for five options, had the list translated into Spanish, then used GPT to search Spanish websites. She placed the order the same afternoon.

She avoids the time trap by collapsing the task instead of carving out space around it. That shift happens when you use AI before you even decide how big the task is. You shorten the loop between idea and execution by making automation your first thought instead of your follow-up plan.

That shift shows up in:

• Weekly campaign planning
• Grocery and household sourcing
• Health data collection and interpretation
• Relationship communication
• Language translation
• Internal documentation
• Research and tooling

Using AI early gives you leverage later. The benefit compounds because the habit builds while the work gets done. You do not need to schedule blocks to test new tools when the tools already sit inside your daily motion. Tasks feel lighter. Backlogs shrink. Feedback loops tighten. Work gets faster without getting rushed.

Key takeaway: Start every task by asking how AI can reduce effort. Use it for first drafts, decision support, comparisons, translations, or summarization. Build the habit where automation becomes the default, not the bonus. That way you can eliminate drag from your daily work instead of adding more planning to your calendar. Time saved will come from instincts, not intentions.

Back to the top ⬆️

AI’s Endgame: Play-to-Earn and Mandatory Human Quotas

AI agents are starting to erase entire teams, not just automate workflows. Olga sees a new kind of economy taking shape, one built less on jobs and more on digital participation. She believes new industries will rise from this shift, especially around entertainment. The idea of “play-to-earn” already exists in Web3 circles, but she sees it expanding into the mainstream. If AI takes care of core labor, people will spend more time doing what they already enjoy. Gaming companies will benefit from higher engagement, more ad inventory, and a growing pool of players who treat games as part-time income.

“Everyone loves gaming. If people don’t need a regular job, they’ll just play more,” Olga said. “And companies will love that.”

New Income Models for the Post-Work Era

She connects this future to the world of WALL-E, where humans live in chairs, do very little, and still generate income. That scenario sounds exaggerated, but the building blocks are already here. People are spending more time on platforms that reward attention and participation. Olga sees the rise of low-effort earning models (watch-to-earn, learn-to-earn, click-to-earn) as a natural outcome. These models reward people for being online, completing surveys, or submitting information, and they sit at the intersection of boredom and necessity. If AI is handling everything else, people will find ways to generate micro-income in the time they once spent commuting.

Model	What It Looks Like	Why It Gains Traction
Play-to-earn	Gaming as a part-time income source	High engagement, scalable ads
Watch-to-earn	Ads and content that pay per view	Monetizes attention spans
Learn-to-earn	Paid certifications or skill tasks	Upskills without full-time jobs
Click-to-earn	Surveys, microtasks, content tagging	Low barrier, constant flow
Human quotas	Mandated human roles inside companies	Slows automation, protects jobs

Slowing the Replacement Curve

Governments will not just watch this unfold. Olga believes regulatory frameworks will enforce human participation inside organizations. One example she offers is a quota system where companies with a certain number of users are required to employ a minimum number of humans in customer-facing roles. A company with 10,000 customers might be forced to keep five human support reps on payroll, even if an AI system could handle the volume alone. These quotas would function as brakes, designed not to stop innovation, but to slow down the rate of replacement. That way companies have time to redesign around more inclusive structures.

The branding around AI already feels fractured. Olga called out one company that ran a billboard in Times Square saying “Stop Hiring Humans.” That same founder later celebrated hiring an “amazing team” on LinkedIn. These contradictions create friction. Messaging tone matters. Olga compared Shopify’s internal AI memo to Duolingo’s. Shopify’s framed AI as a partner to help employees thrive. Duolingo’s memo focused on replacing contractors. Both said the same thing. One landed. The other sparked pushback. Sequence and values made the difference.

A Bigger Mandate for Ops Leaders

Olga wants ops leaders to widen their focus. Tooling is important, but it is only part of the work. Most companies are still learning how to automate basic processes. Some startups building AI tools do not even use them internally. That gap creates risk. Ops teams need to take the lead in educating their orgs, supporting cross-team adoption, and building confidence across departments. That way you can move beyond implementation and help leadership face harder questions about governance, ethics, and workforce design.

“Even the companies building these tools aren’t AI-savvy internally. They’re focused on customers, but their teams are still in infancy,” Olga said.

Operations has always been about making things work. Now it also has to make sense of who works, when, and why.

Key takeaway: AI agents are already changing how companies structure work. As teams shrink and new earning models rise, leaders need to move fast without losing the plot. Focus on more than tooling. Build internal education programs, prep your org for regulatory constraints like human quotas, and get involved in shaping the values behind your AI roadmap. That way you can keep your systems efficient and your teams human.

Back to the top ⬆️

What Happens When You Optimize Your Body Like a Martech Stack

Olga organizes her life around a few things that actually matter. She chose relationships, work, personal growth, and cooking. These are not wishy-washy preferences. They are fixed priorities with consequences. She doesn’t try to squeeze in everything else. She made peace with letting the rest go.

She builds her days around physical stability. Energy, focus, emotional regulation, creative thinking, none of it works if your body is dragging. Since the pandemic, she has followed the same morning routine every day. No snooze button. No decision paralysis. She wakes up at the same time, uses three sleep trackers, skips caffeine, drinks loads of water, and gets sunlight in her eyes before the day even starts. She walks the dog in a cap, not sunglasses, to make sure her brain gets the light cue. These are not quirky habits. They are practical protocols.

“If my body works, everything else becomes easier,” she said. It wasn’t a flex. It was a statement of fact.

She takes cold showers. She cooks at home. She keeps her diet tight. These routines create frictionless mornings and reduce cognitive load. They eliminate unnecessary decisions so her actual work gets more of her attention. The goal isn’t discipline for its own sake. The goal is clarity. If you know what matters, it becomes easier to stay present, say no, and stay sharp.

Every trade-off reinforces the system. When you define your priorities with precision, you get the freedom to stop performing for everything else. You get to focus without apology. You get to protect your time and your energy like they are resources worth defending because they are.

Key takeaway: Choose four areas that truly matter to you. Build physical routines that create consistency and reduce decision fatigue. That way you can protect your energy and show up with clarity, without wasting effort on things that sit outside your priorities. Clear priorities and repeatable health habits create the conditions for sustained performance.

Back to the top ⬆️

Episode Recap

Olga had been building with AI tools for months. Not in a dramatic way. Just in the quiet, persistent rhythm of someone tired of redoing the same formatting task every week. She opened ChatGPT, trained a few custom workflows, and started reclaiming hours she didn’t know she’d been wasting. 

Then she signed up for a weekend course. It was supposed to be a deeper dive into prompt engineering, but it ended up changing how she thought about the entire structure of her work. The course introduced autonomous agents; systems that don’t wait for a prompt, they run the process. That was the moment she realized she’d only been nibbling at the edges of what was possible.

At Semrush, AI adoption was scattered. Every team had its own account. Budgets were creeping. Results were inconsistent. Olga pulled together the cost data and sketched out a better plan. Shared tools built on APIs. One system for copy refinement. Another for content generation. A feedback loop tied to brand voice and performance data. It worked because she started from work she already owned. She wasn’t trying to transform the company. She was just trying to make things faster, cleaner, less painful.

She avoided the usual procurement hurdles by using tools the company already trusted. Gemini. Vertex. Nothing new to clear. For public-facing projects, she hired freelancers and made sure contracts protected her team’s ownership. Legal never had to get involved. She ran a risk-reward grid across departments to figure out where AI would save time without creating internal blowback. The first wins were small. Content clean-up. Tracking influencer mentions. Rewriting sloppy briefs. But those hours added up, and people noticed.

Now she’s watching the edges of this world blur. The work doesn’t stop, but it’s shifting. People will get paid to play, scroll, contribute. Not traditional jobs, but not quite leisure either. Olga sees governments stepping in soon. Not to block progress, but to slow the rate of replacement. She expects human quotas. Not as sentiment, but as policy. And she wants marketing ops leaders to get ready for that shift now, while most teams are still stuck in experimentation mode.

Olga’s story is practical. She didn’t wait for an AI roadmap. She built systems that worked with what she had, right when she needed them. And she’s still building. Not because she’s trying to future-proof her team. But because it’s already the present, and most people haven’t noticed.

Listen to the full episode ⬇️ or Back to the top ⬆️

Follow Olga 👇

• LinkedIn

✌️

—
Intro music by Wowa via Unminus
Cover art created with Midjourney (check out how)

Apple •Spotify• Pocket Casts •Youtube •Overcast •RSS

Related tags

See all episodes

Future-proofing the humans behind the tech

Apple •Pocket Casts•Google •Overcast •Spotify •Breaker •Castro •RSS

• LinkedIn
• Twitter
• Instagram
• YouTube
• RSS Feed