AI can make almost anything in marketing faster. That does not mean it belongs everywhere. As Sajeel Qureshi, CEO of Computan, put it on a recent episode of the company's Unwatchable podcast, banana peppers are great on a pizza and great on a sandwich, but nobody should put them in a peanut butter and banana sandwich just because they happen to have a jar open. AI works the same way. The right question is not where can we use AI. It is where does AI actually make things better.
TL;DR
- Consumers notice AI generated marketing and it tends to cost brands trust, not build it.
- Customer success and other relationship driven work should keep a human at the center, with AI supporting from behind.
- AI written content and sales emails need human editing and judgment before anything goes out, not after something goes wrong.
- Founder and executive content, and anything where the person is the product, loses value the moment it is handed to an avatar.
- "We have AI" is not a strategy. AI belongs where it removes real repetitive work, not everywhere it technically can be added.
- The best use of AI in marketing is automating the work around the relationship, not the relationship itself.
AI Is Powerful, But It Doesn't Belong Everywhere
AI is showing up in almost every marketing workflow now, from content drafts to prospecting to customer support. The temptation is to automate everything simply because AI can technically do it. That temptation is exactly where a lot of marketing teams go wrong.
The data backs this up. When consumers notice visibly AI generated content in a brand's marketing, only 7 percent say it makes them trust the brand more, while 31 percent say it makes them trust the brand less, a gap of more than four to one.[1] Seventy percent of consumers say they can usually tell an AI generated ad is missing something, and 74 percent say they are more likely to buy from an ad they believe was made entirely by humans.[2] Visibility is the problem. AI behind the scenes, doing research and drafting and organizing, is not what customers are reacting to. AI standing in for a human, badly, is.
Don't Use AI To Replace Human Relationships
Some marketing activities work because there is a real person behind them. Customer success is the clearest example. AI can answer FAQs well, but removing the human entirely can strip out the rapport that customers actually value. Sajeel made this point directly on the podcast: when you remove humans from something like customer success and let a robot answer everything, that is not the point of the interaction. The point is a person giving a quick, real answer and building a bit of rapport along the way.
The data agrees more than it disagrees. Seventy nine percent of Americans strongly prefer a human agent over an AI one, and 84 percent believe human agents are more accurate, even when that is not objectively true.[3] Human handled interactions score 88 percent satisfaction versus 60 percent for AI alone, a 28 point gap that has not closed over the past two years.[4] Half of consumers say they would consider cancelling a service over AI driven support that goes too far.[5]
The better model is AI assisting the human, not AI becoming the entire interaction. Use AI to find information, summarize conversations, prepare responses, and surface customer context. Keep humans responsible for relationship building, sensitive conversations, understanding nuance, and high value interactions.
Don't Let AI Write Your Marketing Without Human Editing
AI generated content can quietly turn into what Sajeel called empty calories on the podcast: writing a blog post from a couple of prompts and publishing it, generating LinkedIn posts with no personal experience in them, producing generic email copy, or creating content simply to hit a publishing schedule. It is thoughtless AI use, content made because AI makes it easy rather than because there is something worth saying.
Joe Jerome, CRO of Computan, made the same point from the buyer's side of that transaction: a couple of prompts and calling it a day is exactly the slothy stuff that reads flat and does not land. Consumers notice. More than half of respondents in one 2026 study said they are annoyed by AI generated social posts, machine personalized emails, and AI written articles specifically.[2]
A better sequence looks like this: human insight, then AI assistance, then human refinement, then publication. AI is genuinely useful for structuring a draft, organizing research, editing, repurposing existing content, and brainstorming angles. The original insight and the final judgment should still come from the marketer.
Don't Send Obviously AI Generated Prospecting Emails
Automation does not excuse a bad message. AI prospecting tools can identify the right people and send at scale, but a prospecting email still has to feel relevant and intentional to land. The problem is not necessarily using AI to draft the first version. The problem is sending that first draft without asking whether it sounds like your brand, whether it is actually relevant to that person, and whether a human would genuinely send it as written.
Joe described this failure mode directly on the podcast: he sees plenty of sales prospecting emails that are clearly AI, and it shows, because a prospecting agent can send the email, but the message itself still needs to be sharp, well written, and genuinely eye catching, not just technically sent. His own workaround was to have AI draft based on his real emails and his sales training, then edit most of it himself before sending, because letting AI send everything unedited is, in his words, always a bad idea.
The numbers back up the instinct. A large A/B test of 50,000 AI generated cold emails against 50,000 human written ones found the AI batch replied at 4.1 percent versus 5.2 percent for the human batch, and got flagged as spam 8 percent of the time versus 3 percent for the human emails, nearly three times the rate.[6] Fully autonomous, unedited AI cold emails trail experienced sales reps by a wide margin, and that gap widens sharply on senior, high value contacts.[7] Automate the research and the workflow. Do not automate away the thinking.
Don't Replace Your Brand's Personality With AI
Marketing is not only about efficiency. Voice, perspective, and personality are part of what makes a brand recognizable, and excessive automation tends to sand that down until every brand starts sounding the same. This is a particular risk for founder led marketing, personal LinkedIn content, executive communication, thought leadership, and community engagement, exactly the places where a distinct voice is the whole point.
AI can imitate a writing style reasonably well. It cannot generate a genuine point of view. That is a human contribution, and it is the same piece missing from the generic LinkedIn posts and prospecting emails covered above.
Don't Create AI Avatars When The Human Is The Product
Sometimes people want a specific person, not an AI version of that person. Joe described exactly this happening to him: he was sent a training built around a virtual avatar of a real person he knew, downloaded the underlying PDF instead, and fed that to AI directly, because he was not willing to sit through three hours of fake him. He said it plainly afterward, that watching an avatar stand in for a real person just was not going to happen.
This risk shows up most in training, executive thought leadership, personal brand content, customer education, and relationship driven sales, anywhere the specific person is genuinely part of the value being delivered. The question worth asking before building an AI avatar is simple: is the person part of the value proposition. If the answer is yes, replacing them removes the exact thing the audience came for.
Don't Use AI For The Sake Of Saying You Use AI
Having AI is not a marketing strategy. Adding AI to a workflow does not automatically make that workflow better. It is worth skipping AI when the existing process already works well, when AI adds complexity without a real payoff, when a person can do the task faster or better without it, when the output still needs more cleanup than doing it manually would have taken, or when there is no meaningful benefit to the customer or the business on the other end.
The podcast comes back to this idea repeatedly in different forms: know where AI fits before reaching for it, rather than applying it indiscriminately because it is available. That discipline is what separates a thoughtful AI workflow from a technically impressive one that nobody actually benefits from.
Don't Confuse AI Quality With AI As A Category
Not every AI implementation is equally capable, and treating all AI as one thing leads to bad conclusions. Different models have different capabilities. Some have real context and access to connected data. Some are one off prompt interfaces with nothing behind them. Some are agentic and wired into real workflows. Some are simply configured poorly.
Joe made this distinction explicitly: when you are working with a strong model, you can choose it, connect it, and give it real context, but plenty of AI experiences people run into are running an outdated or disconnected model behind the scenes, and it shows. So when an AI experience fails, the more useful question is not does AI work. It is which model is being used, what context does it actually have, what systems can it reach, how was the workflow designed, and is it even the right task for AI in the first place.
Where AI Does Belong In Marketing
After all of that, there is a clear list of places AI genuinely earns its keep. Operational workflows are the strongest fit: CRM logging, follow-up reminders, morning briefs, task generation, meeting summaries, and sales pipeline reviews. Connecting disconnected systems is another strong fit, pulling together the CRM, email, meeting transcripts, project management, knowledge bases, and reporting tools that otherwise sit in silos. AI is also genuinely good at surfacing information humans would otherwise miss: who needs a follow-up, which prospect interacted recently, which tasks were never completed, and what actually happened in yesterday's meetings.
Joe described exactly this kind of workflow running at Computan: AI prepares the team for a daily sales meeting by pulling from emails, transcripts, and CRM activity, the team assigns actions during the meeting, and the next day AI checks whether those actions actually got done. That is AI doing the remembering and the connecting, while people still make the calls and have the conversations.
The Better Marketing AI Model: Automate The Work, Not The Relationship
This is the core framework worth taking away from all of it.
| Use AI to | Keep humans responsible for |
|---|---|
| Gather information | Building relationships |
| Summarize conversations | Important conversations |
| Identify follow-ups | Deciding how to approach someone |
| Draft content | Adding perspective and personality |
| Log CRM activity | Strategic decisions |
| Analyze patterns | Understanding nuance |
| Prepare meetings | Having the meeting |
Every row on the left saves real time. Every row on the right is exactly where customers, prospects, and colleagues can tell the difference between AI and a person, and where that difference actually matters.
The Banana Pepper Test For AI
Before adding AI to a marketing workflow, it helps to run it through a short test, borrowed loosely from Sajeel's own analogy. Does AI make this better. Does it save meaningful time. Does it improve the experience for the customer on the other end. Does it preserve the human element in the places where a human element actually matters. Would this workflow still be worth keeping if you removed the word AI from the description entirely.
If the honest answer to most of those is no, AI may be sitting somewhere it does not belong, the marketing equivalent of banana peppers on a peanut butter sandwich. AI is powerful enough to genuinely improve a workflow, or to dominate it in the wrong way. The difference comes down to where it gets placed.
Don't Automate What Makes Your Marketing Human
AI should make marketers more capable, not less human. The goal was never maximum AI adoption. It is thoughtful AI adoption: using AI where it removes repetitive work, connects information that would otherwise stay scattered, and helps people make better decisions faster, while keeping humans exactly where trust, creativity, judgment, and relationships are the actual product being sold.
The smartest marketing teams will not be the ones using AI everywhere. They will be the ones that know exactly where to use it, and just as importantly, where to leave it out.
Sources
- EMARKETER: Visible AI in Marketing Is Four Times More Likely to Cost Brands Trust Than Build It
- MarTech: Consumers Want AI Ads With a Human Touch
- Eesel AI: Chatbot vs Live Agent, What the 2026 Data Actually Says
- Stealth Agents: AI vs Human VA Statistics
- AnswerFirst: 17 Customer Support Statistics Your Business Needs to Know for 2026
- Lobsterpack: AI Cold Emails Get Flagged as Spam More Than Human Ones
- Komo: AI Email Writer, Sales Productivity Data
Frequently Asked Questions
Should marketing teams avoid AI in customer facing content altogether?
No. The issue is not AI itself, it is visible, unedited AI standing in for a human where a real connection matters. AI is generally fine behind the scenes for research, drafting, and organizing. It becomes a problem when it replaces the human element customers actually came for.
Is it ever okay to use AI to write a first draft of marketing content?
Yes, as long as a human adds the original insight and does the final editing before it publishes. The problem is publishing an AI draft unedited, not using AI to get past a blank page.
Why do AI generated prospecting emails perform worse than human written ones?
Recipients and spam filters can often tell the difference, and unedited AI emails get flagged as spam and get lower reply rates than human written or human edited outreach, especially with senior contacts.
Where does AI genuinely help a marketing team?
Operational work: CRM logging, follow-up reminders, meeting summaries, connecting scattered systems, and surfacing information a person would otherwise miss, like who needs a follow-up or what happened in yesterday's meetings.
How do you decide whether a workflow is a good fit for AI?
Ask whether AI genuinely saves meaningful time, improves the customer's experience, and preserves the human element where one actually matters. If removing the word AI from the description would make the workflow pointless, it is probably not a good fit.
Should founders and executives let AI handle their personal content and communication?
Be cautious. Founder led content, thought leadership, and personal brand communication work specifically because a real person's voice and perspective are behind them. Handing that entirely to AI tends to remove the reason the audience was paying attention.
