Automate the list work of sourcing (searching, enriching, deduplicating and tracking) and keep a person in charge of the first message. Outreach sounds like a bot when it is generic, wrong in a small detail, sent at an odd time, or repeated without thought, and each of those has a practical fix. This guide covers what to automate, what to write by hand, how to personalize without it feeling like a template, and the checks to run before anything is sent.
What candidate sourcing automation covers
Sourcing has two halves that behave very differently. The first is data work: finding people who match a brief, filling in missing details, and keeping your records tidy. The second is communication: deciding who to contact, writing to them, and handling what they say back. Automation is strong at the first half and risky in the second.
| Sourcing task | Automation fit | Why |
|---|---|---|
| Building search lists from your criteria | High | Repeatable, rule-based, easy to review |
| Filling in missing fields (company, title, location) | High, with verification | Saves time, but data goes stale and can be wrong |
| Checking against your existing database | High | A matching rule finds people you already hold |
| Checking prior contact and do-not-contact status | High | Prevents embarrassing repeat messages |
| Ranking candidates by fit to the brief | Assist | Software orders the list, a person decides |
| Drafting the first message | Assist | A person approves before it goes out |
| Sending approved messages | Partial | Only after review, with limits |
| Reading and answering replies | Human | Questions and objections need judgment |
The pattern is the same one that runs through any good automation plan: automate the preparation, keep a person on the moments that shape how someone feels about you. For the broader framework, see what to automate first and what to keep human.
Why automated outreach sounds like a bot
Candidates can usually tell within a sentence. These are the most common reasons:
- The generic opener. Lines like "I came across your impressive profile" fit everyone, so they tell the reader nothing.
- Visible mail-merge errors. An empty name field, the wrong company, or a job title in capital letters gives the game away immediately.
- The wrong role for the person. A message about a position that is clearly irrelevant shows nobody checked.
- No useful facts. If the message does not say where the job is, what level it is, or what the company does, the candidate has to ask before they can decide whether to care.
- Odd timing and rhythm. Messages that arrive in the middle of the night, or follow-ups that repeat the same text, feel mechanical.
- No easy way out. If a person cannot decline simply, the message feels like pressure.
Notice that none of these are caused by automation itself. They come from using automation to send more messages instead of better ones.
Personalization that scales
Putting a first name in a template is not personalization. A message feels personal when it explains why this particular person is being contacted and gives them what they need to respond. Build it in three layers:
- Relevance. Decide who gets a message by using tight criteria taken from the brief. A smaller, better-matched list beats a large one, and it makes every later step easier.
- One true, specific detail. Pull a single accurate detail from the candidate's record that connects to the role, such as a type of project they worked on, a technology they use, or the kind of company they have worked for. If your data does not contain a reliable detail for someone, leave the line out instead of inventing one.
- The role facts, up front. Say where the job is, whether it is remote or on site, the level, the pay range if you are allowed to share it, and what the company does. People decide quickly, and facts help them decide.
The difference shows up in the text. Here is a weak version and a stronger one. Both are illustrative, and the square brackets mark details you would fill in from real data:
Weak: Hi [First name], I came across your impressive profile and think you would be a great fit for an exciting opportunity. Let me know if you are open to a chat!
Stronger: Hi [First name], I noticed you led the migration of a billing system at [Company]. I am recruiting a senior backend engineer for a payments company in [City], hybrid, [pay range if available]. The role owns similar billing infrastructure. If that sounds relevant I can send the details, and if not, tell me and I will not contact you again.
The stronger version is not longer or cleverer. It says why this person, what the job is, and how to respond or decline.
Templates that do not feel templated
You still need templates, because writing every message from scratch does not scale. The trick is to template the structure and let the content vary where it matters.
- Build blocks, not one fixed message. Keep a few openers for different types of candidate, a block for the role facts, and a short closing line. Combine them by segment, such as seniority or industry, instead of randomizing.
- Use variables only for data you trust. Define what happens when a field is empty. If a required detail is missing, the message should not send, not send with a gap.
- Write the first version yourself. AI can help draft variations, but give it the real facts and tell it not to add details that are not in your source data. Models can invent plausible but false specifics, and a false detail is worse than none.
- Vary the follow-ups. A second message should add something new, such as a fact about the team or a clearer description of the role, not repeat the first.
Review checkpoints before anything is sent
A short checklist catches most of what makes outreach look robotic. Run it before the first send and again when you change the list or the template.
- Check the list. Open a sample of about twenty records and confirm that the person is in the right field, the right location, and still in the job your data says they are.
- Check contact status. Remove people who have opted out, who you contacted recently, or who are in an active process with you.
- Preview rendered messages. Look at ten messages with real data filled in. Hunt for empty fields, strange capitalization, and wrong company names.
- Trace each personal detail. Every specific line should come from a field in your data. If you cannot point to the source, delete the line.
- Set sending limits. Cap messages per day, per person, and per company, and schedule sends for business hours in the candidate's time zone.
- Send a small first batch. Send to a handful of people, read every reply, and fix what you learn before scaling.
- Put a person on replies. Anything that is not a clear yes or no goes to a recruiter, and opt-outs are processed straight away.
Legal and platform limits
The rules differ by country and by channel, so treat this as a prompt to check your own situation rather than a complete guide.
- Personal data. Laws such as the GDPR require a lawful basis for using personal data, expect you to tell people how you obtained their details, and give them rights to object and to ask for deletion.
- Platform terms. Many professional networks and tools restrict automated activity in their terms, and breaking them can get an account limited or closed. Read the terms for any platform you use.
- Email rules. Requirements for business email, identification, and unsubscribe handling vary by jurisdiction.
This is general information, not legal advice. For a real process, speak to a qualified adviser who knows your market.
How to tell whether it is working
Compare against what you achieved manually before you automated:
- Positive reply rate. Replies from people who are interested matter more than total replies.
- Opt-out and complaint rate. If this rises as volume rises, the messaging has become a problem.
- Bounce rate. A high rate points to poor data.
- Conversion to screening. The share of interested candidates who go on to a conversation.
- Time from vacancy to first screened candidate. This is the speed benefit you are really after.
Be careful with reply rate in isolation. Aggressive tactics can raise it for a short time while damaging your reputation with the people you will want to contact again.
Common mistakes
- Automating the sending before the list is good. Poor targeting at scale only produces poor results faster.
- Fake personalization. Referring to a post or project the candidate never wrote or worked on loses trust at once.
- Repeating the same follow-up. Each touch should give the reader a new reason to respond.
- No caps. Without limits per person and per company, it is easy to over-message the same people.
- Ignoring replies. A fast, human answer to an interested candidate is where the value of sourcing is realized.
- Measuring sends instead of outcomes. The number of messages sent says nothing about whether the work is succeeding.
Summary
Use automation to build and clean your lists, check your records and keep track of activity, and keep a person on the first message, on replies and on the final decision about who to contact. Personalize the reason you are writing, put the role facts first, and review before you send. For the wider picture of where automation fits in an agency, read about twelve agency processes you can automate or browse the rest of the blog.