of most builders’ work. We use instruments comparable to Cursor, Windsurf, OpenAI Codex, Claude Code, and so forth, to turn out to be way more productive at work. Nonetheless, from discussions with folks working in non-technical strains of labor, I typically see lots of processes that may be optimized utilizing AI.
On this article, I’ll present how different non-tech industries can profit from utilizing AI. I’ll spotlight this by automating the method of discovering prospects with AI, a typical gross sales job that requires lots of guide work. The aim of the article is to spotlight how even non-technical folks could make use of the newest coding brokers to create highly effective automation instruments.
All through the article, I’ll be highlighting my essential suggestions with:
This can be a essential tip
Why do we’d like automated prospect discovering
Enterprise improvement representatives work with:
- Discovering fascinating prospects
- Retrieving their contact data, comparable to their LinkedIn profile or e-mail tackle
- Reaching out to prospects, making an attempt to arrange a gathering
From there, an account government often takes over, although I’ll be specializing in the way to optimize the primary three steps.
This three-step course of is often fairly intensive, as discovering prospects on-line requires scouring by lots of LinkedIn profiles or different web sites to seek out fascinating firms. After discovering an organization, you sometimes begin in search of particular folks within the group to achieve out to. This may often be key resolution makers, who could be a center supervisor in a bigger firm, or a CFO in a smaller firm. After discovering the right particular person, it’s good to get their contact data, often discovered on LinkedIn or on the corporate’s web site. Lastly, you need to attain out to this particular person with a customized message.
Discovering fascinating prospects
I’ll begin growing this instrument utilizing Claude Code. You’ll be able to primarily use any coding instrument there, comparable to Codex, Cursor, Windsurf, Replit, and so forth. The principle level is the instructions you employ to make the applying.
All the time use planning mode earlier than creating a brand new software
I all the time begin off utilizing plan mode, which tells the mannequin to learn your immediate, create a step-by-step plan, and ask any clarifying questions. That is very helpful, because it helps you slender down the scope of your software and ensures you present your ideas on any ambiguity. For instance, I obtained prompted on:
- Which location are you concentrating on (in my case, I’m taking a look at prospects in Norway)
- Which programming language would you like your software in (I selected Python, however TypeScript or different languages are in fact additionally attainable)
- What are some web sites you need to have a look at (in Norway, proff.no has lots of data on firms and workers, so yow will discover comparable instruments, and even ask your coding agent to seek out these web sites itself utilizing an internet search)
- What output format do I choose (I selected an Excel sheet)
These are all nice inquiries to make clear, which is why utilizing plan mode is so helpful.
Additionally, be sure that to immediate your mannequin to make use of publicly out there API’s, and solely search for data from related firms. Fetching private data may very well be in opposition to the laws, relying on the place you reside.
All the time present your coding agent as a lot instruments as attainable. It may very well be MCP servers, OpenAI credentials, or entry to any packages by an API.
Moreover, I additionally made an OpenAI key, which I instructed Claude Code it may load from my .env file with OPENAI_API_KEY. That is helpful as a result of lots of operations will in all probability require utilizing an LLM to seek out or course of data. Thus, offering Claude Code entry to a robust API service is essential. When you’ve got different related API’s, you also needs to present Claude Code with the documentation for these API’s and inform it to make use of them.
I, for instance, knowledgeable Claude that it has entry to internet search, and may carry out it with the operate under:
response = shopper.responses.create(
mannequin="gpt-5",
instruments=[{"type": "web_search"}],
enter="What was a optimistic information story from right this moment?"
)After I answered all of the clarifying questions, I instructed Claude to begin constructing, and it constructed an software to discover a record of prospects, returned in CSV format. This covers each step 1 and a couple of, of first discovering fascinating prospects, and getting a maintain of their contact data.
After discovering all of those prospects, you also needs to do a guide evaluation, making certain correctness. Moreover, I like to recommend prompting GPT-5 or an equal mannequin to undergo your outcomes and confirm any inconsistencies.
Have an LLM undergo your outcomes to confirm correctness
Lastly, it’s additionally necessary to abide by laws when discovering prospects. You must solely be discovering related firms on-line, after which discovering people manually to abide by GDPR laws. Thus, to seek out further data, such because the title, e-mail, and function of particular person prospects, I manually discover the knowledge from the businesses supplied with my software.
Reaching out
After discovering contact data, you now want to achieve out. You’ll be able to learn lots of stats and data on the way to carry out chilly emails, however I’ll not get into that right here, since my focus is on the expertise and the way we are able to use it to optimize our processes.
Up so far, I assume you’ve fetched an inventory of related prospects, together with their contact data, and also you’re prepared to begin reaching out. We now need to create custom-made messages to every particular person, which is fortunately a process that LLMs are actually good at.
LLMs are capable of craft customized messages
For instance, at this stage, I’ve the next data per prospect:
- Particular person title
- Particular person e-mail
- Particular person function
- Firm title
- Firm dimension
- Firm income
Additionally, basically, including extra data to your prompts is often higher. In the event you choose a selected type or tone in your emails, you need to add that data. A good suggestion can be to point out examples of you’re earlier emails, highlighting the way you write emails your self, and thus utilizing few-shot prompting to enhance output high quality.
Add as a lot data to your prompts as attainable
I’ll now use this data to draft up a custom-made message. This may be finished comparatively merely utilizing GPT-5 for instance:
immediate = f"""
You might be an professional at creating customized emails. You might be given data
about a person and should create an e-mail to achieve out to them for the
first time.
Title: {title}
E mail: {mail}
Function: {function}
Firm title: {company_name}
Firm dimension: (company_size}
Firm income: {company_revenue}
Create each a topic tag, and the complete e-mail, together with no different feedback
or reasoning.
"""
shopper = OpenAI(api_key=OPENAI_API_KEY, base_url=API_URL)
import os
from openai import OpenAI
end result = shopper.responses.create(
mannequin="gpt-5",
enter=immediate,
reasoning={ "effort": "low"},
textual content={ "verbosity": "low" },
)When you could have the define of the message, you’ll be able to then manually tweak and optimize it for the particular person you’re reaching out to, given the knowledge you’ve discovered.
Now you can attain out utilizing these emails. To keep away from breaching any phrases of service and spamming folks, I like to recommend reaching out manually and never utilizing an automatic service. I don’t condone spam emails or comparable. AI is just used that can assist you and make processes more practical, to not take away all people from the loop essentially.
Conclusion
On this article, I’ve highlighted how you need to use the newest coding instruments, comparable to Claude Code, to automate some processes. On this article particularly, I lined the way to optimize the method of reaching out to prospects through the use of AI to mechanically discover related firms and mechanically create emails whereas staying compliant. We’ve seen immense progress with AI in the previous few years, however I nonetheless consider AI is lagging behind on the implementation facet. Thus, in case you might be quick at integrating AI into your day by day life, you’ll be able to have an enormous benefit over your friends.
👉 Discover me on socials:
🧑💻 Get in touch
✍️ Medium
You may also learn my different articles:
