Inbound is drying up for a lot of companies, and they all know why.
Buyers used to type a question into Google, scan ten links, and click one. Now Google writes the answer at the top of the page. In a Pew Research Center study of 68,879 searches, people clicked a regular result on 8 percent of visits when an AI summary appeared, against 15 percent when it didn’t. Only 1 percent clicked a link inside the summary.
Fewer clicks means fewer visitors, and fewer visitors means fewer leads. So sales teams look for another way to fill the pipeline, and the obvious answer is sitting in their inbox: an AI SDR.
What is an AI SDR?
An AI SDR (sales development representative) is software that finds prospects, writes emails to them, and sends follow-ups, with little or no human help. You will also see it called an AI sales agent or an AI BDR (business development representative). The pitch is simple. A thousand personalized emails a day, for less than the cost of one junior salesperson.
I run AI outbound every week. Some of it works well. A lot of what gets sold as AI outbound is spam with better grammar.
What AI SDRs do well
Let me start with the good news, because there is plenty of it.
AI is excellent at research. Give it a company name and it will read the website, the news, the job posts, and the annual report, and tell you in a paragraph what the company is working on. That used to take a salesperson 20 minutes per account. Now it takes seconds.
AI is good at first drafts. Point it at that research and it writes a reasonable opening email. Not a great one, but a good starting point for a person who knows the buyer.
AI is good at the boring parts. Cleaning a list, finding the right contact, logging replies in the CRM, reminding someone to follow up. Nobody misses doing that by hand.
Used this way, AI makes a good salesperson faster. The trouble starts when it replaces the salesperson.
Three ways AI outbound goes wrong
1. Volume burns your domain
Because sending is cheap, the tools encourage you to send a lot. Gmail and Yahoo have spent the last two years making that expensive.
Google’s sender guidelines tell bulk senders to keep their spam complaint rate below 0.1 percent and never reach 0.3 percent. That is three complaints per thousand emails. Send a thousand cold emails a day to people who never asked to hear from you, and you can cross that line in an afternoon.
Cross it, and your deliverability collapses: your emails start landing in spam. Not only the cold ones. Your invoices, your proposals, and your replies to customers all come from the same domain.
2. Fake personalization fools nobody
Most AI personalization is a template with a variable or two dropped in. “I loved your recent post about supply chains, {first name}.” Every buyer has seen it a hundred times, and it’s the same AI-written filler that has taken over LinkedIn.
Spam filters have too. They look for the same message sent to many people with a few words changed, and that is exactly what template personalization produces.
Buyers react even worse than the filters. In a 2025 Adience survey of 350 business buyers, a quarter said they were put off by outreach that was obviously written by AI. The same survey found buyers were more annoyed by poor discovery questions and irrelevant decks. The problem isn’t AI. It’s lazy selling, and AI makes lazy selling cheap.
3. The wrong list, faster
An AI SDR sends to whoever you point it at. Most bought lists are out of date, too broad, or both. The tool doesn’t know that the head of procurement left last spring, or that the company stopped importing from your region two years ago.
A bad list used to waste a salesperson’s week. Now it wastes 10,000 emails and your domain’s reputation along with them.
What works instead
The companies getting results from AI outbound do four things differently.
They send less. A list of 200 accounts that fit beats a list of 20,000 names. AI makes it cheap to send more. The win is using it to send less, to the right people.
They find a real reason to write. A new market entry, a trade show, a tender, a product launch, a new regulation. Something true and specific about the buyer that makes the email worth reading. AI is very good at finding these reasons. It’s bad at making them up, so don’t let it.
They keep a person in the loop. AI researches and drafts. A salesperson reads every message before it goes out and handles every reply. That one step catches the factual mistakes, the awkward phrasing, and the emails that should never have been sent.
They measure replies, not sends. Emails sent is a vanity number. Positive replies and meetings booked are the numbers that matter. If a campaign gets complaints or silence, stop it and find out why before you send more.
A checklist before you turn on an AI SDR
Run through these before a single AI-written email goes out:
- Separate your domains. Send outbound from a domain that isn’t the one your customers and invoices use.
- Set up authentication. SPF, DKIM, and DMARC on every sending domain. Gmail requires them for bulk senders.
- Warm up slowly. Email warm-up means new inboxes start with a handful of emails a day and grow over weeks, not hours.
- Watch your complaint rate. Set up Google Postmaster Tools and check it weekly. Stop at 0.1 percent, not 0.3.
- Build the list yourself. Start from the accounts you want to win, not from a list someone sold you.
- Write the reason first. If you can’t say in one sentence why this buyer should hear from you this week, don’t write.
- Read every message. A person approves each email before it sends. No exceptions in the first month.
- Follow the law where you send. CAN-SPAM in the US, CASL in Canada, and GDPR in Europe all apply to cold email. Canada and Europe are stricter.
- Track replies and meetings. Not opens, not sends.
Outbound and AI search go together
Outbound only fixes half the problem. When a buyer gets your email, the first thing many of them do is ask ChatGPT or Google about you. If the AI answer names three competitors and not you, your email just did their marketing.
So the work comes in a pair. Get named when buyers ask AI, which is what an AI search visibility audit covers. And reach buyers directly, without spamming them, which is what the AI sales audit is for. In two weeks I check where your pipeline comes from, fix your sending setup, and build one outbound workflow on your own data that your team runs.
AI won’t sell for you. Used well, it gives your best salespeople more time with the buyers who matter.
Common questions
What does an AI SDR do?
An AI SDR (AI sales development representative) is software that finds prospects, researches them, writes outreach emails, and sends follow-ups with little or no human help. It is also called an AI sales agent or AI BDR.
Do AI SDRs work?
They work well for research and first drafts. They work badly when they send high volumes of templated email with no person checking it. That approach raises spam complaints, gets domains filtered, and puts buyers off.
Can an AI SDR get my domain blocked?
Yes. Google tells bulk senders to keep spam complaints below 0.1 percent and never reach 0.3 percent. A high-volume cold campaign can cross that quickly, and then all email from the domain, including invoices and customer replies, starts landing in spam. Send outbound from a separate domain.
Will an AI SDR replace human SDRs?
Not the good ones. AI takes over the research, list cleaning, first drafts, and CRM logging. A person still decides who to contact, approves each message, and handles the replies, which is where deals start.
How many cold emails can I send per day without hurting deliverability?
There is no safe number for everyone. New inboxes should start with a handful of emails a day and build up over several weeks. Watch your complaint rate in Google Postmaster Tools and slow down well before 0.1 percent.