The short answer
Mostly no, and the human is what decides it. At $10,000 a year per customer, a plausible outbound programme produces roughly $96,000 of first-year revenue against a fully loaded sales development rep costing $110,000 to $127,000 before any account executive time. An AI SDR shrinks the software line. It does not shrink the line that decides the answer.
Why does the arithmetic fail under $10,000?
Work it in public. Ten meetings booked a month is 120 a year. Assume eight in ten are held, which gives 96.
At a 10% close rate that is 9.6 deals. On a $10,000 product that is $96,000 of first-year revenue.
The cost side, sourced: the Bridge Group's 2025 survey of 351 B2B companies puts median sales development rep base pay at $55,000 and median on-target earnings at $80,000. Applying the US Bureau of Labor Statistics employer-cost load of 1.33 to 1.46 gives roughly $106,000 to $117,000, and tooling takes it to about $110,000 to $127,000 fully loaded. Account executive time to run 96 meetings is real and additional, and no independent source publishes a figure for it.
So the programme is under water before the account executive is counted. Note also that the ten meetings a month is Bridge Group's median quota, recorded at 60% attainment — the lowest in that survey's history — and that the eight-in-ten held rate is an assumption, because no independent source publishes a no-show rate for booked B2B meetings.
An AI SDR reduces one of those two lines. Software cost is the line it shrinks. Human cost is the line that decides the answer.
The time required to hold 96 meetings does not shrink because a model wrote the emails. It shrinks only if you change who holds the meeting, and under $10,000 that usually means nobody holds it.
Does the buyer even want the call?
Often not. Nielsen Norman Group's B2B research, running 79 participants across 179 sites, found pricing ranked highest of all buyer priorities, 29% above product availability. Buyers at this deal size are trying to work out whether they can afford you without talking to anyone.
A sales conversation inserted into a self-serve purchase adds friction to a decision that was already going your way. That is the mechanism by which outbound at small deal sizes loses money twice.
Where does the same money produce a measured effect?
| Where the money goes | What the evidence found | Study and sample | How strong |
|---|---|---|---|
| Cold outbound, AI or human | Firm-initiated AI produced no measurable sales effect | Fang et al., a randomised trial of 44,614 consumers | Strong evidence of no effect |
| Responding to inbound with AI | A pre-sale chatbot raised sales 16.3% | Fang et al., 44,614 consumers | Moderate, one platform |
| Onboarding and activation | Proactive onboarding education halved first-week churn and raised cumulative eight-month usage 46.6% | Retana, Forman & Wu, M&SOM 2016, 366 treated of 2,673 | Strong, but human-delivered |
| Pricing clarity on the site | Pricing ranked highest in buyer priorities, 29 points above availability | Nielsen Norman Group B2B, 79 participants, 179 sites | Moderate |
| Personalising outbound harder | About 0.43 percentage points of extra response | Outreach personalisation study, 76,977 recipients | Real but very small |
| Proactive save attempts on at-risk customers | Raised churn from 6% to 10% | Ascarza, Iyengar & Schleicher, JMR 2016 | Strong evidence of harm |
| Ranking customers by profit-weighted response | At least 4% higher firm profit at identical spend | Lemmens & Gupta, Marketing Science 2020 | Strong |
Two rows of that table point the opposite way to what people expect. Personalising harder buys almost nothing. Reaching out to at-risk customers made things actively worse.
What should you do instead? Answer inbound properly.
The clearest split in the AI and sales evidence is by who started the conversation. Fang et al., pooling seven separate randomised experiments at one retail platform, found customer-initiated AI raised sales 16.3% among 44,614 consumers, while push messaging to 13.7 million produced no significant effect.
Outbound asks a stranger to care. Inbound answers somebody who already does.
At under $10,000 that is where the return sits. You have usually paid for the enquiry already, through ads or content, and then lost it because nothing answered while you were busy. The cost of answering does not scale with deal size, which is why it works where outbound does not.
What should you do instead? Fix the pricing page and the quote.
Nielsen Norman Group's B2B study of 79 participants across 179 sites put pricing at the top of buyer priorities, 29% clear of product availability. At small deal sizes the pricing page does the work an SDR would otherwise do badly.
Quote support has its own evidence. Karlinsky-Shichor and Netzer, writing in Marketing Science in 2024 on 17 sales representatives and 67,851 quotes, found a human-machine hybrid delivered a 7.8% profit improvement against 4.9% for full automation. The per-quote value was $14.85 for low-expertise representatives and $5.21 for high-expertise ones.
The machine advising a person beat the machine acting alone. That result has held up better than anything in the autonomous outbound category.
What should you do instead? Spend it on the first week of the customer relationship.
Retana, Forman and Wu, publishing in M&SOM in 2016 on 366 treated customers of 2,673, found that proactive onboarding education halved first-week churn, cut week-one support questions by 19.6%, and raised cumulative usage across eight months by 46.6%. The authors note the direct effect on activity decays within about a week, and the intervention was delivered by people rather than software.
At $10,000 contract value, halving early churn is worth more than any plausible increase in cold reply rate, and the effect compounds because the customers you keep renew.
Timing decides whether proactive contact helps or harms. Education at the start halved first-week churn. Intervention at the end, in Ascarza, Iyengar and Schleicher's 2016 JMR study, raised churn from 6% to 10%.
When is an AI SDR worth it under $10,000?
Three situations, and they are narrower than the marketing suggests.
When retention makes the real contract value much larger than the first-year number. A $9,000 product that renews for four years is a $36,000 decision, and it should be modelled that way.
When the AI is handling inbound response and routing rather than cold prospecting. That is the customer-initiated case, and it is the only side of that split with a positive randomised result behind it.
When expansion revenue is the actual business and the first sale is a foot in the door. Model the account, not the deal.
What does nobody know?
No randomised trial has tested an AI SDR at small deal sizes. No vendor in the category publishes cost per meeting held. The best available read on outcomes is Gartner's early-2026 survey of 210 chief sales officers, where 25% reported a return of 50% or better and 20% reported a negative return of 50% or worse.
The adoption data says something quieter. The US Census Bureau's BTOS work, published as CES-WP-26-25, found 18% of US firms using AI, with sales and marketing the most common function among adopters at 52%. Only 2% reported any labour reduction.
Almost everybody points AI at this function. Almost nobody removes a salary because of it.
If your deals are under $10,000 and you came here hoping to be told to buy one, the answer is do not. Spend the money on the people who already raised their hand, and on keeping the ones who signed.