How AI Is Reshaping Account-Based Marketing, Without Losing the Human Element

Snowflake’s Hillary Carpio on ABM in the AI Era: How to Scale Impact Without Losing the Human Kathleen Booth (VP of Marketing, Sequel) sat down with Hillary Carpio (VP of Growth Marketing, Snowflake) to unpack what it really looks like to run ABM at enterprise scale, supporting 400+ sales reps and 5,000 target accounts, while rethinking how marketing teams operate in a world shaped by AI. 1) Innovation has expanded, and so has who gets to do it “Now with AI, the art of the possible is not only larger in terms of what we can create, but it's also larger in terms of who can create it.” Hillary’s framing is important because it shifts the AI conversation away from novelty and toward capability distribution. Her point isn’t just that AI unlocks new outputs, it changes the creator map inside an organization. The subtext: innovation no longer needs to be bottlenecked in specialized teams (ops, data science, dev). That democratization becomes a strategic advantage if leaders also set guardrails and create a path from experimentation to operational leverage, an idea that reappears later when she describes “random acts of AI” versus enterprise transformation. 2) Stop “doing more with less”, reinvest the capacity you free up “If you essentially have… five people on a team and those people can now do… 10 people's worth of work… can you now create an AI native function that never was needed and never existed before?” This is one of the strongest executive-level reframes in the session. Hillary doesn’t accept the default narrative that automation equals cost-cutting. Instead, she treats AI as a way to create organizational slack, then redeploy it into new, higher-leverage work marketing historically couldn’t staff. It’s also a direct bridge into her later points about org design: if your operating model doesn’t change, AI becomes a productivity hack. If your operating model does change, AI becomes a catalyst for new functions (e.g., orchestration factories, signal aggregation, personalization workflows) that make marketing more adaptive and scalable. 3) AI content isn’t the problem, skipping the thinking is “We're missing the critical thinking layer.” Hillary pinpoints why so much AI-generated messaging feels polished but empty: teams are outsourcing synthesis, meaning, and point of view. In the conversation, Kathleen connects this to the lived buyer experience (overflowing inboxes, “AI slop cannon,” indistinguishable conference messaging). Hillary’s advice implies a practical workflow standard: AI can accelerate drafting and ideation, but the human must still own the connective tissue, what it means, why it matters, what the customer should do, and what hard questions the message should withstand. 4) “Random acts of AI” are useful, until they fragment the organization “When you are building something, are you solving for now or are you solving for the future?” This is the tension at the heart of AI adoption inside big marketing teams: empower individuals to experiment, but avoid a proliferation of disconnected tools, duplicated effort, and messy data. Hillary’s point lands because it respects both needs. Early experimentation creates confidence and fluency. But transformation requires alignment around shared infrastructure, unified data, and scalable workflows. This theme ties directly to Snowflake’s ABM tooling approach later in the session: AI isn’t just embedded as features, it’s being shaped into coordinated “workspaces” that help teams act from a shared control plane. 5) Org design: separate strategy from execution to prepare for agentic marketing “We really wanted to create a center of excellence… so that they can be the innovators of how to then get it off their plates and become audience strategists.” Hillary describes a deliberate restructuring: rather than endlessly adding headcount, Snowflake split campaign responsibilities to create an execution-focused COE that can standardize and automate delivery, freeing others to move toward audience strategy and performance insights. The significance is less about titles and more about future readiness. She’s anticipating an “agentic enterprise era” where orchestration, content variation, and workflow automation will accelerate. In that world, strategy and insight become even more valuable, and execution needs systems, templates, and AI-enabled leverage to keep up without burning teams out. 6) ABM personalization: use AI to recommend and aggregate, not to decide “We are not personalizing at scale with AI making the decisions… What we are doing is using AI to help aggregate signals and make recommendations… with a human in the loop to then make the decision.” This is a clear stance on a hot debate: should AI autonomously personalize journeys at account scale? Hillary says: not yet. Instead, Snowflake uses AI to compress research time and surface next-best actions, then relies on marketers (and sales alignment) to decide what to do. In other words, AI accelerates diagnosis and options, not judgment. This is also where Snowflake’s “unified records” advantage shows up: with customer/prospect data, intent, sales conversations, and use cases accessible together, recommendations can be rooted in reality, not generic persona assumptions. 7) The ABM Workspace: one place to see movement, and take action “I… open this ABM workspace, and it tells me, here's your accounts that have had movement… Here's what we recommend you do next… [and] the ability to actually action all of those things from this one central place.” Hillary’s description is a blueprint for where enterprise ABM is heading: integrated signal detection + recommended actions + workflow execution in a single interface. What matters most is the operational intent. Instead of ABM being a set of disconnected plays (ads here, email there, insights elsewhere), it becomes a guided system that helps a marketer prioritize, coordinate, and execute. This directly reinforces her earlier “do more with what you have” thesis: consolidating decisions and actions reduces overhead, freeing marketers to focus on creative strategy, sales collaboration, and experience design. 8) Cutting through AI noise requires human resonance, not just better targeting “At the end of the day… if you're touching on a true business problem… it should be a gift that you are giving them… ‘get them promoted.’” Kathleen’s observation that conference floors have become a wall of indistinguishable messaging is a common market reality right now. Hillary’s answer is not “more personalization” in the shallow sense, it’s deeper relevance to the individual’s stakes. Her “get them promoted” line is memorable because it forces marketers to translate value props into human outcomes: credibility, wins, internal influence, career success. That’s also why she points to ABM initiatives that feature real executives speaking directly to peers, because credibility and empathy travel farther than auto-generated copy, even when the targeting is perfect. 9) Direct mail that works doesn’t feel like swag, it feels like respect “What is something that can sit on their desk that it truly feels like a gift and doesn't feel like swag?” In ABM, direct mail often fails when it’s transactional or disposable. Hillary’s team anchors direct mail strategy in permanence and meaning: something that earns space in a buyer’s environment, not something that gets tossed. She also notes that “free AirPods aren’t that incentivized” for senior audiences, an important reminder that executives don’t respond to the same incentives as mid-market buyers. This ties neatly back to her human-centric ABM philosophy: the goal isn’t gimmicks; it’s demonstrating understanding, and earning attention in a crowded, automated world. 10) AI ops as a force multiplier: triage and “command centers” for funnel health “We’re building like a command center… to say… ‘Here’s the broken piece of the pipe’… I actually don't want it to just alert you. I wanted to go fix it and then tell you that it fixed it.” This closing section is a glimpse into where marketing operations is headed: from reporting to remediation. Hillary gives two examples: an AI-driven reply triage system for inbound responses (“polar reply”), and a pipeline leakage command center designed to detect where leads fall out and recommend (eventually execute) fixes. It’s the logical endpoint of her earlier thinking about transformation: not just automating tasks, but redesigning the operating model so systems watch the machine, diagnose issues, and increasingly resolve them, while humans focus on strategy, alignment, and creativity.