Claude Code + AIsa : j’ai créé un agent IA qui automatise ma prospection commerciale — Transcript
Full transcript
- 0:00One of the questions you ask me most on
- 0:02this channel is: how do you find
- 0:04clients when you offer artificial
- 0:06intelligence services? And that is
- 0:08exactly what we are going to look at in
- 0:10this video. If you are new to the
- 0:12channel, I’m Etel Cédric, actor and
- 0:13data scientist. And in this video, I
- 0:16will show you step-by-step how to use
- 0:18Claude Code to create an AI agent
- 0:20capable of searching for clients 24/7
- 0:23for your business. So, get comfortable,
- 0:26let’s get started. OK, so we’re
- 0:29going to head over to Claude Code and
- 0:31start the project together. For this
- 0:33video, I assume everyone has used
- 0:35Claude Code at least once. If that’s
- 0:38not the case, it’s very simple to
- 0:40install. You simply go to Google and
- 0:43search for Claude Desktop. You will
- 0:46install the application, create an
- 0:48account, and once you open the
- 0:50application, you will see this
- 0:52interface. So, we can start the project
- 0:55. On my PC, I created a "client
- 0:58prospecting" folder for this project
- 1:00with a "Claude Project" directory. And
- 1:03in that directory, I put two Markdown
- 1:05files. The first file is the "claude.md
- 1:08" file, which represents the entire
- 1:10memory and context of my project. I’m
- 1:13going to open it so we can read through
- 1:15it briefly together. So, in my "
- 1:17claude.md" file, what did I specify?
- 1:20First, I presented the project to it. I
- 1:23told it that this project consists of
- 1:25developing a sales prospecting agent
- 1:27for Neflow AI. So, Neflow AI is my
- 1:31artificial intelligence agency. I
- 1:35defined the goal, which is to create an
- 1:37application that will automate part of
- 1:39the research and qualification of local
- 1:42prospects. Basically, we are going to
- 1:46provide a set of information to our
- 1:47application, specifically the business
- 1:50sector. It could be the hospitality
- 1:53sector, for example, the restaurant
- 1:56industry, or SMEs specialized in
- 1:58digital services. The second piece of
- 2:02information we will provide to our AI
- 2:04agent is the city or geographic area.
- 2:06It could be a country, a city, or a
- 2:08small neighborhood so that it can
- 2:10conduct a targeted search. The third
- 2:13piece of information we will ask the
- 2:14agent for is the number of prospects we
- 2:16want. Once we provide this information
- 2:19to the agent, it will then search for
- 2:22companies. It will quickly analyze
- 2:24their digital presence and generate
- 2:26personalized prospecting emails.
- 2:29Secondly, I will specify the main
- 2:31objective, the target workflow. OK? In
- 2:35a third step, I describe here the user
- 2:38inputs. OK, in a fourth step, I will
- 2:40therefore specify the data sought. What
- 2:43am I looking for in companies, exactly?
- 2:45The name, address, phone number, rating
- 2:48, number of reviews, website, social
- 2:51media, and useful information to
- 2:53understand their digital presence. In
- 2:57fifth, I will therefore specify some
- 2:59qualification criteria for the leads.
- 3:01So each company is filtered into a
- 3:04priority of either high, medium, or not
- 3:07relevant. So that is generally what my
- 3:09cloud file looks like. MD looks like.
- 3:11The second file is the prompt. It is
- 3:13the prompt that will allow us to create
- 3:15the AI agent. So I am opening it so we
- 3:17can see together what it looks like. So
- 3:19here, what is the agent's mission? You
- 3:22are the sales prospecting agent for
- 3:23Neflow EAI. So your mission is simple.
- 3:26Find local companies matching the
- 3:28requested criteria. I would like to
- 3:32clarify that I will put both files in
- 3:33the description so that you can
- 3:35download and modify them according to
- 3:37your own context. Now that we have
- 3:40presented the context of this project,
- 3:42we can move on to practice. We are
- 3:44going to go into Claude. I am opening
- 3:45it here. You can see that this is the
- 3:47chat section. So you can write here,
- 3:50you hit enter, and you get a response.
- 3:52That is not what we are going to use.
- 3:54What we are going to use is the Claude
- 3:55Code section. So you are just going to
- 3:57click there. OK. So these are the
- 4:00different projects I have worked on.
- 4:02The first thing to do when in Claude
- 4:04Code is to provide it with the working
- 4:06directory. It is a very simple thing to
- 4:08do. You are simply going to click here
- 4:10on the folder. OK. You go to your work
- 4:13folder, cloud tuto. I have the sales
- 4:17prospecting, and I will provide it with
- 4:19the Claude Project directory. I click
- 4:22on select a folder. OK. So here, we
- 4:26have just provided Claude with access
- 4:28to our working directory. To launch the
- 4:30creation of the AI agent, what we will
- 4:32do is very simple. We are going to copy
- 4:35this prompt file. We go into Claude
- 4:37Code and we are going to paste it there
- 4:39. OK, we have our prompt file here. We
- 4:43are just going to tell Claude, execute
- 4:47these instructions please. OK. So here,
- 4:50we have provided the prompt file to
- 4:53Claude. Knowing that in our working
- 4:57directory structure, we have the
- 4:59cloud.mmd file. OK. Which is right
- 5:03there. So Claude will be able to read
- 5:06it. The second super important thing to
- 5:08do is to choose the model. So here, we
- 5:10have different models. You can click
- 5:12right there on Opus 4.6. You have
- 5:14different models there. Uh, Fable 5.1,
- 5:17which is currently the most
- 5:18high-performance model from Anthropic.
- 5:21We have Opus 5, we have Sonnet, we have
- 5:23Haiku, and you can click here for more
- 5:26models. So you have different models.
- 5:29Now, for really specialized tasks that
- 5:32will require coding, I recommend you
- 5:35use at least Opus 4.6, really Opus 4.6.
- 5:40And for the most relevant tasks, you
- 5:42can go up to Fable 5.1. But one thing
- 5:47to keep in mind is that Fable 5.1 is
- 5:49the most powerful model, but it is also
- 5:51the most expensive. So if you want good
- 5:55value for money, I recommend using Opus
- 5:584.7 or 4.8, for example. Okay? So for
- 6:02this video, I'll stick with Opus 4.6
- 6:04because we're working on a fairly
- 6:06simple task. We want to create an AI
- 6:08agent that will perform online company
- 6:10research. So I've chosen Opus 4.6 here.
- 6:14Once that's done, I can click on
- 6:17execute. And here, Claude will first
- 6:20try to gain access to the working
- 6:22directory. Then, it will read the
- 6:25various files. Okay, perfect. So now,
- 6:27it is iterating. You can see the
- 6:30iteration time is currently 48 seconds
- 6:32and the number of tokens consumed is
- 6:35184. Okay, so now it's in a reflection
- 6:38phase. We'll wait just a little bit.
- 6:41Okay, perfect. So Claude Code has
- 6:43completed a certain number of
- 6:44iterations. It has almost finalized the
- 6:47application creation, but it needs one
- 6:49important piece of information. You can
- 6:52see here, it's asking me a question. It
- 6:56says dependencies are installed. Before
- 6:59starting the server, you need to
- 7:01configure your Anthropic API key. Do
- 7:04you already have a .env.local file with
- 7:07your Anthropic API key? This is a very
- 7:11important step because since your
- 7:14application is supposed to search for
- 7:17companies online, you need an Anthropic
- 7:20key to perform this specific task. That
- 7:24is why Claude Code is asking if you
- 7:26already have a local environment file
- 7:28with the key provided. So we'll just
- 7:30tell it no, I don't have a .env.local
- 7:33file with my Anthropic key entered. I
- 7:36execute. It will then create this file
- 7:40itself. We will just need to input our
- 7:43API key. Okay. So now, it has created
- 7:46the local .env file and is simply
- 7:48asking me to enter my API key. And if
- 7:51you've never done it, it's very simple.
- 7:54You just go to Google and type in Cloud
- 7:57Platform. Click on the first link. OK,
- 8:04perfect. In your case, if you've never
- 8:07created an account in Cloud Console,
- 8:08you will simply need an email address
- 8:10to create an account. Once that's done,
- 8:14you can create an API key by clicking
- 8:17on get an API key. Here, in my specific
- 8:19case, I had already created an API key.
- 8:21So I can go to the API keys section and
- 8:24my API key appears there. So if I
- 8:27remember the key, I can copy it and
- 8:29come paste it into Cloud Code. And if
- 8:33not, I can simply create a new key. So,
- 8:37I remember my API key. I will therefore
- 8:40just come here and paste it. OK. So as
- 8:45soon as I provide my key to Cloud Code,
- 8:47it can continue creating the
- 8:49application. OK, Cloud Code has created
- 8:52the application, it is accessible at
- 8:55this URL. We will simply copy the link,
- 8:58go to Google, and paste it there. OK,
- 9:02so here we have our Neflori application
- 9:04. So here we can define our different
- 9:07criteria. We will, for example, put
- 9:09restaurant, the city, we will put Paris
- 9:11. And the number here, we will, for
- 9:13example, put one restaurant. OK, we
- 9:14launch the search. So when we click on
- 9:17search in the application, our AI agent
- 9:21begins to look for prospects. So here,
- 9:24in this instance, it will look for a
- 9:27restaurant in Paris, OK, that has a
- 9:30poor online presence. It will then see
- 9:33if it is possible to offer services to
- 9:36this restaurant. It will then generate
- 9:39an analysis report for us. So we will
- 9:41wait a bit for the search to be
- 9:43finalized. OK, perfect. The application
- 9:46has finished the search. It found a
- 9:49restaurant, in this case, the Suite
- 9:52Insulty. OK, so we'll go to view, and
- 9:57here we see a 4.6 rating, 692 reviews,
- 10:02at € 30. OK. Is that what the
- 10:05application is telling us? So here, the
- 10:08application tells us 4.7 with 410
- 10:10reviews. So that is problematic. There
- 10:16is truly a discrepancy between what the
- 10:18application provided and the reality.
- 10:21OK. And that is a major limitation
- 10:24because the application does not
- 10:26actually use a specific real-time
- 10:29search API. The application used
- 10:32Anthropic's search API, but this API
- 10:35likely pulled from Claude’s training
- 10:38data, which led it to provide this
- 10:41incorrect information. To address this
- 10:45issue, we need to use specialized APIs,
- 10:48and in this case, we could use several
- 10:51of them. The Serper API, which allows
- 10:55for real-time search, and the Tavily
- 10:58API, which allows for searching and
- 11:01returning well-structured data. The
- 11:05Twitter X API and other specialized
- 11:07APIs that will allow us to perform
- 11:10instant searches to see if a restaurant
- 11:13has an online presence. Now, the
- 11:17problem with this approach is that for
- 11:19an application like this, you would
- 11:22typically need at least 5 different
- 11:24APIs to ensure the most optimal search
- 11:27possible. And what you would have to do
- 11:31is go to each API's site, retrieve the
- 11:33documentation, understand how it works,
- 11:35and integrate it into the application.
- 11:38That is, therefore, the first
- 11:39constraint. If we need five API keys
- 11:42for our application, we have to go to
- 11:45five different platforms. You have to
- 11:48create accounts, generate API keys, and
- 11:50read five sets of documentation to
- 11:53understand how to call these different
- 11:56API keys in our application. The second
- 11:59major constraint is cost tracking. Five
- 12:03platforms means five different ways to
- 12:05track costs. And that can quickly turn
- 12:08into a mess in terms of cost monitoring
- 12:10. And personally, to address these two
- 12:13issues, the solution I use daily is the
- 12:16One API platform. Okay, you simply go
- 12:19to Google. Search. There you go, hit
- 12:25enter, and open the first link. Okay,
- 12:29so I’m going to quickly introduce you
- 12:30to this platform. It is a platform that
- 12:34aggregates various API keys so that,
- 12:37when you develop your application, you
- 12:40only need to provide a single API key
- 12:43—your only API key—and you provide
- 12:46that API key to your application. And
- 12:51this API key is connected to a set of
- 12:53tools, namely APIs (there are over
- 12:565,000) connected, tools (there are over
- 12:5940), LLM models (there are over 110),
- 13:01and payment methods. So, to enrich my
- 13:06application, what am I going to do?
- 13:09Since it is currently limited by the
- 13:12use of the Anthropic API key which can
- 13:14only use web search, which is very
- 13:17limited, what we are going to do is
- 13:19generate an API key for API EA, which
- 13:22is automatically connected to a set of
- 13:25APIs. I can show you here the various
- 13:29APIs that are aggregated. There are
- 13:31more than 5,000 of them. For example,
- 13:33we have Twitter here, which is super
- 13:35interesting when checking an company's
- 13:37presence. We have Reddit, super
- 13:38interesting when doing information
- 13:40research. We have YouTube, we have
- 13:41Tavily search. That is for retrieving
- 13:44information from the internet in a
- 13:46structured way. We have financial
- 13:49information, we have Instagram to know
- 13:50if a company has a presence on
- 13:52Instagram. Uh, we call it Perplexity
- 13:55for in-depth research. So personally,
- 13:58AISA is really a platform that allows
- 14:01me to overcome this major issue of
- 14:03online research and also the multiple
- 14:06use of different API keys. So to use it
- 14:10, it's very simple. OK? The first thing
- 14:13to do is simply create an account. OK?
- 14:15You click there on get started. I was
- 14:17already connected. So it automatically
- 14:19opens a dashboard. The first thing to
- 14:22do is to create an API key. So you will
- 14:25click there on new API key. OK. So I go
- 14:29to API keys and I can see there I had
- 14:32already created an API key. OK, so once
- 14:35you create your API key, how are you
- 14:38concretely going to use this platform?
- 14:42You go to get started. OK, here you
- 14:45have the three steps to follow. Really
- 14:47very simple. Personally, I find this
- 14:49platform to be a real game changer for
- 14:52people who want to build tools that use
- 14:54different API keys from different
- 14:56platforms. It is really super efficient
- 14:59for tracking. Everything is centralized
- 15:02with a single API key and it is just as
- 15:05interesting for cost tracking. So here,
- 15:09you will simply go to your usage log
- 15:10and you can see here the costs and
- 15:12token usage. No need to look at five
- 15:15platforms at the same time to be able
- 15:16to evaluate your costs. So you will
- 15:18come here to get started. OK, I will
- 15:21copy this command and I'm going back
- 15:23into Cloud Code. I will ask it to
- 15:26enrich my application using Ace One.
- 15:29enrich my application using Ace One and
- 15:38particularly use two tools, namely
- 15:42DataForSEO and Apollo. DataForSEO will
- 15:48allow for a complete search of the
- 15:50internet based on certain keywords and
- 15:53Apollo will allow for finding leads. So
- 15:56we are going to specify that. Data, for
- 16:03and Apollo please. OK, I'll put a dot
- 16:08and paste the previous command. Once
- 16:11that's done, I'll execute it. I'm
- 16:15running it and waiting. Cloud code is
- 16:19now getting to work. And there, it
- 16:21clearly tells me, I will first save the
- 16:24new key and read the documentation. to
- 16:28understand how to use DataForSEO and
- 16:31Apollo via Eisa. So we wait. It has
- 16:35retrieved the doc, it is currently
- 16:38retrieving the doc. So it tells me I
- 16:41have a clear picture of the available
- 16:43APIs. DataForSEO for business listings.
- 16:46But I'm not localizing well, but Apollo
- 16:48is excellent for prospecting. Here is
- 16:50my integration plan. Apollo mixed
- 16:53company search. Find companies by
- 16:56sector plus location. Apollo
- 16:59Organizations Enrich to enrich each
- 17:02company. Description, headcount, phone,
- 17:05network. Perplexity sonar existing.
- 17:08Let's find Google ratings, reviews,
- 17:10additional info. Tavily search existing
- 17:14web presence, Cloud, which will do the
- 17:17analysis plus email generation. So now
- 17:20it's starting the implementation. OK,
- 17:22perfect. So the other code has
- 17:23finalized. the registration of the
- 17:26application, it tells me the Apollo
- 17:28integration is working perfectly. OK,
- 17:31so we're going to test this. We'll go
- 17:34to Google within the application. We're
- 17:37going to redo the search. We'll put in
- 17:40a restaurant in the city of Paris.
- 17:43We'll look for one. We'll launch the
- 17:46search. OK. So what is it telling me?
- 17:49Company search with Apollo plus
- 17:51Perplexity. So Apollo is a really
- 17:55powerful tool when you need to find
- 17:57leads on the internet. So it tells me
- 18:01searching web presence using Tavily,
- 18:03analyzing five websites. OK. OK.
- 18:07Perfect. The search is finished. So the
- 18:11application preselected a certain
- 18:14number of companies and finally
- 18:17selected the restaurant Crocus Paris.
- 18:21OK. So the restaurant Crocus Paris.
- 18:24Here we have the address, the phone
- 18:27number, we have the website. Uh, so the
- 18:31priority is medium. So in terms of
- 18:34findings, the site is built on the Wix
- 18:36platform. OK. And the email sent. Hello
- 18:41, I am contacting you regarding the
- 18:43restaurant Crocus located on Rue Didot
- 18:46in the 14th arrondissement. While
- 18:49checking your online presence, I
- 18:51noticed that your current site is
- 18:53hosted on a generalist platform, Wix,
- 18:55which often limits customization and
- 18:58can provide a less distinctive image
- 19:00than a bistronomic and Mediterranean
- 19:03cuisine like yours deserves. At Nexi,
- 19:06we support independent restaurants with
- 19:08the creation of custom websites
- 19:10designed to reflect their culinary
- 19:12identity and facilitate customer
- 19:14contact. Menu, ambiance, reservation.
- 19:18Would you be available for a quick chat
- 19:20in 15 minutes to discuss what could be
- 19:22envisioned for Crocus? Best regards,
- 19:25Nexi, and my email address. So that is
- 19:27perfect. So here, typically, I can copy
- 19:30the email by clicking here, copy the
- 19:34email, and go to any email tool to send
- 19:37my message. With a tool like this,
- 19:40frankly, you can actually automate your
- 19:44client search. And here, I started with
- 19:47a tool where I have three parameters.
- 19:50The industry sector, the city, and the
- 19:52number of leads. You can add as many
- 19:55criteria as you wish, for example, the
- 19:57number of company employees, revenue,
- 20:00etc. I assure you, with such a tool,
- 20:02you can truly automate your prospecting
- 20:04. So, rather than spending hours every
- 20:07day manually searching for clients on
- 20:08the internet, you launch your tool. The
- 20:11tool loads clients for you
- 20:13automatically, and you can go further
- 20:15by connecting an automated email
- 20:17sending system. So it is truly a game
- 20:19changer. So definitely do not hesitate
- 20:23to test it on your end. I will put all
- 20:26the resources in the description that
- 20:28will allow you to reproduce this AI
- 20:30prospecting agent. I assure you, once
- 20:33you have managed to develop such an
- 20:35agent, you will have the ability to
- 20:37automate your client search. Now,
- 20:41speaking of the different resources, I
- 20:43will send you the Claude MD file, okay?
- 20:47Which represents the memory of your
- 20:48project. Then, the Prompt Claude Code
- 20:51file, okay, which will allow you to
- 20:53specify your instructions to let Claude
- 20:56know exactly what it needs to do. So,
- 20:59there is the possibility to customize
- 21:02these two files based on your company
- 21:04and its characteristics, and also what
- 21:07you wish to use for the prospecting
- 21:10application. In addition to that, in
- 21:13the description, I will also add the
- 21:15URL for the AI One platform. So, to use
- 21:18it, it is very simple. You simply type
- 21:20EASA. Do not hesitate to give me
- 21:25feedback on your use of this platform.
- 21:28It aggregates thousands of APIs. So you
- 21:30can truly use it for all types of
- 21:33projects. And what is really very
- 21:35practical with this platform, besides
- 21:37the fact that it aggregates several
- 21:39thousand APIs, is that you have the
- 21:41possibility to track your various costs
- 21:43on a single platform. So if I go here
- 21:47into my dashboard, okay, you can see
- 21:50the number of API keys I have, the
- 21:52requests, and the expenses I made today
- 21:55. Okay? So for the search I did earlier
- 21:59, it cost me 24 cents. Okay. So if I go
- 22:03into usage logs, I can see here for
- 22:05each of the APIs used in my application
- 22:08how much it cost me. This is a real
- 22:12game changer because it will actually
- 22:15allow me to compare different APIs. If
- 22:19I find that one API is too expensive
- 22:21compared to another, I can therefore
- 22:23use the ones that are cheaper. So here
- 22:26for example, I can see for the search
- 22:28from earlier, here are all the APIs
- 22:30that were launched. So here, I can see
- 22:32right here tav extract, how much did
- 22:34that cost me? I go all the way to the
- 22:37right, it cost me 0.054. OK. Uh, Tavily
- 22:41search 0.03. So this is super
- 22:46interesting if you are a business and
- 22:48you want to track your various costs in
- 22:50an ultra-precise way because for a
- 22:52company in the service sector or in the
- 22:54SaaS industry, you need to be able to
- 22:57save every penny. So, being able to
- 23:00track the different APIs you use for
- 23:02your applications on a single platform
- 23:04is a real game changer. If you found
- 23:07the video useful, don't forget to leave
- 23:09a like, subscribe so you don't miss the
- 23:11next tutorials on Claude Code and agent
- 23:13creation. See you very soon for a new
- 23:16video. This was Etel. Peace out.
About this transcript
This page contains the full transcript of Claude Code + AIsa : j’ai créé un agent IA qui automatise ma prospection commerciale by Ethel Cédric, generated from the public captions YouTube serves with the video. The transcript has 3,468 words across 510 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.