Glossary · GTM roles and motions
What is an AI SDR?
The short answer
An AI SDR is software that does the first steps of a sales development rep's job: it builds a list of prospects, researches each account, writes and sends the first messages, follows up and hands interested replies to a person. What it knows about each company and contact decides what it can write, so its results depend on its data.
Real data
The SDR job an AI SDR takes on, in US job postings
US job postings from the 30 days through November 10, 2026 whose whole title is a standard SDR or BDR form, such as "Sales Development Representative" or "BDR". What their descriptions ask for is the work an AI SDR is built to automate.
Returned by the BlitzAPI API on
/v2/jobs/searchUS postings for human SDRs and BDRs, last 30 days
Three exact forms per title, such as "Sales Development Representative", "Senior Sales Development Representative" and "SDR". One company can post several roles.
/v2/jobs/searchWhat the 2,222 descriptions ask for
One request per task. A posting counts when its description holds one of the words in parentheses. The meeting row also counts phrases such as "meeting quota", so it is an upper bound.
/v2/jobs/searchAI and workflow tools named in the 2,222 postings
A posting counts when its description names the tool, including postings from the companies that make it.
How does an AI SDR work?
An AI SDR works through the same steps as a human SDR, in the same order, for every account on its list. What changes is who does each step: software reads the data, drafts the message and sends it, and a person steps in when a buyer replies.
- Step 1Build the listAccounts and people that fit the target profile
- Step 2Research each accountThe company record, recent hires, open jobs
- Step 3Write the first messageWith a reason to write that belongs to that account
- Step 4Send and follow upEmail first, sometimes calls or social messages
- Step 5Hand off the repliesA person takes over once a buyer engages
Replies, bounces and booked meetings feed the next list.
Most products sold as AI SDRs combine a data source for steps 1 and 2, a language model for step 3 and a sending tool for step 4. Some stop at a draft that a rep approves; others send on their own within limits the team sets.
AI SDR vs human SDR: what is the difference?
An AI SDR and a human SDR do the same work at a different scale and with different blind spots. Software does not get tired of research and can write a first draft for every account in a segment; a person notices when something in the data looks wrong.
| Human SDR | AI SDR | |
|---|---|---|
| Limit | Hours in the day | The data it can reach and the sending rules |
| Research | Reads widely, judges what matters | Reads what its data sources return |
| Mistakes | One at a time | The same mistake on every account until a rule stops it |
| Best at | Conversations, judgment, follow-up after a reply | Volume, consistency, the first touch |
The two often work together: the AI SDR prepares and sends the first touches, and the human SDR or the account executive takes the conversation once a buyer replies.
Are AI SDRs replacing human SDRs?
AI SDRs have not replaced human SDRs: companies still hire them, with 2,222 SDR and BDR postings located in the US in the 30 days through November 10, 2026. Their descriptions name the tasks AI SDR products automate: the word meeting appears in up to 1,419 of them, qualifying in 1,206, research in 1,013 and personalization in 350. A single 30-day count cannot show whether that hiring is rising or falling.
A smaller group of postings names AI or workflow tools: 60 name Clay, 45 Claude and 32 ChatGPT, including postings from the companies that make them. Building and running AI SDR software is part of what a GTM engineer does.
What does an AI SDR need to work well?
An AI SDR needs three things to work well: the right data, clear rules and a person who checks its output. The data matters most, because the software cannot tell a good record from a stale one:
- Accounts that fit, from a defined target profile, so that the volume goes to companies that can buy.
- Current roles, because people change jobs and a message to someone who left reaches no one.
- Verified addresses, because a guessed email that bounces hurts the reputation of the sending domain, as the email bounce page explains.
- A reason to write, such as a new hire, an open role or an expansion, that belongs to the account and not to a template.
Show the answer key
- The language model that wrote the messageThe message was fine; it went to a mailbox that no longer existed.
- The contact dataThe AI SDR worked from a role and an address that were out of date. A current role and a verified address would have stopped the send.
- The sending scheduleTiming changes reply rates, not whether a closed mailbox accepts mail.
- The target profileThe account may fit perfectly; the person in the record had moved on.
Where does an AI SDR get its data?
An AI SDR gets its data from the APIs it is connected to, and BlitzAPI provides the B2B data side of that stack. People Search finds the people who match a role across many accounts, Profile Enrichment returns the current role and history of a contact, and Find Work Email returns a verified work address or found: false, in which case nothing is counted and there is nothing to send.
Job postings supply the reason to write, and each of these endpoints is also a tool of the BlitzAPI MCP server, so an agent can call it directly. The GTM agents guide shows how an outbound agency assembles these calls into agents it controls step by step.
Sources and method
- Written by
- Antoine Blitz, Founder, BlitzAPI
- Published
- API calls run on
The calls ran on November 10, 2026 through the BlitzAPI MCP server, which calls the same endpoints; they are shown here as the equivalent REST requests.
A title search for "AI SDR" returns mostly human SDR roles at companies that build AI products, so the evidence reads the human job, whose tasks are the ones AI SDR products automate.
The counts are job postings located in the US, not companies or people: one company can post several roles. Words are matched whole and case is ignored, so "meeting" does not count "meetings" unless both are listed, and a common word can match another sense of it.