What does an inbound SDR do (and what AI now handles)?

An inbound SDR (sales development representative) answers the people who raise their hand: demo requests, form fills, chat handoffs and trial signups. They reply fast, check whether the person fits, and book the right ones onto an account executive's calendar.

Nearly every verb in that job description is about speed or sorting. That tells you what the role was built for, and why so much of it is moving to AI agents. Agents now take the first touch, and the human role moves to the conversations worth a person. I call the point where that handoff happens the worth-a-person line, and this piece is about where to draw it.

What does an inbound SDR do all day?

The work splits into six tasks.

1. Work the inbound queue. Demo requests, contact forms, chat transcripts passed over by a bot, trial signups and, at some companies, content downloads.
2. Respond and research. Look up the company and the person, then send the first reply or make the first call.
3. Qualify. Find out whether the account fits the ideal customer profile, what problem they have, and when they need it solved. (These lead qualification questions are a common starting set.)
4. Route and book. Send the lead to the right AE by territory, segment or product, and get a meeting on the calendar. (Our guide to lead routing covers the rules most teams use.)
5. Log it. Update the CRM so the AE walks into the first call knowing what was said.
6. Chase. Follow up with the people who filled a form and went quiet, and rebook the no-shows. (See lead follow-up timing and cadence.)

The difference from an outbound SDR is direction. Outbound SDRs create interest in people who never asked. Inbound SDRs handle interest that already exists, and their tool stack looks different too (see the SDR tool stack, job by job). That sounds easier, but the buyer is already evaluating, and every hour of delay is an hour on someone else's website.

What metrics is an inbound SDR measured on?

Five numbers show up on most inbound SDR scorecards:

Two of the five measure speed, three measure sorting. None measures whether the buyer got their question answered. For the first metric, see speed to lead and how to improve it.

Why was the inbound SDR role built around speed and triage?

The usual answer is that SDRs protect AE time: someone cheaper filters leads so expensive people only see good ones. But if filtering were the core of the job, most SDR hours would go into deep discovery. In practice, they go into replying, chasing and scheduling.

My theory: the role was built around a website that couldn't talk. A buyer with a question could find the answer on a page, or fill in a form and wait for a person. The SDR was the person on the other end, a patch for that wait.

The Storylane Buyability Index shows how big that wait is. It checked 10,580 buyer questions against 529 B2B websites, and 41.4% were not fully answered: half answered, gated, sent to sales or not found at all. Pricing questions were fully answered only 46.4% of the time. (The questions are generated per company, as what a buyer would plausibly ask, not logged from real buyers, and the sites were chosen by the people who ran the scans.)

Each unanswered question becomes a form fill in the SDR queue or a buyer who leaves. So the job is not "qualify leads". It is "answer the questions the website couldn't, before the buyer gives up".

What does AI now handle in the inbound SDR role?

Once the website can hold a conversation, the wait disappears, and so does most of the speed job. A buyer asks about pricing at 11pm and gets an answer at 11pm, with no lead left waiting.

Timing matters most. In Storylane's Modern Buying Trends 2026 report, 51.4% of 8,211 conversations with website agents across 19 companies started outside working hours, in the buyer's own local time, and those conversations run just as long as the ones inside working hours. No team of inbound SDRs covers that schedule.

The triage job moves too. A good website agent can do the sorting:

• Answer the product, pricing and fit questions from the company's own material.
• Ask discovery questions in the conversation: the current setup, the problem, the timeline.
• Qualify against the ideal customer profile, written in plain English rather than form fields.
• Book the meeting with the right rep, using the same routing rules the SDR used (here is how to book meetings from your website without a live SDR).
• Write the summary into the CRM so the AE has the context. On a HubSpot site, that includes routing to the right owner and lifecycle stage.

We have a name for the lead that comes out of this: an agent qualified lead. If you are comparing tools for this layer, our roundup of the best inbound AI SDR tools sorts them by job.

What is left for the human inbound SDR?

Most "AI replaces SDRs" pieces skip this part, and the data surprised me.

Buyers sort their own questions. The same report compared 26,382 questions buyers asked website agents across 19 companies with 5,687 prospect questions from 402 human discovery calls. Pricing opened 9.4% of agent conversations and 2.5% of sales calls. Integrations and rollout made up 16.5% of questions to the agent and 29.7% of questions on human calls. Buyers take quick, low-stakes questions to the agent, where asking costs nothing, and questions that need a judgement about their situation to a person.

And they rarely ask for a person just to have one. Across all conversations in the report, 3.1% of the questions buyers put to website agents were a request to speak to a human. That share rises as the conversation deepens: 1.6% of single-message conversations ended in a request for a human, against 25.4% of conversations with 11 or more messages. The buyers who want a person are the ones who have already done the reading.

So the worth-a-person line is not "complex leads go to humans". It is closer to this: a conversation is worth a person when the buyer needs a judgement about their own situation, or when the deal needs people who were never in the conversation. The second part is bigger: the same report found that nine in ten of the buying committee never join a sales call. The person chatting on your website at night is usually one of many (more in what 637 B2B deals taught us about buying committees).

That gives the human inbound SDR four jobs worth doing:

1. Take the judgement conversations. "Will this work with our setup?" and "How long will this take us?" are where a person adds the most. The SDR picks these up from the agent with the full conversation in hand, so the first human touch starts at question five, not question one. A clean handoff from conversation to sales is what makes this work.
2. Build the account. One visitor asked about pricing. Who else at that company needs to see this? The SDR finds the other people in the committee and gives the champion something to share.
3. Work the conversation record. Every agent conversation is a written record of what a buyer wanted to know. An SDR who reads them each morning knows which accounts are warm and what they care about before the AE does.
4. Teach the agent. The person who has handled the most objections is the best person to write how the agent should answer them.

The strongest objection is that buyers want a human and an AI first touch feels cold. The data points the other way: buyers asked for a human in 3.1% of their questions, and those who ask get one sooner, because the SDR is no longer buried in forms. For the cost and quality tradeoffs, see AI SDRs and human SDRs.

How should you redesign the inbound SDR role?

Start with what your inbound actually contains. Pull the last 90 days of demo requests, chat transcripts and contact forms, and sort each one into two piles: needed speed (an answer, a price, a booking) or needed judgement (a specific setup, a multi-team rollout, a procurement process). The first pile is the agent's. The second is the SDR's.

Then change the scorecard so it rewards the new job:

Finally, give the SDR ownership of what the agent says. They hear the objections. They should write the answers.

Where Storylane fits

We build a website agent, Storylane RepX, so I am biased about where the first touch should happen.

RepX is a website agent (an AI SDR for inbound) for teams whose buyers arrive with questions about the product. It is trained on your website, docs, decks and sales call recordings, and it answers in text, voice or video. It pulls up interactive demos in the conversation. It runs discovery, qualifies on your ICP written in plain English, handles objections the way your best reps do, and books meetings. Playbooks set different qualification questions, email capture rules and meeting links per page, guardrails keep pricing and competitor answers within your rules, and you can test the agent before anything goes live. Qualified leads and conversation summaries flow to HubSpot, Salesforce and Slack. In HubSpot, each conversation is logged as a note with the full transcript, a summary and the topics covered, so the SDR picking up a conversation sees everything the buyer already said. Hospitable used RepX to take over its SMB motion: 36 meetings booked and 7 closed-won deals in the first month, with a simple property-count question routing bigger accounts straight to sales. RepX starts at $2,000 a month, and the price depends on website traffic.

The one thing to take into Monday

The inbound SDR role was a patch for a website that couldn't answer questions. Once the website can, the patch comes off, and what remains is the part of the job that always mattered most: judgement, and the people who were never in the room. Draw your worth-a-person line on purpose, before your buyers draw it for you.

If you want to see how an agent takes the first touch on your own site, book a demo of RepX.