Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More
197 contributions to AI Automation Society
One instruction file per project, not one for everything
My first setup had one big instruction file for the AI agent that covered everything I do. It got too long, parts of it went out of date, and in the end nobody read it. The reason is simple. A session that has everything open carries everything in its context. The agent reads rules for projects it isn't working on, and the rules that matter get lost between them. So I split it by folder. I have eleven top-level folders, and each one has its own instruction file. It loads by itself as soon as the agent touches a file in that folder. For each project, the file answers four questions: what is it about, where is the current state, what comes next, which tools exist. The top file only has pointers and the rules where breaking them was expensive. Anything about one project stays in that project. I also work with one main session per folder. I give it the folder name and it finds the state on its own. If you build for several clients or products with an agent, try this before you write the next long prompt file: one folder, one instruction file, one session.
0 likes • 2d
@Wasilis Paliakoudis That is exactly how I keep it: the project file only says where the current state lives and never holds it. The one thing I still can't do is notice when the pointer itself goes wrong, like when the folder it points to gets renamed. So far I only find out when the agent comes back empty.
0 likes • 5h
@Madonna Lapid Good point, the why is what I lose first. Right now a rule is just a line, and after a few weeks I can't tell why it's there. I'll add a short reason to each one, and if I can't write the reason anymore, that rule goes out.
First Website with the process
Hello guys, I stated the process and here is the website I build for a dentist clinic, i contact them and they interested to see the website so I send à WhatsApp message and waiting now. So happy now 😊😊 Can you give me feedback for this? I will appreciate ✌🏾 Https://cabinetlumiere.roschcorp.com
First Website with the process
1 like • 9h
Congrats on getting a "yes, show me" from a clinic, Alexis. Building the site before writing to them is the right order, and I'd keep doing that. Now that you're waiting: don't send a "just checking in". Send one specific observation about their site instead, for example "your booking page needs three taps on a phone, I'd cut it to one". Then they have something to answer, not just something to ignore. I do the same with my own outreach. "I make great 3D visuals" got me nothing, while "56 product photos and on none of them you can see how the lid comes off" got replies. Which part of the dentist's day were you trying to fix with this site?
Can I get clients in 7 days?
Hey I joined this group few months back and I have claude subscription but I don't know what to build I want to ask can I get clients in 7 days?
2 likes • 1d
Seven days is realistic for a first conversation, not for a signed client, and the conversation depends on one decision: what you build for whom. Tomorrow, pick one kind of small business and read 20 of their open job ads. When a company hires someone for a task like entering invoices or chasing payments, it describes the manual work in its own words, and that's your thing to build. From 299 job ads I pulled, 46% came from recruiters and staffing agencies, so I threw those out first. Only 62 were left with three or more tasks I could actually automate. Twenty ads can already give you your first three companies to write to. Which kind of business do you know best from your own work so far?
Research Agent for Cold Calls
Hey guys, I’ve made over 30,000 cold calls in the past year, so I have a pretty good understanding of what to say and how to say it. Now I’m looking to automate the research and script personalization. What works really well for me is researching each company right before calling. I enter the company name or website into a prompt I’ve built, get a personalized script based on my existing template, and make the call. The problem: this sometimes takes more than 30 seconds per company, which slows me down and breaks my calling flow. I could research everything in advance, but some phone numbers turn out to be wrong, and some contacts no longer work at the company. That means spending time researching leads I can’t actually reach. I haven’t tried batch research yet, but before spending hours doing it manually, I’d rather build a repeatable workflow that processes an entire lead list for each service I offer. Ideally, it would: - Research each company’s specific services and target customers. - Personalize my existing script without changing its wording or structure beyond the designated placeholders. - Identify or verify the relevant decision-maker and, where possible, suggest an alternative if the original contact isn’t the right person. If it turns out the person is not a part of the company it would be great if the agent could check if its the case so the suggestion of other potential decision makers also would save me time! Important Note: I have different projects and therefore the pitch needs adjustments especially if it's a new project. So it would be good if I have the flexibility to change the one part that need the adjustment! No matter which service I am reaching out the structure remains the same just the service pitch differs. The goal is to have concise, call-ready scripts prepared so I can focus entirely on calling. Has anyone built something similar, or can you recommend a setup for this?
1 like • 2d
Those 30 seconds are the price of doing the research live, so I'd move it out of the call flow instead of making it faster. Run a cheap check first: is the number still valid and is the contact still listed on the company's own team page? Drop the ones that fail. Then research the survivors in the background while you're on the previous call, so the script for the next lead is ready when you hang up and you never research someone you can't reach. I'd test it on 10 leads before building more and see how many die in the check. After 30,000 calls: which part of your script changes most from company to company?
3 numbers I check before adding volume to any outreach sequence
Volume makes a bad sequence fail faster. Before I send more, I check three numbers on the first 100 first messages. 1) Calls booked per 100 first messages. I want at least 1. 2) Positive replies. I want 5% or more. 3) Whether the sequence has run its full follow-ups. Most bookings land after follow-up 5, so a count taken on day 3 lies. The math: 50 to 100 messages a day at 1 call per 100 is about one call a day, so roughly 30 calls a month. If a number is under the line, I fix the message, not the volume. Which of the three do you track today?
1 like • 3d
One step you can take tomorrow: before you work out messages per day, count your real bottleneck on LinkedIn, which is connection requests, not messages. In my own data, 3,671 LinkedIn connection requests from 20 users got 36% accepted. That comes to roughly 280 requests for every 100 new contacts. With LinkedIn's weekly cap on invitations, that is the limit, not timing, because weekday and time of day barely moved the acceptance rate. Are your 100 first messages going to people you're already connected with, or to strangers?
1-10 of 197
Lorenz Wieseke
6
957 points to level up
@lorenz-wieseke-4768
AI Automator | Founder of Inboundy.app https://inboundy.app

Active 5h ago
Joined Sep 16, 2025
Powered by