Back to Blogs
Drew Neisser is the founder of CMO Huddles and a globally recognized authority on B2B marketing. He’s an AdAge columnist, LinkedIn TopVoice, leading CMO coach, podcast host & friend of penguins everywhere.
CMOs should build three capabilities before scaling AI across go-to-market operations: clean data, clear processes, and shared customer context. Drawing on four CMO Super Huddle workshops led by LeanData CMO Jim Bell, this practical guide explains how to strengthen that foundation, introduce agents within defined workflows, and connect AI activity to customer and revenue outcomes.
“We’re purposely not buying any more technology until we can get our culture straight,” one CMO confessed during a workshop at the CMO Super Huddle. Dirty CRM data, organizational silos, and inconsistent pipeline management were already preventing marketing from doing its job. Adding more AI would only accelerate the confusion.
That observation captured the central lesson from four workshops led by Jim Bell, CMO of LeanData: AI performance depends on the go-to-market foundation beneath it. According to research Jim shared, 93% of GTM teams have deployed at least one AI agent, yet only about one-third believe their infrastructure is ready.
Jim organized the challenge around three requirements: clean data, clear processes, and shared customer context. Together, they provide CMOs with a practical framework for deciding whether a promising AI experiment can become a dependable business capability.
Workshop quotes have been lightly edited for clarity and anonymized, except for comments from Jim Bell.
“Bad data” is a popular diagnosis because almost everyone can agree with it. Unfortunately, telling a CMO to clean all company data is about as useful as recommending that they boil the ocean before making tea.
Jim offered a more practical approach. “You’re never going to perfect your data, even at the giant companies we work with,” he explained. Instead, companies need checkpoints that improve and protect data as it enters and moves through a specific workflow.
One CMO described how unreliable persona and industry fields had stalled their personalization efforts: “We were struggling with the most basic things: industry and persona. I wasn’t confident enough to do the segmentation we needed because I didn’t trust the data. You can do more harm by sending the wrong thing than something more generic.”
The data challenge also extends well beyond structured CRM fields. One CMO observed that “half the things that matter in an opportunity tend to be in unstructured data.” Sales-call recordings, emails, meeting notes, support interactions, and other conversational sources often contain buying signals and decision-makers that never appear correctly in Salesforce.
Several CMOs pointed to recorded sales calls as a particularly valuable source. “Our Salesforce data is getting better, but it still sucks,” one said. “Gong is a source of truth because it captures what’s actually happening. We can roll that up across 350 salespeople and break it down by territory, region, or specialty. That’s gold.”
Another described using AI to convert call recordings into structured Salesforce updates automatically. The result was better information entered immediately instead of incomplete notes appearing days later, if they appeared at all.
Sometimes, removing the human from data entry improves the human understanding of the customer.
CMOs do not need to solve every data problem before moving forward. They should select one important workflow and identify the information required to run it reliably. That means establishing the authoritative source for each field, setting validation rules at critical entry points, incorporating relevant unstructured data, and assigning someone to monitor quality over time.
Many GTM processes work because experienced employees quietly compensate for their weaknesses. They recognize a duplicate account, know which salesperson should receive an unusual lead, or notice that an otherwise attractive customer is in the middle of a serious support escalation.
Those judgment calls rarely appear in the official process map.
One CMO described the problem as “all these assumptions we have as humans.” They continued, “It’s tacit knowledge that humans possess. We wouldn’t even think to document it until the thing hallucinates, and then you realize it hallucinated because of something you knew inherently.”
Jim made a similar point: humans naturally pause when something seems suspicious. Agents follow the information and instructions they receive, even when an experienced employee would recognize that the output makes little sense.
This is why one GTM engineering leader stopped thinking primarily about prompting agents and started “onboarding them like a real human being.” Their team supplies the context behind the instructions, includes the people affected by the workflow, and creates a process for updating the agent as new information emerges.
Another CMO treats agents as defined roles inside the operating model. “We give the AI a job description, reporting lines, and metrics,” they explained. “That helps us define the process, justify why we want the role, and clarify the handoff between the human and the AI.”
That level of definition also helps prevent agent sprawl. If an agent lacks a clear job, owner, input, output, success metric, and handoff, the organization probably does not need to build it.
Jim also emphasized the distinction between inference and deterministic execution. AI can interpret an inquiry or summarize an interaction, but certain GTM decisions need to produce the same correct result every time. Account ownership, territories, contact information, compliance requirements, and escalation rules cannot depend on an agent’s creative interpretation.
Before automating a GTM process, CMOs should document:
One leader recommended starting with repetitive, bounded work: “Anything you do ten times a week, let’s look at automating that. We’re starting with the small ones before saying, ‘Let’s build the pipeline,’ because our stack and tools aren’t ready for that, and we don’t have a clear owner.”
The goal is to earn the right to scale.
An agent can have accurate data and still create a terrible customer experience if it sees only one departmental slice of the relationship.
Marketing may be launching an expansion campaign while customer success is managing a severity-one issue. Sales may be pursuing a product that the customer has already rejected. Three teams may contact the same executive with different messages because their systems and incentives do not reflect the full relationship.
One CMO described the need for “one voice, one company” after discovering multiple teams pursuing the same people at the same financial institutions. Their organization created a customer-focused team and began combining conversational intelligence with outreach data to understand who was contacting each account and whether the customer was in a good position to hear from them.
Another CMO found that AI-assisted customer research improved executive meetings and shortened the path to revenue. Their team used external research, account information, recorded conversations, and executive commentary to prepare for customer discussions. In one case, that broader context revealed that the sales team was promoting the wrong solution.
“We were able to provide something more relevant to the customer rather than a generic presentation,” the CMO explained. The process also returned insights to product teams faster and helped marketing demonstrate its contribution to sales velocity.
Shared context requires coordination across marketing, sales, customer success, support, product, and operations. It also raises an uncomfortable question: Who owns the complete customer experience?
Marketing has a strong claim because it already worries about the buyer journey, brand, market understanding, and the consistency of the company’s story. As one CMO argued, the traditional funnel helped create functional silos. Now that companies are rediscovering the customer journey as a flywheel, “Who’s in charge of the flywheel? Marketing.”
Ownership will vary by organization, but someone must be accountable for connecting the signals. Otherwise, AI will help every silo move faster in a different direction.
The workshops also revealed a useful tension. AI can produce personalized content at extraordinary speed, yet that very capability is making personalized content less remarkable.
One CMO worried about receiving increasingly polished materials from every vendor: “We can produce beautiful personalized landing pages, emails, meeting follow-ups, and event follow-ups. But now I’m getting them from all kinds of vendors. There’s so much that I’m going to start tuning it all out.”
Another CMO saw the countertrend: “AI is going to make human experiences and human touches much more valuable. We could sit and watch webinars or read content all day, but this is what we want.”
A third shared an example involving personal invitations to a London event. The outreach took more time because a team member wrote directly to a carefully selected group instead of automating the messages. Prospects responded warmly even when they could not attend. “People could sense it was a real person writing,” the CMO said. “It feels like it’s going back to the human.”
AI should help teams understand customers, prepare better conversations, reduce administrative work, and respond more intelligently. That creates room for the judgment, creativity, and personal interaction that customers increasingly value.
More personalized output is easy. A more valuable customer experience takes work.
CMOs can begin with one customer-facing or revenue-critical workflow and conduct a focused review:
Jim’s framework is valuable because it transforms AI readiness from an abstract technology conversation into an operating discipline. Clean data, clear processes, and shared context will not eliminate every AI mistake. They will give the organization a much better chance of recognizing, correcting, and learning from those mistakes before they reach the customer.
Start by assessing one important GTM workflow against these three requirements. Then use the CMO Huddles AI Maturity Calculator to evaluate the broader organizational capabilities supporting your AI transformation.
An AI-ready GTM foundation combines trustworthy data, a documented end-to-end process, and shared customer context across marketing, sales, customer success, support, and operations. It also assigns clear ownership for quality, governance, cost, maintenance, and business outcomes.
No. Jim Bell recommends improving and protecting the data required for a defined workflow rather than waiting for a company-wide cleanup. CMOs should establish authoritative sources, validation checkpoints, and an owner who monitors quality over time.
Begin with repetitive, bounded work that has a clear owner, observable output, and measurable result. Document the process and its exceptions before introducing an agent, then test internally before moving into higher-risk customer-facing use cases.
Use AI to assemble customer context, prepare better conversations, reduce administrative work, and coordinate outreach across teams. Preserve human judgment and interaction for moments where trust, creativity, and nuanced understanding create the greatest value.