Skip to content

Blog

AI SDR: What AI Sales Agents Do Well and Badly

An AI SDR is software that does parts of a sales development rep's job: finding prospects, researching them, writing and sending outreach, answering replies, and booking meetings. It is good at research and first drafts at a scale no person can match, and weak at judgement, at deciding when not to send, and at sounding like someone who knows your market.

This post explains what AI SDRs and AI sales agents do, where they help, where they fall short, the rules any tool sending on your behalf has to follow, and how to decide whether one fits. It's written for sales and marketing leads at B2B companies weighing a tool against people.

What an AI SDR is

A sales development rep, or SDR, starts conversations. They build a list of accounts, find the right people, write to them, follow up, handle the first replies, and book meetings for account executives. An AI SDR is a product that tries to do some or all of that with language models and automation.

What's sold under the name varies a lot. At one end are writing assistants that draft emails for a person to send. At the other are products that run outbound from list to meeting with little human involvement. Two examples, as their own sites describe them in November 2026:

  • Artisan describes Ava as an AI BDR that finds and enriches B2B leads, writes personalised outreach, sends it through its own sequencing, handles replies, and books meetings.
  • 11x describes Alice, for outbound, as researching accounts and running multi-step sequences across email, social, and calls, and Julian, for inbound, as calling back new leads and qualifying them by voice and chat around the clock.

We haven't tested either product, and we're not recommending one. They're here to show the range of what "AI SDR" covers. Check any vendor's claims against a trial on your own market.

What AI sales agents do well

Used for the right jobs, these tools save real time:

  • Research at scale. Reading a company's website, recent news, job ads, and product pages, then summarising why it might need you, is slow for a person and quick for a model. This is where AI does the most for outbound.
  • First drafts. A model given good research and a clear brief writes a reasonable first email in seconds. A person can then make it right far faster than writing from scratch.
  • Consistency. Every lead gets the follow-up the sequence says it should, on time, without anyone forgetting.
  • Speed on inbound. An automated first response can reach a new lead in seconds at any hour. The research on speed to lead statistics suggests that matters, as long as the reply is useful.
  • Sorting replies. Classifying replies as interested, not now, wrong person, or unsubscribe is a task models handle well, and it keeps opt-outs from being missed.

What they do badly

The same tools fail in predictable ways:

  • Judgement about whether to send. A person reading a prospect's site might decide the timing is wrong, the company is a poor fit, or the contact just left. A system optimised to send will usually send.
  • Specific knowledge of your market. A model writes plausibly about anything, which is the problem. Buyers in technical B2B markets spot generic claims quickly, and a confident wrong detail about their business does more harm than a plain email.
  • Getting facts right. Research summaries can include things that aren't true: an invented product line, an outdated role, a wrong location. Anything a model says about a prospect needs checking before it's sent under your name.
  • Volume as a strategy. If each email is cheap to send, it's tempting to send many more. Mailbox providers watch for exactly that, and a damaged sending domain hurts all your email, not just the campaign.
  • Live conversations. Handling an objection, a pricing question, or a detailed technical query is where buyers decide whether you're worth a meeting. Most teams keep a person in those conversations, and for good reason.

AI appointment setters

An AI appointment setter is the narrowest version: it answers replies or inbound leads and tries to book a meeting. It can work well for simple, high-volume requests where the next step is always the same, such as booking a demo.

It works less well when the right next step depends on the lead. A B2B buyer asking whether you can handle their volume, integrate with their system, or meet a deadline wants an answer from someone who knows. An appointment setter that books every lead regardless also fills your sales team's calendar with meetings that should have been an email. Our guide to lead qualification covers how to decide which leads deserve a call.

The rules any AI SDR has to follow

A tool sending on your behalf doesn't change the rules that apply to you. Three matter most, as of November 2026:

  • Email law. In the UK, the ICO's guidance on business-to-business marketing allows marketing emails to corporate subscribers without prior consent, but treats sole traders and some partnerships as individuals, and every email must say who sent it and offer a way to opt out. In the US, the FTC's CAN-SPAM compliance guide says the law makes no exception for business email, and requires accurate sender details, a postal address, and an opt-out honoured within 10 business days. You stay responsible for email sent for you.
  • Mailbox provider rules. Google's email sender guidelines require every sender to authenticate their email and keep spam complaint rates below 0.3%, and senders of more than 5,000 messages a day to Gmail to add one-click unsubscribe and DMARC. An AI SDR that sends high volumes from your domain has to meet these, or your email stops arriving.
  • LinkedIn's terms. LinkedIn's help page on prohibited software and extensions says tools that scrape the site or automate activity on it, such as sending messages or adding contacts, break its User Agreement and can get accounts restricted. A product that automates LinkedIn actions puts your reps' accounts at risk.

How to decide whether an AI SDR fits

Ask these before you buy:

  1. What exactly will it do without a person? Research, drafting, sending, replying, booking: get the list in writing.
  2. Who checks what it says about prospects before it goes out under your name?
  3. Which domain and mailboxes will it send from, and who watches their reputation?
  4. How does it handle opt-outs, and how quickly?
  5. What does it do on LinkedIn, and is that within LinkedIn's terms?
  6. How will you measure it? Meetings that turned into real opportunities, not emails sent or meetings booked.
  7. What happens to your data and your prospects' data, and where is it processed?

A trial on a small, well-defined segment, with a person reviewing every message, tells you more than any demo.

For a wider look at where AI agents fit in a business beyond sales, see our post on AI agents for business, with examples.

Where Webika fits

Our B2B lead generation service uses AI where it's strong and people where it isn't. AI researches each target account; we write the sequences from that research, check them, and run the sending on separate domains, never your main one. We don't automate actions on LinkedIn. We report meetings booked and reply rates as results, never as a promise, because they depend on your market and your offer. If you'd rather own the system, we can build an agent that researches accounts and drafts emails for your team to approve.

Book a free call to talk through whether AI, people, or both suit your outbound.

Talk it through first

30 minutes, booked online. Free, with no obligation.

Book a call