The AI SDR pitch in 2024 was beautiful. Infinite personalization at near-zero marginal cost. Every prospect gets a custom email referencing their last LinkedIn post, their company's latest product launch, and a relevant data point from a public source. Volume scales. Quality stays high. Reps stop spending time on research and start spending time on actual selling.

Two years in, the verdict from the buyer side is brutal. Reply rates fell, brand damage went up, and the senior buyers your team most wants to reach have effectively opted out of inbound entirely.

This post is about what actually happened, why the AI SDR category misread the constraint, and what the teams who are still hitting quota in 2026 are doing instead.

From the trenches

I will give you the honest math from when Boomerang itself was running scaled outbound infrastructure. The same kind of motion most teams are running today.

Our scale: about 5,000 cold emails per month, signal-based personalization, the full AI tooling stack. Result: roughly 2 meetings per month. That is 1,000 emails per meeting at the headline ratio. We could argue our targeting was slightly better than the category average. It was not meaningfully better.

The infrastructure to maintain this was a nightmare. Multiple sending domains, warm-up rotations, deliverability monitoring, constant tuning. The team time to keep the machine running was real. And the actual yield was 2 meetings.

Even if we had doubled conversion to 10 meetings per 5,000 emails, we were burning through our addressable market at a rate that did not justify the brand cost. Every senior buyer we hit added to the universe of people who associated us with cold spam. We stopped completely.

This is not a moral position. The math did not work and the brand damage was real. The companies still running this motion in 2026 are mostly doing it because the infrastructure exists and nobody has done the audit. The audit is the thing that ends the contract.

What the AI SDR vendors got right

Before tearing the category apart, I want to be honest about what was real.

The cost-per-touch math was correct. AI tools genuinely did reduce the marginal cost of a "personalized" outbound touch from about 15 dollars (rep time, list pull, light research) to roughly 10 cents (LLM inference plus a thin orchestration layer). That is a 150x cost reduction. Every operator's instinct says you should use that.

The personalization quality, on a single message basis, was real too. A 2024-era AI SDR could produce a cold email that, in isolation, was indistinguishable from a well-researched human-written note. It would reference the prospect's role, a recent company event, and a plausible reason for the outreach. If you read one in a vacuum, you would say "that is a good email."

Both of these things were true. The category was not built on a lie. It was built on a misread of where the constraint actually was.

What the AI SDR vendors got wrong

The constraint was never personalization quality. It was the buyer's tolerance for outbound as a category.

When the cost of "personalized" cold dropped 150x, the volume of "personalized" cold sent to every senior buyer rose by a similar factor. The buyer's inbox stopped being a place where the occasional well-crafted note from a curious seller appeared. It became a firehose of indistinguishable AI-generated openers, each one referencing the buyer's last quarter, each one suggesting "15 minutes" to "explore synergies," each one feeling personal in isolation and identical in aggregate.

The buyer adapted. The new defensive reflex looks like this:

• Skim subject line, look for cold flags (templated language, "quick question," follow-up to a non-existent email)
• If detected, delete without opening
• Mark sender as spam if the pattern repeats
• Eventually, install filters that route anything with cold characteristics directly to a folder nobody reads

This is not a hypothesis. This is what the average B2B buyer's inbox actually looks like in 2026. Average cold email reply rate in B2B has fallen from 3 to 5 percent in 2020 to below 1 percent today. Among C-suite recipients at companies over 1,000 employees, the reply rate is essentially statistical noise. Cold to a CFO at a public company has a reply rate that is functionally indistinguishable from zero.

The AI SDR vendors solved the wrong problem. They made personalization scalable. The market needed less outbound, not more. By solving a problem the market did not have, they accelerated the destruction of the channel they were operating in.

The brand damage angle nobody talks about

There is a second-order effect that does not show up in any AI SDR vendor's case study, and which most teams are slow to admit.

Every AI-generated cold email that gets sent on your behalf carries your brand. The sender field says your company name. The signature has your domain. When the buyer marks it as spam, they are marking your company as spam. When the buyer rolls their eyes at "I was looking at your LinkedIn profile and noticed your role in scaling demand generation at [Company]," they are rolling their eyes at your company.

For most B2B companies, the AI SDR motion sends 20 to 50 times more outbound touches per month than the previous human SDR motion. The total brand surface area exposed to potential negative impressions went up by that same factor. The conversion rate per touch went down by 70 to 90 percent. Net effect: significantly more brand exposure to significantly less qualified attention, with significantly more negative sentiment per impression.

I have talked to CMOs in 2026 who can show measurable declines in brand search volume that correlate with their AI SDR rollouts. The mechanism is straightforward. Buyers who would have arrived at the brand through organic discovery, content, or a referral now arrive with a pre-formed impression based on the cold notes they have been receiving. That pre-formed impression is "another company doing AI spam." The brand never gets a clean first impression.

This is the part of the AI SDR story that the dashboards do not show. The reply rate dashboard shows you something is broken. It does not show you the brand erosion that is happening in parallel and that will be much harder to reverse.

Why "better personalization" does not fix this

The vendor response to declining reply rates has been predictable. More signal sources for personalization. Job change data. 10-K filings. Conference attendee lists. The pitch is: if the personalization is sharper, the buyer will respond.

The buyer's response to sharper personalization is identical to the buyer's response to weaker personalization. Delete.

The buyer is not making a quality assessment of the email. The buyer is making a category assessment of the channel. Cold inbound is now classified, in most senior buyers' minds, as "spam, regardless of how good the individual message is." No amount of additional signal in the email body changes that classification. The classification was triggered by the volume, not the quality.

The category cannot personalize its way out of this hole. The hole is structural, and was created by the category itself.

What is still working

In the same period that AI SDRs were destroying the cold channel, three other motions were quietly outperforming.

Warm intros from customers. Conversion to meeting in the 40 to 60 percent range. Brand-positive (the customer doing the intro elevates your brand rather than degrading it). Scales linearly with customer base size and customer success investment.

Board and investor introductions. Lower volume but extraordinarily high quality. Conversion to meeting often above 70 percent. Each board member can credibly make 5 to 8 high-quality introductions per year, which sounds small until you do the math on what 8 high-conversion meetings produces in pipeline for an enterprise sales motion.

Employee alumni networks. Most B2B companies have employees who used to work at companies in their target market. Those former-coworker introductions convert at 30 to 50 percent. The data is sitting in your team's LinkedIn graph. Almost no team mines it systematically.

These three channels share three properties that the AI SDR motion specifically lacks. They are positive-brand (the channel itself enhances perception of your company). They are conversion-rich (the math works at a fraction of the volume cold needs). They are durable (they are not getting worse over time, unlike cold which is structurally degrading).

The teams hitting quota in 2026 are running a motion where these three channels carry 40 to 60 percent of pipeline. Cold still has a place for very high-volume, low-ACV segments. For anything over 30k in ACV, the math has shifted permanently toward warm-first.

What to replace AI SDR with

If you are reading this and you have an active AI SDR contract that is not producing pipeline, the answer is not to switch to a different AI SDR vendor. The answer is to rebuild the motion around the channels that actually work.

The four super-connector groups that produce warm pipeline at scale are your team (employees and their alumni networks), your customers (and customers' former colleagues at prospect accounts), your investors and board, and your advisors. Most B2B companies have all four groups but treat them as informal, ad-hoc resources. The shift that pays off is to make them an operational, instrumented channel with weekly cadence, measurable conversion rates, and an actual ops layer behind them.

This is exactly what Boomerang does. We map the relationships across all four super-connector groups, surface the warmest path to each account in your pipeline, draft the intro request in the connector's voice, and close the loop when the meeting books. It is the operational layer for the channel that actually works in 2026, instead of the channel that worked in 2022 and was killed by AI.

You can build this internally. Some teams do. The minimum viable version is a shared spreadsheet of every employee's prior companies and current LinkedIn graph, a quarterly board-intro batch process, and a CSM-led customer reference program with structured ask cadences. If you build this carefully and assign ownership, it will produce more pipeline than your current AI SDR.

The tooling layer exists when the manual version stops scaling. The point is that the channel is the answer, not the tool.

What to do this quarter

Two moves that change the trajectory.

One. Audit the actual contribution of your AI SDR investment. Pull the last six months of "AI SDR-sourced" meetings. Look at conversion to opportunity. Look at conversion to closed-won. Compare to warm-sourced pipeline. Most teams discover the AI SDR motion is producing meetings that close at 30 to 50 percent of the rate of warm-sourced meetings, at significantly higher brand cost. The contract probably does not survive that audit.

Two. Stand up a 90-day pilot of warm-led pipeline. Pick three super-connector groups (team, customers, board is a good starting trio). Assign ownership. Set a weekly cadence. Measure conversion at every step. After 90 days you will have hard numbers to compare to AI SDR and a decision becomes obvious.

If you want a starting point for the warm-led motion, our warm introduction software page covers the architecture in detail. For the executive-altitude version of this argument (why senior buyers specifically stopped opening cold), see our path to power pillar.

The AI SDR category will continue to exist. Some teams will keep using it for very specific use cases where the math still works. For most B2B companies pursuing deals over 30k ACV, the category is in the same position intent data was two years ago. Vendors are still selling it, dashboards still light up, the pipeline contribution is shrinking and the brand cost is rising. The teams that pivot now to warm-led motions are buying themselves 18 months of competitive advantage before the rest of the market catches up.

Shankar Ganapathy is the co-founder of Boomerang, the operational layer for relationship-led pipeline. Before founding Boomerang, he led product in the account planning signals space.