AI did not make buyers easier to reach. It made everyone cheaper to interrupt.

A modern AI sales agent can research a company, identify a plausible buying signal, compose a personalised email, schedule follow-ups, evaluate the response, and hand a qualified opportunity to a human salesperson.

Major CRM vendors now market agents that research leads, draft and send personalised outreach, and follow up with limited human supervision. Microsoft describes an autonomous sales qualification agent that can qualify leads at scale, while HubSpot offers a prospecting agent for automated research and outreach. This is no longer a speculative feature of the sales stack. It is becoming normal infrastructure.

For an individual company, the economics are compelling. Work that previously required hours of human research and writing can now be completed in seconds. The rational response is to contact more prospects.

But when every company follows the same logic, the collective result is not more relevance. It is more interruptions.

B2B outreach is becoming a tragedy of the commons.

The shared resource is not the inbox

In the traditional tragedy of the commons, individuals acting in their own interest gradually exhaust a resource shared by everyone.

In outbound sales, that shared resource is not email itself. Email is effectively unlimited. The scarce resources are buyer attention, channel trust, and sender reputation.

Every unsolicited message consumes a small amount of each. A useful message may repay that cost by helping the recipient discover a relevant solution. An irrelevant message makes a withdrawal without creating value in return.

No individual sender owns the recipient’s attention. No individual sender controls how mailbox providers respond to unwanted messages. Yet every sender has an incentive to consume a little more of both.

The damage initially looks manageable. Reply rates decline slightly. More messages land in spam. Buyers become less receptive to unfamiliar senders. Teams then respond by increasing volume.

That creates a cycle: lower production costs lead to more messages; more messages produce less attention per message; lower response rates encourage yet more volume; complaints and disengagement rise; mailbox providers tighten their filtering; and the channel becomes less effective for everyone.

I call this interruption inflation: when the supply of automated outreach grows faster than the supply of human attention, each message purchases less attention than the one before it.

The data shows a market under pressure

At Belkins, we see this market at considerable scale. For our 2025 cross-channel benchmark, Belkins and its research partners analysed 16.5 million cold emails, more than 20 million LinkedIn outreach attempts, and more than 5 million cold calls. The size of that dataset matters because the pressure is not limited to one campaign, industry, or channel.

A separate Belkins analysis of more than 7.5 million cold emails sent during 2025 used a strict methodology: replies were measured against all emails sent to net-new contacts, not only against recipients who opened an email.

The average reply rate was 0.45%. More importantly, the first half of 2025 averaged 0.50%, while the second half averaged 0.40%, a 20% decline within the same year. December reached a low of 0.35%.

This does not prove that AI caused the decline. Buyer behaviour, economic conditions, targeting quality, deliverability, seasonality, and messaging all affect performance. The newer study also uses a different denominator from previous Belkins research, so its absolute reply rate should not be directly compared with earlier reports.

What the data does show is that buyer response is not infinitely elastic. Increasing the supply of messages does not create a corresponding increase in buyer attention.

The infrastructure is also becoming less forgiving. Google advises bulk senders to keep their user-reported spam rate below 0.1% and prevent it from reaching 0.3% or higher. Rates above 0.1% can negatively affect inbox delivery, while rates of 0.3% or more have an even stronger effect.

Put differently, a spam rate of 0.1% represents roughly one complaint per 1,000 delivered messages. A rate of 0.3% represents three complaints per 1,000. That is not much room for error.

AI can generate thousands of convincing sentences, but a tiny number of recipients can still damage the sender’s ability to reach everyone else.

Personalisation is becoming a commodity

For years, the sales industry treated personalisation as the solution to poor outreach. The basic idea was reasonable: research the recipient, mention something specific, and demonstrate that the message was written for them.

AI is rapidly making surface-level personalisation a commodity. Almost any system can identify that a prospect recently raised funding, hired a senior executive, launched a product, appeared on a podcast, or published a LinkedIn post.

Once every sender has access to the same facts, mentioning those facts no longer proves that meaningful attention was invested. It merely proves that the sender has access to software.

Personalisation without selectivity is just better-targeted noise.

True relevance requires more than inserting personal details. The recipient must plausibly have the problem, the problem must matter now, the proposed solution must be credible, and the value of the message must justify the interruption.

AI can help evaluate those conditions. But most sales systems are still optimised around the wrong output. They measure how many personalised messages the machine can produce, rather than how many unnecessary messages it correctly decides not to send.

When everyone can manufacture the appearance of relevance, relevance stops being primarily a copywriting problem. It becomes a judgement problem.

Reputation is now part of the product

We learned this lesson before the current AI sales boom. In 2019, Folderly began as an internal solution after recurring deliverability problems began affecting outreach. The original purpose was to identify technical issues, improve inbox placement, and recover mailboxes that were being classified as spam.

The deeper lesson was not simply that companies needed another deliverability tool. It was that authentication, infrastructure, sender behaviour, reputation, targeting, and message quality form one connected system.

A perfectly written email creates no value when it does not reach the inbox. At the same time, no technical deliverability system can transform an unwanted message into a wanted one.

Reputation is partly technical, but it is also behavioural. Every campaign teaches mailbox providers and recipients how much trust to place in the next message.

That makes sender reputation a business asset rather than a technical afterthought.

The next sales stack needs a trust budget

The future of outbound is not zero outbound. Buyers still benefit from discovering solutions they did not know existed, and new companies still need ways to reach markets that are unfamiliar with them.

But the next generation of sales systems must treat attention and trust as finite resources.

Use AI to reduce uncertainty before increasing activity

The system should gather stronger evidence that a company has the right problem, timing, and buying conditions before initiating contact. Better research should produce fewer weak messages, not merely more polished ones.

Give every account a contact budget

A prospect who has already received several unanswered messages should not be treated as an untouched opportunity simply because another department, channel, or agent owns the next sequence.

Coordinate channels around the buyer

An email, LinkedIn message, phone call, advertisement, and automated follow-up are not separate interactions from the buyer’s perspective. They are all withdrawals from the same attention account.

Measure negative externalities

Positive replies and meetings matter, but so do spam complaints, unsubscribes, negative replies, domain health, inbox placement, and the number of contacts repeatedly approached without engagement.

Keep human judgement closest to the interruption

AI is excellent at increasing the amount of context available to a salesperson. It is less reliable when asked to decide, without supervision, whether a commercially plausible message deserves a stranger’s attention.

Restraint will become a competitive advantage

AI will not kill outbound sales. Undisciplined use of AI might damage it.

The companies that win will not necessarily be those that generate the largest number of personalised messages. They will be the ones that develop the clearest understanding of who should be contacted, why the timing is justified, what value the interaction creates, and when silence is the better decision.

The most valuable action taken by an AI sales agent may eventually be a non-action: do not send this message.

When interruption becomes cheap, restraint becomes a competitive advantage.

Get the TNW newsletter

Get the most important tech news in your inbox each week.

Expertise from selected TNW Council members, admitted through an application and review process to a fee-based program. Opinions expressed by the authors are their own.