How to Build an Outbound Engine With Clay and n8n

What is an outbound engine, and why build one?

An outbound engine is the repeatable path from a list of companies you want to work with to a booked meeting. The difference from ordinary outbound is that software does the research, the data work and the first draft, and a person only does the parts that need judgement: deciding who is worth contacting and approving what goes out.

The reason to build one is simple. When sales teams chase cold lists by hand, most of their time goes into finding emails, copying details between tabs and writing the same opening line again. That work is slow, inconsistent and hard to measure. A system makes every step visible, so you can see where accounts drop out and fix that step instead of guessing.

At Attio I built this kind of workflow. It covered 200 target accounts with a 72% data match, a 9% reply rate and 14 meetings in 6 weeks. The rest of this guide walks through the same shape so you can build your own.

Step 1: How do you choose the target list?

Start narrower than feels comfortable. A few hundred accounts that match your ideal customer profile beat thousands that loosely fit. Write the profile down as filters you can actually apply: industry, size band, region, the tools they already use, and one buying signal such as a recent hire in the team you sell to or a new product launch.

Keep the list in one table with a stable ID per account. Every later step adds columns to that same row, which is what lets you trace a meeting back to the exact data and message that produced it.

Step 2: What is waterfall enrichment and how does Clay do it?

Enrichment means adding the facts you need to contact the right person: their name, role, verified work email and a detail worth mentioning. No single data provider has all of it. Clay's answer is the waterfall: it searches providers in sequence until one returns a valid match, and you pay only when a provider finds a result. Clay lists more than 150 databases you can line up this way.

Order the providers from cheapest or most accurate for your market to the most expensive, and stop at the first confident result. Track the match rate as its own number. If it is low, the problem is usually the list (wrong company domains, people who left) rather than the providers.

• Verify every email before it enters a sequence; bounces hurt the whole domain.
• Store which provider returned each field, so you can drop providers that add cost without adding matches.
• Keep one research field per account: the specific reason you are writing to them.

Step 3: How do you write a first line that gets replies?

The first line decides whether the rest gets read. It works when it is about the reader's own situation, not about you. A language model can draft it from the research field you stored, but give it strict rules: one sentence, one specific fact, no flattery, no invented details. Anything the model cannot back with a stored fact should be left out.

Psychology explains why this works. A small, specific gap in knowledge creates an urge to close it, so a line that names something true about the account and hints at a useful idea earns the next sentence. A generic compliment closes the tab.

Step 4: Why add a Slack approval step with n8n?

Fully autonomous outbound sends mistakes at scale. A human approval step costs seconds per message and catches the drafts that are wrong, awkward or aimed at the wrong person. n8n supports this directly: a workflow can pause and wait for a person to approve or deny, and the request can arrive in Slack, Microsoft Teams, Gmail and several other channels.

In practice the account owner sees the draft, the research it was based on and two buttons. Approved messages go to the sending tool; rejected ones go back with a reason, which becomes a rule for the next batch.

How do you keep the engine improving week after week?

An outbound engine is never finished. Each week, look at where accounts dropped out and change one thing. If the match rate fell, check the list source and the provider order. If replies fell, compare first-line styles and retire the weakest. If meetings stalled, look at who replied: the right company but the wrong person usually means the role filter needs tightening.

Write every change down with the date and the number it was meant to move. After a month you will have a short record of what actually works for your market, which is worth more than any template. Rejected drafts from the Slack approval step are the richest source of lessons, because each one shows a rule the model has not learned yet. Turn the common ones into instructions for the drafting step.

Finally, keep volume honest. The point of the engine is better conversations, not more email. If a change raises sends but not meetings, undo it.

Step 5: What has to be in place before you press send?

Deliverability comes before volume. Gmail requires bulk senders to set up SPF, DKIM and DMARC and to keep spam rates below 0.3%, and Microsoft has rejected non-compliant bulk mail to Outlook.com, Hotmail.com and Live.com addresses since May 5, 2025. Send from a dedicated subdomain, warm it up gradually and watch replies and complaints daily.

Then measure the engine end to end. The four numbers that matter are accounts reached, data match rate, reply rate and meetings booked. Add attribution on every send so a closed deal can be traced back to the list, the data and the message that started it.