Is AI DDoS'ing the Buyer and Killing Cold Outbound?

I was talking to a friend recently who is the CEO of a data startup. He has a good product. He needs customers. So, naturally, he started doing outbound.

It didn't work.

Then he hired people who told him they could solve it. One promised 150 qualified leads. Another promised results in a month. Then three months. Then six months. Different people. Different strategies. Different tools. Different messaging.

After all of that? One lead came through. Not one customer. One lead. And this guy wasn't sitting around waiting for the phone to ring. He was trying everything and anything people told him worked.

Then I spoke with another CEO friend. Same story. Cold email. LinkedIn. SDRs. Agencies. Better targeting. Better messaging. More personalization. Almost nothing.

That got me wondering: what if they aren't doing anything wrong? What if cold outbound itself is breaking? And if that's true, we may be about to make the problem dramatically worse with AI.

I remember doing this back around 2011. Cold outreach wasn't some magical customer-acquisition machine then either. Conversion rates were low. You sent a lot of emails to get a few conversations. But it worked enough.

The problem today is different. There are vastly more startups. Every startup needs customers. Every startup has salespeople. Every salesperson has access to databases containing millions of contacts. And now everyone has AI.

AI can do the whole job:

• AI can research your company.
• AI can find the right executive.
• AI can read their LinkedIn posts and figure out what they care about.
• AI can write a personalized message.
• AI can send thousands of them.
• AI can follow up forever.

We have essentially driven the marginal cost of creating a reasonably personalized sales message toward zero. Which sounds wonderful for sellers. But think about what happens on the other side.

Imagine you're a CRO, CIO or CDO. Ten years ago maybe you received a handful of relevant sales emails. Now you could receive dozens or hundreds. And every one says: "Nitin, noticed your company is investing heavily in AI..." or "Loved your recent post about data quality..." or my favorite: "Given your role at XYZ, I thought this might be relevant..."

We used AI to personalize spam.

The seller thinks: this email is incredibly personalized. The buyer thinks: delete.

Not because the buyer doesn't have the problem. Not because the product isn't interesting. Not because the salesperson wrote a bad email. There is simply too much incoming information to process. Cold outreach is landing on deaf ears not because people don't want to listen, but because listening to everyone has become impossible. That distinction matters.

The numbers say the same thing. Belkins tracked 7.5M cold emails it sent in 2025 and counted replies against emails sent, not opens: 0.45% for the year, and falling within the year, from 0.54% in February to 0.35% in December.¹ Instantly's benchmark across its own platform reports a higher 3.43% average, but even there the top decile is defined by a 10.7% reply rate, which means nine out of ten senders are doing worse than that.²

And the closer you get to the inbox my friends were targeting, the worse it gets. Founders and owners reply at 0.57%. C-level at 0.42%. VPs at 0.32%. Companies with more than 10,000 employees, the ones the enterprise startup needs, reply at 0.22%.¹ Roughly one reply, not one meeting, per 450 emails.

This is the part I find fascinating. Our solution to declining outbound conversion appears to be: send more outbound. Human SDR sends 50 messages. AI SDR sends 5,000.

If every startup deploys AI SDRs, we don't solve outbound. We DDoS the buyer.

And then buyers respond rationally. Their email filters get better. LinkedIn filtering gets better. Executive assistants use AI. Companies deploy AI agents to screen incoming communication. This is already underway at the infrastructure layer: since 2024, Google and Yahoo require anyone sending around 5,000 messages a day to authenticate with SPF, DKIM and DMARC, offer one-click unsubscribe, and keep spam complaints under 0.3%, or get filtered wholesale.³ And on the human layer, Gartner's latest survey found 67% of B2B buyers now prefer a rep-free buying experience, up from 61% a year earlier, with 45% already using AI during a recent purchase.⁴

Eventually we get the wonderfully absurd situation where:

Seller AI
  -> writes message
  -> Buyer AI
  -> decides whether a human
     should ever see it

Two enormous language models talking to each other so that two humans don't have to. At that point we should probably ask whether we're solving the right problem.

This is where I think the model needs to flip. The traditional SDR model is:

Database -> Persona -> Message
  -> Send -> Follow up -> Pray

AI makes every step in that process cheaper. But cheaper doesn't necessarily mean better. If everyone can generate infinite outbound, the scarce resource is no longer the ability to contact someone. The scarce resource is attention. And right behind attention are trust, timing, relevance, and a legitimate reason to talk.

So maybe the AI SDR of the future shouldn't be a message-sending machine. Maybe it should be an opportunity-detection machine. Its job isn't "who can I email today?" Its job is "what changed in the market today that gives us a legitimate reason to talk to someone?"

That's a very different system. Observe, detect, understand, enter. Imagine an AI agent continuously watching your market:

• A company hires a new CRO.
• Another announces they're moving upmarket.
• A CEO talks publicly about retention problems.
• A company suddenly starts hiring 20 enterprise salespeople.
• Someone asks a question in a community about exactly the problem your product solves.
• A portfolio company announces a strategic initiative related to your product.

The AI doesn't automatically send 5,000 messages. It tells the founder: something changed here. This company may have the problem you solve. Here's why. Here's who matters. Here's the context. And here's the most credible way into the conversation.

Maybe that's a direct message. Maybe it's a thoughtful comment. Maybe it's an introduction. Maybe it's sharing some research. Maybe it's inviting them to something. Maybe it's doing nothing yet. The important thing is that there is now a reason for the interaction to exist.

This creates another problem. What if you're entering healthcare and know nobody in healthcare? The traditional answer is: buy a database, find 5,000 healthcare executives, start emailing.

I think there may be a better answer. Don't find 5,000 prospects. Find 20 nodes. Find the consultants, fractional executives, investors, operators, community leaders and adjacent vendors who already have trust inside that market.

1 founder -> 5,000 strangers

1 founder -> 20 trusted people
          -> their networks

You're borrowing trust rather than trying to manufacture it inside an inbox.

There's an even bigger inversion. Instead of asking "how do I get the CRO to respond to me?", ask: what could I create that would make the CRO want to find me?

A benchmark. A proprietary dataset. A diagnostic. A calculator. A research report. A private peer group. A teardown. A tiny free product. Something genuinely useful.

Imagine publishing: "We analyzed 10,000 enterprise opportunities and found the seven signals most correlated with deals stalling." Now the conversation starts differently. You're not asking someone for 30 minutes. You're giving them something worth 30 minutes. That asset becomes your SDR.

Founders hear constantly that they need to "build an audience." I'm not sure. I'd rather have 50 CROs who know who I am than 50,000 random followers.

Create a 50-person CRO group. No selling. Ask everyone to anonymously contribute the biggest revenue problem they're dealing with. Analyze the problems. Give the intelligence back to everyone. Repeat monthly.

After six months you haven't created a lead database. You've created something much more valuable: a network. And networks are difficult for AI to spam into existence.

That's the irony. AI is making communication essentially free. Which means communication itself becomes less valuable. When everyone can generate a beautifully researched, perfectly written personalized email in three seconds, receiving one tells me almost nothing.

But someone I trust saying "you should talk to this person" means a lot. A useful piece of original research means something. A thoughtful conversation means something. Someone showing up consistently in a community means something. A founder who has spent six months genuinely helping 50 people means something.

AI makes content abundant. Trust remains scarce.

So perhaps cold outbound isn't the problem. Maybe cold outbound is simply experiencing the inevitable economics of abundance. When sending messages was expensive, people sent fewer messages. When sending became cheap, volume increased. AI makes sending nearly free. So supply can become effectively infinite. But human attention remains stubbornly finite. That equation cannot balance.

Which makes me wonder whether we're entering the end of the traditional SDR era. Not because AI SDRs replace human SDRs. But because AI SDRs destroy the channel SDRs depended on.

The next generation of sales technology may therefore look nothing like today's AI SDR. It may not optimize "how many people did we contact?" It may optimize "how many legitimate reasons to engage did we discover?"

And the winning GTM motion may shift from:

Find -> Message -> Follow up

to:

Earn -> Observe -> Enter

Earn attention before someone is buying. Observe when circumstances change. Enter when you actually have a reason to be there.

Because the world doesn't need another agent capable of sending 10,000 personalized emails. We already solved sending. The problem now is getting someone to care.

If you run outbound today, or sit on the receiving end of it, I'd like to know: when was the last time a cold message earned a reply from you, and what did it have that the other hundred didn't?

The two founder stories at the top are from private conversations and are told with their details rounded. Reply-rate benchmarks come from vendors who sell outbound tooling and measure their own traffic, so they are directional, not universal; Belkins counts replies per email sent while Instantly's figure reflects its own customer base, which is why the two differ so much.