Skip to content
Solutions · Self-Hosted AI SDR

Self-Hosted AI SDR Agent for Outbound and Follow-Up

AI SDR agent built on your CRM and run in your own cloud account or servers. It researches leads, drafts personalised outbound email, follows up, sorts replies and books meetings, with a rep approving what goes out. Prospect data, email threads and logs stay in your environment, and every message leaves from your own domain.

Book a Strategy Session Free 30-minute call
0.3%Spam-rate ceiling in Google's Gmail sender guidelines; Google advises staying below 0.1%
10Business days, at most, to honor an opt-out under CAN-SPAM (FTC) and CASL (CRTC)
5,000Daily messages to Gmail at which Google's bulk-sender rules apply, including one-click unsubscribe
What It Does

What an AI SDR Agent Does for Your Sales Team

An AI SDR, also called an AI BDR or AI sales development representative, handles the written prospecting between a new lead and a booked meeting.

Lead research

It reads the lead's website, your CRM history and your licensed data sources, then writes a short account brief with a reason to reach out.

AI outbound email

It drafts a first-touch email from that brief in your reps' voice, then queues it for a rep.

AI lead follow-up

Follow-ups go out on your cadence and stop as soon as a lead replies, books or opts out.

Reply triage

Replies are sorted by intent. Opt-outs are suppressed at once, and the rest reach a rep with a suggested answer.

AI lead qualification

Engaged leads are scored against your criteria, such as role and company size, with notes on the CRM record.

Meeting booking

As an AI appointment setter, it offers times from the rep's calendar, sends the invite and handles reschedules.

Built on Your CRM

An AI Sales Agent That Works Inside HubSpot, Salesforce or Pipedrive

Your CRM stays the system of record. The agent reads leads from it and writes every email, reply, score and meeting back.

HubSpot

HubSpot runs a remote MCP server at mcp.hubspot.com that lets an AI agent read, write and update CRM records, per HubSpot's developer docs. We scope the agent to the objects the SDR workflow needs.

Salesforce

Salesforce Hosted MCP Servers accept a custom agent over per-user OAuth 2.0 with PKCE, and tool calls respect field-level security and sharing rules, per Salesforce's docs. Separate read-only and create-and-update servers keep writes to what you allow.

Pipedrive

Pipedrive's API supports OAuth 2.0 apps with scopes chosen at registration, giving each app granular access, per its developer docs. The agent requests only the scopes the SDR workflow uses.

Calendars and data providers connect the same way through our AI integration services.

Why Self-Hosted

Why Sales Teams Choose a Self-Hosted AI SDR

Hosted AI SDR tools such as Artisan and AiSDR run in the vendor's cloud and sync with HubSpot and Salesforce, per their product pages. Teams choose a self-hosted agent for three reasons:

1 Prospect records and email threads kept in the company's own cloud account or servers.
2 Outreach from the company's own domain and mailboxes, at limits it sets.
3 Prompts, approval rules and logs that sales and compliance leads can read and change.
The Self-Hosted Answer

An AI SDR agent that runs inside your environment.

The agent, its lead store, prompts and logs run in your AWS, Azure or Google Cloud account or on your servers. In a fully self-hosted build, drafting runs on an open-weight model on your GPUs. Teams that prefer a frontier model can route drafting through their own provider contract instead.

Lead data and email threads in your cloud account
Sends from your own domain and mailboxes
Open-weight model on your GPUs, or your own provider contract
Human Approval Points

Where a Rep Approves the AI SDR's Work

The agent drafts, sorts and schedules. A person decides at five points, each one logged.

1

Target list sign-off

A sales lead approves which contacts enter outreach. The agent checks consent and suppression status in the CRM before drafting.

2

First-touch approval

New sequences start in a review queue. Reps approve or edit each draft, and their edits shape later drafts.

3

Send rules for follow-ups

For a proven template, the sales lead can let follow-ups run under set rules, such as daily caps per mailbox, and return them to review at any time.

4

Reply handoff

Interested replies, questions and objections go to the owning rep with a suggested answer to send, edit or replace.

5

Stage changes and audit log

The agent proposes stage changes with its qualification notes for the rep to confirm. Every draft, approval, send and opt-out is logged with its approver.

Deliverability and Anti-Spam

Consent, Opt-Out and Sender Identity Built Into Every Send

The agent applies the email rules your counsel sets, starting with CAN-SPAM (FTC), CASL (CRTC) and the sender requirements of mailbox providers.

Consent

The FTC's CAN-SPAM guide says the law applies to business-to-business email too. The CRTC's CASL FAQ lists three requirements: consent, identification information and an unsubscribe mechanism. The agent stores each contact's consent basis in the CRM and sequences only contacts that meet your counsel's rules.

Opt-out

CAN-SPAM requires opt-outs honored within 10 business days and a mechanism that works for at least 30 days after sending (FTC). CASL requires unsubscribes processed within 10 business days and a link valid for at least 60 days (CRTC). The agent adds an unsubscribe link to every message and suppresses removal requests across all sequences as soon as it reads them.

Sender identity

CAN-SPAM requires accurate headers, subject lines that reflect the content, a clear ad disclosure and a valid postal address (FTC). CASL requires senders to identify themselves and anyone they send for (CRTC). The agent sends under each rep's real name from your domain, with your postal address in the signature.

Deliverability

Google's Gmail sender guidelines ask every sender to use SPF or DKIM and keep spam rates below 0.3%. Above 5,000 messages a day to Gmail, senders need SPF, DKIM and DMARC plus one-click unsubscribe on marketing mail, and Google advises spam rates below 0.1%. The agent sends at volumes your mailboxes support, drops bounces and reports opt-out rates per sequence.

Your counsel decides how these rules apply to your lists.

Voice Follow-Up

Hand a Lead to a Voice Agent When a Call Fits Better

The AI SDR agent works in email. When a lead asks for a call, it passes the lead, its brief and the email thread to a voice agent.

Outbound and inbound calling run on our AI call center agent, and callbacks can run on a custom conversational voice agent. Both write to the same CRM record, so the rep sees one history.

Compared

SaaS AI SDR Tools Compared With a Custom AI SDR Agent

The best AI SDR for a team depends on where it wants lead data and email to live. Hosted AI SDR pricing follows lead volume (AiSDR and Artisan pricing pages, November 3, 2026).

AreaSaaS AI SDR toolCustom self-hosted agent
Where it runsThe vendor's cloudYour cloud account or servers
Lead dataThe vendor's lead data where offered, synced with your CRMYour CRM and the data sources you license
SendingYour mailboxes, or vendor-supplied mailboxes and domains where offeredYour own domain and mailboxes, at the caps you set
PricingPer person prospected, or plans sized to lead volumeBuild, hosting and model usage, shaped by scope
ApprovalsThe review options the product offersApproval points you define at each step
Best fitTeams that want a ready-made productTeams that want prospect data, email and logs in their own environment
What Shapes the Engagement

What Shapes an AI SDR Agent Build

Each build is scoped in a strategy session around these factors.

  • Your CRM. HubSpot, Salesforce or Pipedrive, plus the objects the agent reads and writes.
  • Lead sources. The data providers and forms that feed the agent.
  • Mailboxes and volume. How many reps, domains and mailboxes send, and at what daily volume.
  • Approval model. Which steps need a rep and who reviews the logs.
  • Model hosting. An open-weight model in a self-hosted AI deployment, or a provider API under your contract.
  • Compliance review. The consent rules and suppression lists your counsel signs off.

The AI SDR is one of the AI agents for sales, service and back-office work in our AI automation agency catalog.

Talk to Us

Talk to an Engineer About Your AI SDR Agent

Bring your CRM, your sequences and your email rules. We'll walk through the agent design, where it would run and what shapes the engagement.

Book a Strategy Session →
Ask us about
Lead research and personalised outbound email
HubSpot, Salesforce and Pipedrive connections
Self-hosted models and lead data storage
Rep approval queues and audit logs
Consent records, opt-outs and suppression lists
Voice follow-up through a call center agent
Questions

Frequently Asked Questions on AI SDR Agents

An AI SDR is an AI agent that does a sales development representative's prospecting: research, outbound email, follow-ups, reply sorting and booking. An AI sales assistant helps a rep inside their own tools on request. This agent runs the SDR workflow on your CRM, with a rep approving the steps you choose.

As of November 3, 2026, AiSDR's pricing page lists pay as you go at $1.50 per person prospected, with Teams plans priced by volume. Artisan's pricing page lists Team (about 2,500 leads contacted a month) and Scale (about 6,000) plans, priced to lead volume, mailboxes and dialer seats. A custom agent's cost follows its scope: the CRM, mailbox count, approval model and where the model runs. A strategy session compares the two for your volume.

It applies the rules your counsel sets: consent basis recorded per contact, an unsubscribe link in every message, opt-outs suppressed across sequences and messages sent under each rep's name with your postal address. The FTC's CAN-SPAM guide and the CRTC's CASL FAQ both set 10 business days as the outer limit for acting on an opt-out.

Ready to Build Your AI SDR Agent?

A free 30-minute strategy session to map your outbound workflow and outline the self-hosted AI SDR agent that fits.

Discuss Your Project