AI BDRs are burning through their own market

The experiment

I was running a 4-month experiment with an AI BDR tool. I gave it everything it needed: an explanation of the product, the value it generates, customer case stories, etc. Then defined who I wanted to target. The tool built the lists and drafted first emails; then, after multiple iterations, I got instructions to the point where the AI didn’t drift, didn’t position the product wrong, and stuck to the communication style I wanted.

It ran research on each contact, online and on LinkedIn, as a good BDR would, and tried to find ways to personalise the messages. Here I was able to see the first issue: it used anything for personalisation. Even stuff that wasn’t relevant and created very strange bridges to start pitching. Or it congratulated a prospect on a new role; the only problem was that person had started in the role 1,5 years earlier. For other contacts, it opened with “I saw your post about leaving the company X” and sent that message to the company X email 2 months after that farewell post on LinkedIn.

I do believe that these problems can be ironed out in the near future. At a certain point, I just had to be content that the majority of emails were decent and let it run autonomously, instead of me checking every outgoing email. If I checked all of them, I would have been a bottleneck, killing the main value: volume.

So did we get results? After four months and after thousands of messages and 3-4 thousand prospects later, we have opened a handful of conversations, out of which only a few have led to opportunity. Not that great a result. As opportunities are still open, I cannot tell what the CAC for this experiment was. Also, the reputational risks of this experiment are something I cannot measure directly, but something to be aware of.

We decided to pull the plug and start testing other ways to do outbound in today’s noisy world. We had used AI to reach thousands of prospects, but it had not opened enough promising conversations to justify continuing with this approach. It could be that wrong assumptions in targeting or offer contributed to the results.

Better personalisation does not solve the whole problem

AI can insert relevant details to make the messages sound personalised, but relevance is not the same as valuable insight.

Most AI-generated outbound follows the same recognisable pattern:

1. Mention a recent event. Whatever you can find from public sources: new hire, new big client, expansion plans, mergers, etc.
2. Connect it to a presumed business problem, even when it has no connection.
3. Introduce the product.
4. Ask for a short meeting.

The details vary, but the structure and logic remain almost identical. The recipient does not experience this as thoughtful personalisation. They experience it as a better-produced template.

Good personalisation in cold outreach no longer makes you stand out. I myself have received hundreds of messages with the same template.

Everyone can send thousands of personalised cold emails; they just need to make the decision to use the tools. What’s missing is the understanding that the recipient is a human being. A company closing a huge customer and a problem derived from that is not necessarily relevant to the recipient and the problems he is facing in his role.

Researching a trigger is easy. Understanding why it matters to this particular person, and saying something they have not heard before, is much harder.

• Personalisation says, “I know something about you.”
• Personal significance says, “I understand something important to you.”
• Insight says, “I may see something you have not considered.”

The outbound arms race

Suppose AI-generated outreach becomes as good as or better than the work of a competent human BDR. Even if AI could send unique messages with perfectly varied angles and “templates” by judging what type of approach could work for the prospect, the market still does not improve, because every sender gains the same capability. There will still be the issue of volume; recipients’ inboxes will still be flooded.

This creates an outbound arms race:

1. AI lowers the cost of reaching out to more and more people, which has already happened.
2. Companies send more messages in different channels.
3. Buyers receive more apparently personalised outreach.
4. Buyers raise their filtering threshold.
5. Sellers increase volume or improve automation.

This continues until inboxes turn into a nuclear wasteland.

I recently saw a discussion on LinkedIn where an AI BDR received a response from the AI assistant of the recipient. AI assistant tried to unsubscribe, and the AI BDR tried to convince the assistant that the message should be delivered to the human. Is that going to be the future? AI BDRs trying to get past the first gatekeeper?

Attention and trust in your market are not infinite; they are finite resources. The AI arms race burns that resource very fast. Think for a moment: you receive 50 cold outreaches in a week; how long do you bother to open any emails or DMs from strangers, or, better question, how many do you open today? Most likely, 90% of them are irrelevant to you, or the timing is not right anyway.

So the more volume we produce, the less it helps you get attention. This will also make it harder for the human BDRs if they continue competing for attention on the same battlefield as tireless machines.

AI BDRs do not merely compete for attention. They reduce the value of the attention channel itself.

What becomes scarce

Historically, the bottleneck was labour: finding prospects, researching accounts, and writing messages took time.

AI removes much of that bottleneck. But it exposes a scarcer one underneath: credible reasons for a buyer to care.

The new constraints are things AI cannot (at the moment) manufacture cheaply:

• Reputation
• Trust
• Original insight
• Genuine relationships
• Demonstrated expertise
• Strong timing
• A meaningful point of view
• Offers that are unusually valuable or specific

Sending a message costs almost nothing now. But developing something worth saying still takes work. That can mean sharing relevant proprietary insights, offering custom analysis, or expressing a real opinion with the risk of disagreement from the recipient. The effort alone does not make the message valuable; it still needs to mean something to this particular person. AI can help prepare the message, but the knowledge, experience, and relationships behind it take effort to develop.

What I want to test instead

I have been thinking lately about what will make sense in outbound. For us, increasing volume is no longer the direction I want to pursue. My bet is that as AI-generated outreach grows, volume alone will become a less effective way to compete for attention. I want to test a smarter, higher-quality, lower-volume approach with a better hit rate. But that approach also needs to make economic sense. How much effort per account can we justify with our average contract value? How much should that effort improve the number of qualified opportunities and our win rate?

Here are some thoughts I have had.

Earning attention and trust

Network-driven distribution: Introductions through customers, investors, partners, colleagues, and communities. Someone you already know and who is ready to vouch for you can open doors. The main limitation here is your own network size. Well-networked people and community relationships become valuable assets.

Public expertise before private outreach: Publish genuinely useful thinking, then contact people who have already encountered it. If your audience already recognises your name and face as someone who knows his stuff, it can help to open doors. This can be a company brand or personal brand. For smaller companies, building a brand that really stands out is not an easy ask. Personal brands can help here. One caveat in personal brands too: if you are planning your GTM and you have not invested in building your brand and reputation well before, it will take a lot of time and effort. Also, AI is making this harder in digital channels; a lot of AI-generated content is filling the space, and getting any visibility, even with smart thoughts, is hard.

Small events and in-person meetings: Create contexts where trust can develop before a sales conversation begins. This could be the direction we move from digital channels after the AI BDR apocalypse. Back to the old days when sellers built relationships and trust face-to-face.

Providing value before asking for a meeting

Customer-specific work: Send an analysis, prototype, benchmark, teardown, or recommendation that has standalone value. Not an easy task. If you can really provide some real value to the prospect at the first touch, you have a better chance to get their attention. But do not send me an email: “I noticed multiple issues with your website...”

Choosing the right moment

Trigger-based restraint: Contact fewer companies, but only when there is a strong reason to believe the timing is right. Right timing combined with the perfect angle could do the job, and here I believe AI can help you.

My focus now is building trust and creating opportunities for our sales team to have human-to-human interactions. One approach I want to test is bringing a small group of relevant people together around a problem they share. The event needs to be useful in its own right, with room to compare experiences and discuss what is working. We would judge it by the relationships and qualified follow-up conversations it creates, and whether those justify the work involved.

Where AI can help

I’m not against using AI in your GTM; I’m myself using AI daily. The product I currently work with is an AI co-worker for sales. I do believe there are really good use cases for AI. But not for generating higher volume. AI can help in many ways.

Constructive uses include:

• Identifying a small number of genuinely relevant accounts
• Detecting high-confidence timing signals
• Preparing humans for conversations
• Summarising complex company context
• Challenging weak assumptions about a prospect
• Improving clarity after a person has chosen the argument
• Preventing irrelevant outreach from being sent

But using it primarily to increase sending capacity is likely to do more damage to everyone instead of helping you.

The winners may not be the companies with the most sophisticated message-generation systems. They may be the companies that create a strong brand and reasons for prospects to recognise, trust, or seek them out before the message arrives.

The future of outbound may still involve AI. But it will not be won by whoever can send the most convincing imitation of a thoughtful message.

It will be won by whoever has done enough real work that the thoughtfulness is not an imitation.