The Marketers Making AI Work in ABM All Changed One Thing First

A 2026 audit by Digital Applied across 14 B2B SaaS GTM teams found that agentic prospecting fed by intent data runs at a 31 to 47% false-positive rate. 

Meaning for every 10 accounts the system flagged as in-market, 3 to 5 weren't. According to one RevOps director in the audit, "Half our AI-SDR capacity was running against accounts that were not actually in-market. The intent-data layer was the bottleneck. The AI was working fine."

That last line highlights an issue so many ABM teams are dealing with. AI is only as effective as the underlying data and systems you provide.

We recently interviewed three practitioners to dive into this exact topic. What we found were three stories that led to the same conclusion.

For AI tools to be most effective, you need to do the groundwork first. Here’s what that looks like practically.

Eighteen months ago, Justin Lopez, Sr. ABM Manager at Bonterra, had a handoff problem.

Bonterra’s intent signals and engagement data lived in separate systems, which meant behavioral signals and the actions they should trigger weren’t connected automatically. Someone had to bridge them manually. And manually meant slowly.

"Successful handoffs weren't common enough," says Lopez.

To fix the problem, Lopez rebuilt the process and set up a scoring agent that evaluates contact-level fit across Bonterra's Influ2 ad audiences—which specific contacts are actively showing buying signals, ranked by fit. 

Sales now receives those contacts with context already attached.

The problem Lopez fixed isn't unique to Bonterra. 

Salesforce's 2026 Connectivity Benchmark Report found roughly 50% of enterprise AI agents currently operate in complete isolation with no shared data or a unified view.

Picture detectives working the same case from separate rooms, each with their own board of clues they've never shared. Each one might be close. Together, they'd have it.

Once scoring and handoffs were a single motion, the program moved when buyers did, rather than when someone got around to it.

Nadia Davis, VP of Marketing at CaliberMind, has seen what’s possible in ABM with AI from a unique angle. She's watched the shift to agentic tools play out across the programs that run on CaliberMind's analytics platform, and she pointed to one scenario in particular. 

For years, audience management meant static firmographic filters, a CSV export, and a manual upload into LinkedIn or an ABM platform. 

Even when CRM integrations improved, segments stayed largely frozen in time. Most marketing engagement data doesn't live in Salesforce, and getting it into your activation platform meant either custom builds or accepting that your audiences were always a step behind. As a result, programs ran against stale data by design.

Agentic tools are making that process easier. Instead of exporting and uploading, you can connect your data sources to an agent, describe what you want in natural language, and the agent pulls the most recent data across every connected source to build the segment in real time.

But Davis is careful about how she frames this. This entire approach only works when it's built on a trusted, unified data layer that pulls from every source a program touches, not just what's in Salesforce.

Without that foundation, connecting Claude or an agentic tool to disconnected data sources produces the same stale results. The data layer is what determines the reliability of the output.

Lori Aizer Bryenton, a fractional CMO with over 25 years of experience, shared a story that many marketers can relate to right now. Her program wasn't struggling with signal quality or audience staleness. It was struggling with personalization at scale.

She needed to build a laundry list of assets, including:

• Account-specific one-pagers
• Tailored email sequences
• Competitive framing built around what a specific account actually cared about

Creating that content meant starting from scratch every time. Pulling messaging from one document, competitive context from another, and digging through call notes for what a particular account was focused on. That process put a ceiling on how many accounts the team could work at once.

She tried to shortcut it with AI-generated outreach triggered by account signals. The hypothesis was if you give AI the right signals and prompts, it should be able to produce something close to what a good rep or marketer would write. 

But she found the content read like AI content. It wasn’t obviously robotic, but in the way a savvy buyer can feel when something was generated rather than written. 

She realized the gap between AI’s first draft and something a targeted prospect would actually engage with was where the real work lived. "We were treating that gap as a QA step," Bryenton says. "It's actually a creative and strategic step, and it can't be skipped or rushed."

A lot of marketers would’ve tried solving the problem by buying a better AI writing tool. But Bryenton knew there was a deeper issue an AI tool couldn’t fix.

Instead, her team created a structured context library in a format AI could use consistently across every session. The library included: 

• ICP definition
• Approved messaging hierarchy
• Brand voice guidance
• What they'd learned from recent wins and losses

She also highlighted something that rarely gets brought up in the AI + ABM conversation: "Almost nobody talks about the unglamorous work of building and maintaining a structured knowledge base."

Once that existed, the production math changed. A set of account-specific assets that used to take a full day now takes a few hours. AI handles the first 70 to 80 percent, while a human still owns the strategy, judgment call on whether the message actually lands for this buyer, and final sign-off. 

Nylo Pereira, Head of B2B Marketing at SRG put it plainly in a LinkedIn post earlier this year.

Every story we shared is proof: 

• Lopez rebuilt the handoff motion. 
• Bryenton rebuilt the knowledge base that content production runs on. 
• Davis named the data layer as the foundation agentic tools depend on to work.

None of them pointed to “a better AI tool” as a solution. They highlighted what needs to be addressed before AI can be truly effective and reliable.

We know the conversations around new AI tools and agentic marketing are dominating LinkedIn right now. But before you get completely Claude-pilled and turn to AI to fix structural issues in your ABM program, take a step back and make sure your foundation is solid first.