An LLM like ChatGPT or Claude writes a decent cold email. For a lot of founders and sales leaders, that’s where the idea comes from: “We already pay for it. Why not build our own outbound machine?” Fair question. The trouble is that writing the email is the easy part. Whether outbound works depends on everything around it: following up at the right time, using more than one channel, personalizing with real data, finding the right companies and knowing when to reach out. A chat tool doesn’t do any of that on its own.
What is an LLM actually good at in outbound?
More than people sometimes admit. An LLM is very good at drafting a first email in whatever tone you describe, and at tightening copy that runs long. Give it three facts about a prospect and it’ll turn them into a decent opening line. It’s also a useful sparring partner for subject lines, angles and objection handling.
If you send a handful of emails a week to people you already know you want to reach, an LLM and a spreadsheet will get you surprisingly far. It gets harder when you want something that runs every day, at volume, without you watching it.
What happens to follow-ups when something changes?
Most replies come after a follow-up rather than the first email, so a sequence needs several steps spread over days or weeks. Writing those follow-ups is easy. Sending them only when they should be sent is harder. A real sequence has to deal with situations like these:
- A prospect replies “not now, try me in March”. Tomorrow’s follow-up shouldn’t go out.
- A prospect replies with a question. A scheduled “just bumping this” in the middle of a live conversation looks careless.
- An out-of-office says the person is away for three weeks. Your follow-up for next Tuesday will land in a full inbox and get skipped. It should be pushed back.
- A prospect books a meeting through LinkedIn. Everything else should stop.
In a homemade setup built from a chat tool and a scheduler, each follow-up is usually its own scheduled job and has no idea a reply came in. Handling the cases above means reading incoming mail, classifying it, finding the right scheduled step and changing it. That’s doable. It’s also a project, and when it fails, it fails in front of a prospect. A follow-up sent to someone who already replied leaves a worse impression than sending nothing.
A sales engagement platform treats the sequence and the inbox as one system, so a reply changes what happens next without anyone touching it.
Is outbound only email?
No, and the sequences that work best usually aren’t. People respond on different channels at different times. A typical sequence might start with an email, add a LinkedIn connection request two days later, and put in a call once the prospect has opened a few emails.
An LLM lives in a text box. It can draft a LinkedIn message, but it can’t send it, schedule it as step three, or hand a call task to a rep with the context attached. You can stitch separate tools together for each channel, but then they need to stay in sync. If the prospect replies on LinkedIn, the email sequence has to know.
Adding LinkedIn and phone is often what gets a sequence answered at all, because you reach people where they actually pay attention.
Why is personalization limited with an LLM alone?
Personalization is only as good as the context behind it, and a lot of the best context about a prospect is on LinkedIn: what they post, how they describe their role, what’s changed for them recently. An LLM can’t read LinkedIn profiles and posts on demand, so it works from whatever you paste in or whatever it can guess from a company name.
You’ve probably received the result. The email is fluent, and the “personalized” first line could have been written about anyone. Real personalization needs real data, pulled in automatically for every prospect. Here’s how lemonreach handles personalized outreach without manual research.
Why is prospecting unreliable without a prospecting tool?
Outbound starts with a list, and building a good list has very little to do with writing.
An LLM doesn’t have a live, structured database of companies and people. Ask it for “50 fintech companies in Sweden with 20 to 100 employees” and you’ll get something that looks like an answer. Some of the companies will be real and current, some will be outdated or the wrong size, and you can’t tell which without checking each one.
A prospecting tool queries actual data, so you can filter on industry, headcount, location and funding stage, and on job titles so you reach the right role. You can also filter on web technology. If what you sell only makes sense for companies on Shopify, HubSpot or a specific payment provider, you need to know what their website runs, and a chat tool can’t look that up across thousands of sites.
Then there’s contact data. Finding a person is one step. Getting a verified work email and a phone number that works is another. Unverified addresses bounce, and bounces damage your sender reputation for every email you send after that.
lemonreach’s AI Prospecting handles this part, with verified work emails and direct phone numbers.
What does maintenance really cost?
The first version usually comes together in an afternoon, which is why people underestimate what comes after.
A working outbound workflow is more than a prompt. It’s a data source, enrichment, email verification, prompts for each step, a sending tool, reply handling and somewhere to track status. Any of them can break, and they rarely break at the same time:
- A prompt that worked last month starts producing odd output after a model update.
- A data source changes its format and half your contacts arrive with empty fields.
- You hit a sending limit and nobody notices until replies dry up.
- Two lists overlap and the same prospect gets contacted twice.
Each of these is easy to fix on its own. Together they add up to a part-time job, and it usually lands on someone whose actual job is selling or running the company. The subscription is cheap. What costs you is the hours spent keeping the thing running, and the outbound that doesn’t go out while it’s broken.
What about outreach that starts from a signal?
Most homemade setups are list-based: build a list, write the emails, send them. That works, but it leaves out timing, and timing is the biggest lever outbound has.
Think of a company that’s hiring for a role you help with, or was just in the news, or switched website technology, or has someone posting on LinkedIn about the exact problem you solve. Or one that’s on your website right now. Any of those is a better prospect than a company that only matches your filters, because there’s a reason to write to them today.
lemonreach handles this with monitors. You can set them up for job postings, LinkedIn posts, news such as funding rounds or leadership changes, changes in a company’s website technology, and website visitors.
When a monitor picks something up, lemonreach finds the right people at that company, verifies their details and starts a sequence. Website visitors are the strongest signal of the lot, since the interest is already there. There’s more on that in website visitor outreach.
Building the same thing yourself means watching several sources continuously, deciding what counts as a signal, removing duplicates and wiring each one into the rest of the workflow. Possible, sure. It’s also a lot of plumbing.
What is the upside of building it yourself?
There is one, and it’s real: you get to make everything exactly the way you want.
A custom workflow can follow unusual rules, connect to tools no vendor supports, use your own scoring and write in your voice. If your process really is unusual, or you have a technical person who likes owning this kind of system, that freedom is worth a lot.
What it costs is time. Every custom decision is something you design, test and then maintain. A purpose-built platform gives you less freedom at the edges, but the core works from day one and someone else keeps it up to date.
Build or buy: how do you decide?
Five questions usually settle it.
- How important is outbound to your growth? If it’s a core channel, reliability matters more than flexibility.
- Who will own the system? If the answer is “whoever has time”, it’ll drift. Someone has to look after it every week.
- How much do the gaps hurt at your volume? At ten emails a week, a missed follow-up is a small annoyance. At a few hundred, it starts to damage your reputation.
- Do you need signals, LinkedIn and phone? If you do, list what it would take to build each one.
- What’s your time worth? A cheap tool gets expensive fast if it eats your week.
If outbound is a side experiment, an LLM and a spreadsheet are a fine place to start, and you’ll learn a lot from it. If you want outbound running every week and growing, keep in mind that the email was never the hard part. An LLM can write it. Somebody still has to run everything around it, and it’s worth deciding now whether that should be you.
You can read more about what an AI SDR does in what is an AI SDR, or book a demo and we’ll show you what outbound looks like when the system around the email is already built.
If you decide to buy, compare lemonreach with Artisan, Alta, AiSDR, and 11x.