Back to Newsletters
GenAI value for B2B marketing goes well beyond first drafts. Andy Crestodina's examples show CMOs how to train AI on personas, find website gaps, identify content holes, analyze email engagement, understand LinkedIn or YouTube performance, and build a voice GPT that improves quality without abandoning human judgment or flooding buyers with sameness across channels and teams.
For too many marketing teams, GenAI started as a faster way to make more words and pictures. Useful? Sometimes. Dangerous? Also sometimes, especially if the result is a larger pile of average content wearing a fresh nametag.
The bigger opportunity is using AI to improve marketing judgment: Finding gaps, analyzing patterns, interrogating assumptions, and helping teams see what the human brain misses because it is tired, too close to the work, or already on its fourth dashboard of the day.
That was the heart of a CMO Huddles Bonus Huddle with Andy Crestodina, co-founder and CMO of Orbit Media Studios. Andy has long been one of the clearest practical voices on content, SEO, analytics, and AI. His examples were not “make me a blog post” parlor tricks. They were usable ways to make marketing smarter.
Andy’s starting point is simple: Do not ask an AI tool to judge your marketing until you have taught it who the audience is.
“If you teach the LLM what the target audience is and train it on their information needs, then you can give it anything (an ad, webpage, keyword list, email, etc.) and it can find the missed opportunities.”
This turns AI into a kind of always-on buyer lens. You can ask it to evaluate a webpage, campaign, ad, email, content plan, or keyword list against the audience’s real questions, fears, priorities, and decision criteria.
Good prompts here ask questions like:
The CMO move: Treat persona-trained AI as a reviewer, not a replacement for customer research.
Your website is usually the most important marketing asset you own, which makes it extra impressive how often teams evaluate it by committee mood, internal preference, and whether the homepage hero has finally stopped moving around.
Andy pointed out why AI can help:
“It's very hard for the human brain to look at a page and identify what's missing.”
Once the LLM understands the audience, upload or paste a key page and ask which buyer information needs are unmet. You can do this for the homepage, product pages, landing pages, comparison pages, or late-stage proof pages.
The goal is not to let AI redesign the site. The goal is to surface gaps humans miss: Missing proof, vague category language, weak objections handling, unclear next steps, or content written for insiders instead of buyers.
The CMO move: Use AI to identify missing buyer context before another round of homepage opinion tennis begins.
Most content strategies are built with good intentions and incomplete inputs. Sales has ideas. Product has ideas. Customers have needs. Executives have favorite topics. Somewhere in that mix, the actual buyer’s missing questions can get lost.
Andy’s view:
“It’s easy to miss things,” explains Crestodina, “but AI is amazing at identifying the holes.”
AI can help map the full set of audience information needs, compare those needs against your existing content, and identify areas competitors are not covering well. That does not mean AI should write the final thought leadership. Please do not hand the market another beige sandwich of machine-generated sameness.
Use AI for the map. Then add your company’s distinct point of view, examples, data, customer stories, and scar tissue.
The CMO move: Let AI find the white space. Make humans responsible for the point of view.
Email remains critical for acquisition, retention, and customer expansion. It also produces data that can be annoyingly hard to interpret when engagement lives in one system, conversions in another, and every export seems designed to humble the user.
Andy used GenAI to analyze when his readers were most engaged over several years. He explained:
“It’s almost impossible to do that analysis on your own because if you add dates and times as a secondary dimension to other analytics, it's just a bunch of numbers, but AI does that beautifully.”
His process was straightforward: Import email analytics and web analytics data, then ask AI to identify days, times, segments, or other factors associated with higher engagement and conversion. You can also ask it to visualize the data in a new way.
The caution: Check the math. AI can be helpful at pattern detection, but CMOs should keep a firm hand on validation, data privacy, and business interpretation.
The CMO move: Use AI to speed up analysis, then verify before changing the operating rhythm.
Social and video platforms often provide plenty of metrics while withholding the exact answer you want. Andy shared one example:
“Take all of your data out of LinkedIn (or YouTube) and ask your LLM to show you the topics that grow your subscriber rate the most.”
He added:
“There are no metrics on LinkedIn or YouTube right now that show new subscribers by topic but AI organizes my content into topics and then finds the correlation between topics and new subscribers and puts it on a chart like a heat map matrix.”
That is useful because channel performance is rarely just about format. It is often topic, hook, audience intent, timing, voice, and distribution. AI can help cluster posts, identify themes, and compare performance patterns across content types.
The CMO move: Use AI to find topic-channel fit, not just to generate more posts.
Custom GPTs can be valuable when they are trained on a person’s real writing, transcripts, principles, examples, and preferred structure. Andy created a voice GPT after uploading hundreds of past articles and transcripts.
“I uploaded hundreds of past articles and transcripts from dozens of videos I've done and now it writes much more like me than it could in the past.”
He also noted:
“You can also instruct your GPT to prioritize the knowledge you upload in the answers it provides to future queries.”
For CMOs, this can help teams draft in a consistent voice, repurpose executive thinking, and speed up first-pass content. But it needs governance. What content can be uploaded? Who can use it? How is accuracy checked? What should never be generated without human review?
The CMO move: Build voice tools to improve consistency and speed, while keeping humans accountable for truth, taste, and judgment.
GenAI’s most useful marketing use cases are not limited to faster drafts. The better opportunity is using AI as an analyst, reviewer, pattern finder, and second set of eyes.
CMOs should encourage practical experimentation while avoiding the two classic traps: Pretending AI is magic and pretending it is useless. The sweet spot is disciplined curiosity. Train it well, ask better questions, verify outputs, and keep your human point of view firmly in charge.
Train an AI tool on target personas, then use it to evaluate webpages, campaigns, emails, and content plans for missing buyer information needs.
It can help identify missing proof, unclear messaging, weak objection handling, and unanswered buyer questions. Humans still need to decide what changes are right.
AI can summarize exports, cluster content by topic, find patterns across channels, and suggest hypotheses. Teams should validate findings before acting on them.
A voice GPT is a custom AI assistant trained on a person’s or brand’s content, style, and guidance so drafts better reflect that voice. It still needs human editing.