Do AI SDRs Work? What the Research Says

An email opens by congratulating you on your funding round. It names your product, mentions the hire you announced last month, and asks whether a quick chat would be worth your while. Ten years ago an email like that would have impressed you, because somebody had to look all of it up. Today you get several a week, and none of them mean anything.

Their dashboard logs another personalised touch. You archive it after the first line.

Why do sales teams buy AI SDRs?

Sales teams buy AI SDRs because prospecting is slow, repetitive work that rarely pays off, and anyone who has sent the fortieth near-identical email of an afternoon understands the appeal. An AI SDR is software that does the job of a sales development representative: it finds prospects, writes and sends outbound email, and answers the replies.

The Bridge Group's 2025 survey of 351 B2B companies found 60% of SDRs at quota, the lowest share in the study's history, and median annual attrition of 40%. In the same Bridge Group survey, the median SDR logs 112 activities a day for 4.1 quality conversations. At roughly 27 activities per conversation, prospecting looks like a volume problem.

If most of your outreach is generic anyway, handing it to a machine that never tires is a rational move. Whether it works depends on what the best salespeople do differently.

What do top-performing salespeople do differently?

Top performers researched the buyer before the first message. In LinkedIn's 2022 State of Sales survey of almost 15,000 B2B buyers and sellers, 76% of top performers (sellers at 150% of quota or more) said they always research a prospect before reaching out, against 47% of other sellers. In a 2017 survey of 489 B2B sellers by the sales training firm RAIN Group, 31% rated one-to-one emails written by hand after researching the prospect as very or extremely effective, against 9% for emails personalised by auto-filled fields and 5% for bulk email.

Salespeople who adapt to each buyer perform better. In a meta-analysis of 268 studies covering 79,747 salespeople, Verbeke, Dietz and Verwaal (2011) found that knowing the product and the customer, and adapting how you sell to what you learn about the buyer, had the highest correlations with sales performance of the 18 factor groups they examined.

Better listeners sell more. In a meta-analysis of 16 studies with a combined sample of 3,780, Itani, Goad and Jaramillo (2019) found salespeople who scored higher on listening also had higher sales, mostly as the salespeople themselves reported it, with an average correlation of .38.

Top performers teach each buyer something new, tailored to that person. In a 2013 survey of more than 700 B2B purchases by the sales training firm RAIN Group, buyers rated the winning seller and the runner-up on 42 behaviours. The two largest gaps were educating the buyer with new ideas and collaborating with them, and "understood my needs" and "connected with me personally" were also in the top ten. In a CEB study of more than 6,000 B2B sales reps, Dixon and Adamson (2011) found close to 40% of top performers fit the profile the authors call the Challenger: reps who teach customers something new about their business and tailor the message to each stakeholder.

Buyers want help making sense of information, not more of it. In a 2025 Gartner survey of 645 B2B buyers, 69% said they prefer to validate AI-generated insights with a sales rep. In a 2019 Gartner survey of 1,174 B2B buyers, reported by Adamson (2022), 80% of buyers whose rep helped them make sense of the information completed a large, premium purchase they did not regret, against 50% whose rep mainly told them what to think and 30% whose rep mainly sent more material.

Most of this evidence predates generative AI. Apart from Gartner's 2025 survey of 645 buyers, every study in this section rests on data gathered before ChatGPT launched in November 2022, when research, adapting, listening and a tailored insight each cost a seller time. In early 2022, researching a buyer before the first message was going above and beyond, and in LinkedIn's survey top performers were far more likely than other sellers to say they always did it. Generative AI has since made research, a tailored opening line and a relevant insight in an email nearly free to produce, so on their own they no longer show a buyer that anyone spent time on them. Listening, adapting and helping a buyer make sense of things in a live conversation, and following through over months, still cost a person time.

Do AI SDRs work?

AI SDRs do not work as the sender of outbound email. Researching the buyer, adapting, listening and bringing a tailored insight worked for top salespeople because each one cost the seller time, and a buyer feels singled out only when someone spends scarce time on them. An AI SDR exists to take that time out. It can copy the written output of each habit: research an account, adjust its wording, answer a question and include an insight. It cannot copy the cost.

A signal only tells you something when it is too costly to fake. That is the core of market signalling, set out by Spence (1973), work for which Michael Spence shared the 2001 Nobel memorial prize in economics. A rep's attention shows you were chosen because the rep's week is finite, and every minute spent on you is a minute not spent on anyone else. An AI SDR spends the same nothing on everyone on its list, and a person approving its drafts adds seconds, not the time the AI SDR was bought to save.

Time spent on a buyer builds the relationship, and frequent contact does little. In a Journal of Marketing meta-analysis of 111 samples (a combined 38,077 respondents), Palmatier and colleagues (2006) found a seller's investment of time, effort and resources was one of the stronger builders of customer relationships (a correlation of .46), while how often the seller made contact did little (.16).

Attention that costs nothing proves nothing.

Why has personalised outreach stopped working?

Personalised outreach has stopped working as a signal because anyone can now produce it. A personalised email proves effort only while most personalised emails take effort, and the buyer never has to find out which ones were automated. Before generative AI, an email that named your funding round and your latest hire meant someone had looked them up; now any tool writes it in seconds, so it proves nothing, whoever sent it. Personalisation has become the entry fee, not the signal. The signal moves to time a rep visibly spends on one buyer: a conversation that adapts to them, help making sense of what they know, and follow-through over months. AI will help reps, but it also moves the bar up for everyone.

Is a human SDR better than an AI SDR?

A human SDR sending recycled email fares no better than an AI SDR. In a study with 600 people, Kirk and Givi (2025) found a sales director whose warm email was virtually identical to the one the previous sales director sent last year was rated far less authentic than one who wrote it fresh (3.12 against 5.63 out of 7), and no more authentic than a chatbot doing the same.

A human rep on call 52 of the day, saying what every other prospect hears, is generic too. If your outreach is a script, an AI SDR can run it as well as that rep can. Neither makes a buyer feel chosen.

How should SDR teams use AI?

SDR teams should use AI to find the buyers worth a rep's time, and let the rep spend the time it saves making each buyer feel singled out. The best salespeople will be the ones who make each buyer feel special and heard, and outreach will move towards fewer, better-aimed messages.

Top-performing reps choose accounts differently from other reps. In CEB research reported in Harvard Business Review, Adamson, Dixon and Toman (2012) found the top 20% by quota attainment favoured agile organisations in flux, after a leadership change, an acquisition or new regulation, and placed "more emphasis on a customer's potential to change than on its potential to buy". Spotting leadership changes, acquisitions and new regulation across thousands of accounts every week is the work AI does well.

AI already frees up seller time. In an early 2026 survey of 210 sales leaders, Gartner found AI saves sellers 4.8 hours a week, yet 72% of sales organisations put little of it back into high-value selling. In the same Gartner survey, sales organisations that saved moderate to large amounts of time and reinvested it were 3.1 times as likely to beat their lead-to-opportunity goals as those that reinvested less.

AI can gather account research, but the rep has to read it and decide why this person, now, because the rep is the one who will carry the conversation. On Monday: before any first message, the rep reads the account research and writes one line on what this account is dealing with and one thing worth telling them that they may not know. After each call or reply, the rep notes what the buyer said that changed the next step, and answers every real reply personally the same day.

Let AI choose who to call. Make them feel chosen.

Frequently asked questions