The issue with AI sales agents
It’s our lives now. We open our inbox or LinkedIn DMs, only to be greeted by a wall of messages that all seem to say the exact same things, with very slight variations. They praise your recent post, pivot sharply to a “common pain point”, and ask for a quick 15-minute chat.
The promise was beautiful: AI sales agents were supposed to supercharge the “SDR part” of the sales process, bringing it to new heights. They were going to personalise emails at hyper-speed, find the perfect contacts instantly, and scale outbound sales to levels humans could only dream of.
Instead, we might be building the ultimate telemarketing harassment tool.
Let’s dive into one of the fatal flaws holding AI sales agents back, why the current scaling model is broken (or mis-built, to be more precise), and what it would actually take to build an AI agent that works without causing an exodus from email entirely.
The SDR “Terminator”
On paper, AI sales agents look like the absolute cheat code for B2B growth. A human SDR can perhaps research and send a few dozen highly personalised emails a day before their brain turns to mush. An AI agent will send thousands an hour.
They can scrape the web, find the right contact, match them to an ideal customer profile, and pick a tailored template in milliseconds. It sounds like pure efficiency.
But there’s a massive catch. Because AI can send thousands of emails instantly, companies are letting them do exactly that. The result isn’t a spike in closed deals; it’s a collective closing of corporate garage doors.
“Templates’ fatigue” on steroids
To understand why AI is struggling, we have to look at human SDRs first. For years, the sales cycle has followed a predictable pattern: a LinkedIn influencer shares a “perfect” opening line or email subject, thousands of sales reps copy it, and within a week, the template is completely dead because prospects recognise it instantly.
Human SDRs ruin good templates over the course of a few months.
AI sales agents ruin them in about forty-eight minutes.
When you give AI the power to reach out to an entire market simultaneously, the lifecycle of an effective pitch shrinks to zero. Because these tools lack genuine imagination, they lean heavily on optimised patterns. And when everyone uses the same optimised patterns, nothing stands out.
The “by the book” Trap
You might think the solution is simply removing the obvious giveaways. We all know the classic AI tells by now: the weirdly formal bullet points, the overly enthusiastic greetings, or the strangely structured transitional phrases.
Get Lucas Nuber’s stories in your inbox
Join Medium for free to get updates from this writer.
But even when the engineering teams get feedback and improve those issues, the problem remains. The emails are still too perfect, too polished.
Real human communication is a little messy. It has quirks, occasional casual phrasing, and a distinct lack of flawless corporate symmetry.
AI sales agents are incredibly polished, and paradoxically, that polish is exactly what sets off our automated mental spam filters, our human “gut feeling”. It sounds exactly like what you expect a sales bot to say. The moment a prospect senses they are reading a generated script, the trust evaporates.
The constant retraining threat
If the market catches on to an AI’s style within days, the only logical solution would be to completely change your copy, approach, and persona all the time, every two days even. It would be one way to stay ahead of the curve.
But is that actually sustainable?
Constantly pivoting means continuously retraining or re-prompting your AI agents, tracking which styles are burning out, and trying to stay one step ahead of an increasingly cynical audience. If your sales operations team has to spend all their time managing the AI’s existential identity crisis, you haven’t actually saved any time or labour.
Find some zen: restraint and subtlety
To give a chance to AI sales agents to succeed long-term, the industry would need a big shift in philosophy.
- Enough with the “it’s a numbers game” nonsense: The goal shouldn’t be to send 10,000 emails a day. The goal should be 50 emails that look so incredibly organic, human, and tailored that they demand a response. As always, quality over quantity!
- Weaponised multiple personality disorder: AI must learn to randomly vary its tone, structure, and intent across hundreds of different micro-variations so no two prospects receive the same “vibe”. While keeping it professional and industry-appropriate.
The big challenge right now isn’t data scraping or speed — it’s self-control.
Let’s do it right (this time)
AI sales agents have the technical capability to revolutionise outbound growth, but right now they are driving us toward a future where people simply stop opening cold emails altogether.
Until AI can replicate true human imagination and step away from the “perfectly polished” pitch, it will continue to face diminishing returns. Speed is useless if you’re sprinting straight into a wall.
