Build an agent run outbound stack: tools, cost and owner hours

An agent run outbound stack has seven layers: audience, data, signals, sending, replies, calendar and reporting. With one platform covering most of them, a small team spends about five hours a week as the owner, and the software costs about $249 a month for an outbound bundle or $198 per LinkedIn agent. The hours are an estimate. They fall as you write rules and read fewer threads.

A lot of outbound advice lists tools. This post lists layers first, because the question that matters is which jobs an agent does and which a person still does. The tools follow from that.

What are the seven layers?

Two of the seven stay human on purpose: the audience and the exceptions. Our own data is why. The audience changed meetings per person tenfold on our own LinkedIn outbound, with the same agents and offer, and no tool chose it for us.

Which tools sit in each layer?

You can use one platform for several layers or several tools for one. A small stack could look like this.

• Audience. A document with your ICP and an exclusion list. The ICP template is a start.
• Data. A contact database. Sellinger offers 700M+ contacts, and a lookup of an email, a company or a person costs $0.01. A LinkedIn search import is also possible.
• Signals. Intent signals on LinkedIn with 26 ready made templates in 8 categories, included with every bundle.
• Sending. LinkedIn, email, WhatsApp, Instagram and live chat agents in one platform, each with its own limits.
• Replies. The same agents, with approval mode and takeover.
• Calendar. Cal.com or Calendly, both live integrations.
• CRM and reporting. The built in CRM, or HubSpot with a two way sync, plus a weekly sheet.

Salesforce, Pipedrive and Zoho are shown as coming soon, not live. If you use one of them, plan an export or the API for now.

What does it cost?

Sellinger list prices, checked on 19 October 2026:

Messaging modules and bundles have a 30 day free trial with no card. Competitors price differently and change often, so use the dated, sourced table in the AI SDR pricing comparison instead of a number from memory. For the whole cost model, including data, infrastructure and what a wrong message costs, read the real cost of owning an AI SDR.

How many hours does the owner spend?

This is an estimate, not a measurement, and it is the number vendors leave out. For one campaign:

• Setup, once: a few hours for the audience, the offer, the rules and the knowledge.
• First weeks: an hour a day reading threads and fixing rules.
• Steady state: one to two hours a week reading flagged conversations, approving what needs approval and looking at the numbers.

Plan around five hours a week for a small team with several campaigns. At a $50 hour that is about $1,000 a month of time, so the full month is about $1,215 with a $198 agent. At a $150 hour the time costs about $3,000.

Review time falls with experience. On our own LinkedIn account, 93% of replied conversations since May 2026 ran with no human message, after a long period of setting rules and reading threads. That is a result of our account and our effort, not a promise. For a human comparison, a human SDR costs about $114,000 a year before tools and management, with a 3.0 month average ramp, and our comparison of an AI SDR and a human SDR sets out the sources.

What does the owner do each week?

A short routine keeps the hours down.

1. Monday: read the flagged conversations. The ones the agent stopped on, and the first replies of any new list.
2. Monday: check the numbers. Replies, meetings booked and meetings held per person messaged, by list.
3. Midweek: fix one rule. Whatever the flagged threads showed. One change at a time, so you know what caused what.
4. Friday: queue the next list. A small, scored batch with a reason to write.
5. Monthly: retire a source. If a list gave no meetings in a month, stop using it.

The post on setting up your first AI SDR in one afternoon covers the setup decisions.

What goes wrong in a stack like this?

• Too many tools. Every extra tool is another login, another bill and another place a lead can go missing.
• No one owns the numbers. Without a weekly check, a falling meeting rate goes unseen. On our own account the reply rate held near 40% all year while meetings per person messaged drifted below 1% from June 2026.
• Automating the exception. A price question or a complaint answered by an agent with no rule is a bad day.
• No exit. Decide which results would make you stop a campaign.

How do you grow the stack without losing control?

Add one thing at a time and measure it. A second channel is a good example. Sellinger's defaults move a lead to email after four LinkedIn follow ups and to WhatsApp after five email follow ups, and every value can be changed per campaign. Switch on a second channel for one list, compare its meetings per person messaged with the same list on one channel, and keep it only if it adds meetings.

Do the same for each new tool. If it does not move meetings held, or save a named number of owner hours, drop it. The cheapest stack is the one with the fewest parts that still gives you the meetings.

A checklist

• [ ] I wrote down the seven layers and who does each.
• [ ] Audience and exceptions stay with a person.
• [ ] I have one platform for sending and replies, not three.
• [ ] I priced the software and my own hours.
• [ ] A weekly routine of under two hours exists, with one rule change per week.
• [ ] I measure meetings per person messaged, not only replies.