Twitxy Saved

Jimmy Wins!

Verified account This account is verified. Learn more

@JimDicarlo

Remote home service biz operator since 2014. Now implementing AI and helping other home service operators do the same. El Salvador based. #girldad
Joined October 2015
33 Following 55 Followers 40 Posts

Latest posts 5

Pinned post

Jimmy Wins! Verified account @JimDicarlo
In 2014 I was 35, broke, and driving Uber at night so I could work on a house cleaning business during the day. The thing that got me through was working on my mindset as hard as I worked on the business.

Today that business and two others do about $2M a year between them. A lean team runs them, with an offshore crew I built in the Philippines in 2016 and AI built into how we run operations. I run all of it from El Salvador and can go a week without checking in.

I believe the next wave of home service companies will be AI-native from day one. Now I help other owners build their businesses that way. This account is where I build in the open and share my life as a builder along the way.
Jimmy Wins! Verified account @JimDicarlo
Last night I went to the Build in El Salvador @BuildInES networking night, and it was the best networking event I've ever been to.

The room was packed with serious people from hospitality, btc, real estate, construction, tech, health and wellness, finance and more. Almost everyone there is actively building something. Whatever sector you wanted to connect with, someone from it was in the room.

The BIES team did an incredible job. The venue was beautiful, they thought of every detail, and the energy was buzzing from the moment I walked in.

What stood out most was that I met people I never would have crossed paths with on my own. You could feel how hungry everyone was to connect.

I've been part of BIES for about a year now, and I tell people all the time that I've never had more like-minded friends in my life. I had to come all the way to El Salvador to find that. The people here are high agency, creative and authentic, and many are here for a real purpose. That sense of belonging means more to me every year.

There is a wealth of talent, connections, resources and opportunity in this community. We're building it so that the more it grows, the more opportunity it creates for everyone in it.

This country is going through a renaissance that's hard to capture from the outside. You have to come and feel it. I'm extremely bullish on El Salvador and grateful to be here.
0:17
4 replies 9 reposts 44 likes 1.9K views View on X
Jimmy Wins! Verified account @JimDicarlo
Hiring cleaners has been my biggest headache for years. We now have an AI agent that sorts applicants, follows up by email and text, and books phone screens. A person still does every screen, and I still make every hiring decision. It's still early.
Jimmy Wins! Verified account @JimDicarlo
OpenAI just let businesses claim a profile inside ChatGPT. I think this is a big opening for home service companies, and most owners won't notice it for a while. Here's how I see it for cleaners, plumbers, HVAC and the rest of us.

1/ This reminds me of Google Ads in the early days, when clicks were cheap and impressions were easy to get. The owners who claim and fill out their profile now will have the field mostly to themselves.

2/ The feature I like most is letting people chat with an agent about your business. It can answer the questions people hesitate to call about. When they're ready to book, I'd still hand them to a real person.

3/ An agent can only be as good as what you write for it. Your service areas, what a deep clean includes and your cancellation policy all need to be written down first. Most home service owners still keep that in their heads...

4/ If that's you, you can start today using your call history. There's gold in them calls! Pull six months of call transcripts from your phone system, like Quo, and have AI pull out the 100 questions customers ask most, along with how your team actually answers them. A tool like @StatusLineAI allows you to pull this super fast. That's your FAQ and the start of your SOPs.

5/ More searching will be done by agents rather than people. Someone will ask ChatGPT or Google Gemini to find a cleaner who can come Thursday, and their agent will compare the options. I think the companies with AI business pages and clear, current information will get picked more often. Being seen by the LLMs and agents is becoming increasingly important.

6/ Your information has to match everywhere. If your website, Google profile and ChatGPT profile show different hours, prices or service areas, the AI has to guess...and probably get things wrong. Spend the time to make sure all your info syncs up.

7/ You can't see the conversations, so ask every new customer how they found you. You should be doing this anyway now. Search is shifting more to AI and it's good to at least have an idea of how much your search is coming from that source.

8/ 1.2 billion people use ChatGPT every week. I'd claim my profile today.
OpenAI Verified account @OpenAI Nov 10
Now in beta: Businesses in ChatGPT.

Claim your business, manage your information, and get insights on how people engage with you.

Eligible businesses can also build a Business Agent to answer customers’ questions and help them explore what you offer.

openai.com/index/introduc…
View quoted post on X
Jimmy Wins! Verified account @JimDicarlo
When I put AI into a workflow, I start by splitting the work into deterministic steps and latent steps.

Deterministic steps are code. A trigger, a webhook, or an API call should do the same thing every time. In our hiring process for example, once a person records an approval (ie. move a candidate to onboarding flow), a trigger can send the orientation invite. That step does not need a model.

Latent steps are judgment. These need intelligence or actual AI. Our Slack agent, for example, reads payroll adjustments in plain language, including voice messages, and figures out what was promised. It then puts together a report based on the information collected. That is a good use of a model.

If I put a model on every step, the system costs more and is less durable. I would be paying it to do work that a simple automation can already do reliably.

I keep judgment in the model and the mechanical steps in code. That is how I decide where AI belongs.