How Neural Networks Transformed Content Marketing: Pasha Molyanov's Experience

Contents:

• MyMeet: My Replacement of a Transcriptor with a Neural Network
• Artificial Intelligence in Art: Replacing Illustrators with Generative Platforms
• ChatGPT and Similar LLMs: How Neural Networks Made an Agency's Work Half as Easy
• Cursor and Claude Code: Our Journey to Programming Without Prior Knowledge
• Podcastle: Advances in Creating High-Quality Sound Without Significant Production Costs
• Observed Transformations in Workflow Thanks to the Implementation of AI

Pasha Molyanov, founder of the content marketing agency "Sdelayem", as well as a club dedicated to neural networks, presented his knowledge and experience "Neuroworkshop."

Previously, I used to outsource various tasks. One of them transcribed interviews, another created illustrations, and a third proofread the texts. I no longer employ employees because I now have neural networks. Their use is proving more cost-effective and efficient in terms of speed.

In this publication for the Skillbox Media "Business" editorial team, I will share my experience implementing neural networks at my agency and describe how the use of artificial intelligence has significantly reduced my costs.

• Transcript of Meetings and Interviews
• Artificial neural networks, designed to create images, are a powerful tool in the field of artificial intelligence. These systems are capable of generating visual materials based on various tasks and source data. Using deep learning algorithms, they analyze vast sets of images, allowing them to capture stylistic features and details.

  These types of neural networks find application in a wide range of fields, including art, fashion, advertising, and even scientific research. They can be used to create unique pieces, as well as to enhance existing images or adapt them to specific user requirements. Technological advances in this field open new horizons for creativity and innovation, providing artists and designers with the tools to realize their boldest ideas.

  Thus, neural networks for image generation are becoming an essential component of the modern digital world, transforming approaches to the creation and perception of visual content.

• ChatGPT and other LLMs
• Artificial neural networks designed for programming allow you to create and edit code without the need for in-depth knowledge in this field. These tools significantly simplify the development process, enabling users to generate software solutions using simple commands or descriptions. Thanks to such technologies, even those without specialized programming skills can implement their ideas and create applications. Such systems use machine learning to analyze and generate code, making them useful for both beginners and experienced developers looking to speed up their work or get new ideas.
• AI Sound Engineer
• What transformations have I observed in my work?

MyMeet: My Replacement of a Transcriptor with a Neural Network

The first neural network I started using in my business was MyMeet.

This is a Russian platform offering services for automatic recording of phone calls and video conferences, as well as their transcription, analysis, and compilation of final reports. This system's bot can connect to online meetings on services like Zoom, Google Meet, and Teams and record both audio and video.

How we use this process: At our content marketing agency, we regularly communicate with clients: conduct interviews, refine briefs, and discuss materials and strategies. At the end of each phone call, we need to transcribe the conversation in order to create an article, post, newsletter, or case study based on this information.

Previously, we had a specialized position—a transcriptionist. This work was quite labor-intensive, which affected the agency's financial costs. However, with the advent of MyMeet, it became clear that manual transcription was no longer relevant. A neural network performs this task much faster, more cost-effectively, and with higher quality—its efficiency is a million times higher.

Now we can upload an audio file or connect the neural network to a meeting, and in just a few minutes we have a completed transcript. What used to take hours can now be done in just minutes.

In addition to conversations with clients, I also use MyMeet in meetings within teams. This tool allows me to capture agreements, tasks, and ideas. Sometimes after meetings, we use the MyMeet report as the basis for creating a postmeet—a short summary of our meeting.

Artificial Intelligence in Art: Replacing Illustrators with Generative Platforms

Midjourney is second in our everyday use. Then ChatGPT came on the scene, mastering image creation, followed by Flux, and then Nano Banana. Currently, everyone at the agency chooses their own tool: some continued working with Midjourney, others switched to Flux, and still others prefer Nano Banana.

These are neural network-based technologies capable of creating visual images based on text descriptions.

Midjourney is a pioneer in neural network technologies and has become a popular tool for creating images, collages, and illustrations based on given text descriptions. ChatGPT 4.0 now has the ability to generate visual content directly in conversations, while Flux demonstrates outstanding results in photorealism and meticulous detail control. We shouldn't forget about Google's Nano Banana, which is excellent at stylistic image adaptation.

All of these neural networks are designed for different tasks: Midjourney is ideal for developing artistic and conceptual images, Flux specializes in creating realistic shots, and Nano Banana is used to generate brand visuals.

How we use these resources. At our agency, we create texts, publications, websites, landing pages, and newsletters. Each of these formats invariably requires visual design: covers, graphics for landing pages, images for articles, infographics, and branded images for clients. Previously, all of these elements were created by illustrators. We had a team of several specialists who were paid significant sums for each illustration. They hand-drawn characters, created collages, comics, and banners.

With the advent of neural networks, the landscape has changed significantly. Illustrators are now in demand only for specific tasks that require careful handwork or exceptional quality for printed products. All other visual materials are created by content workers: article authors generate images for their materials themselves, and social media specialists create visuals for publications.

Where we previously paid 2,000 rubles to an illustrator, we can now do without the expense. And instead of waiting a whole day, everything is now completed in just 10 minutes.

This stage turned out to be the second significant change in our company after integrating MyMeet into our workflows. Artificial neural networks have radically rethought the use of visual content - they are no longer exclusively a tool for designers and have become available to the entire team.

ChatGPT and similar LLMs: how neural networks have made the agency's work half easier

The third neural network that we began to actively use was ChatGPT. Over time, we have also integrated other language models into our processes, such as Claude, Gemini, Grok, Perplexity, and DeepSeek.

These are artificial intelligence systems that can learn, analyze, and create text materials.

• ChatGPT is a multifunctional neural network that can create texts, generate ideas, develop scripts, and write code.
• Claude is ideal for learning and correcting long texts.
• Gemini demonstrates high efficiency in creating educational resources, conducting research, and processing multimodal data.
• Grok focuses on creating short materials, current trends, and content suitable for social media.
• Perplexity AI is an intelligent search engine that provides answers along with indication of sources and analytical data.
• DeepSeek is an effective model designed to organize data and conduct deep research.

How We use this. All of these neural networks have been integrated into virtually every aspect of our activities:

• Content Creation. We now generate some simple texts entirely using AI-powered models. Previously, such tasks required hiring and paying human writers, but now editors can generate the necessary material themselves. While talented writers still produce higher-quality texts, neural networks are quite effective for quickly writing simple materials.

• Research Support. Models based on large language models (LLM) significantly assist writers in fact discovery, collecting materials, and formulating interview questions. When an author approaches an expert and isn't sure how to approach the conversation, ChatGPT can help generate a list of well-thought-out questions and quickly master the subject matter.
• Plan and Structure Creation. When SEO-friendly text is needed, the neural network accesses Google, studies the top ten pages, and creates a structure based on competitor content. After this, the author works independently on the text, which significantly reduces the time spent on preparation.
• Creating text passages. Sometimes it is necessary to develop 20 different headings, suggest examples for an article, or provide a simple explanation of a term. Neural networks show excellent results in this regard. A human only needs to select or slightly refine the most successful options.
• Proofreading and analysis. Language models assist editors in the process of reviewing materials. It is possible to use a neural network to identify inconsistencies with the technical specifications, as well as to find repetitions, errors, or factual discrepancies. This significantly speeds up the proofreading process, especially when the volume of texts is large.
• Until recently, our company employed a team of proofreaders, and their salaries cost a significant amount of money each month. In September, we decided to replace half of them with an automated script that processes text using a neural network directly in Google Docs. This tool effectively corrects errors and typos, doing so quickly and cost-effectively.
• Previously, we used specialized agencies to translate texts for foreign clients. Currently, we write materials in Russian and use LLM for translation, and then an English-language editor simply proofreads and checks the meaning. This allows us to reduce the number of contractors.
• Support for managers. Project managers use the LLM as their assistant: they seek advice, research ideas for business proposals, develop formulations, create presentations, analyze financial models, and create job descriptions.

Read also:

Ways to generate ideas for marketers using ChatGPT: recommendations and methods

Cursor and Claude Code: Our Journey to Programming Without Prior Knowledge

I'm not a developer. I have an idea of ​​how things work, but I've never written code, been familiar with programming language syntax, run apps, or used GitHub. All of this was foreign to me—until Cursor and Claude Code came along.

These are neural network-based systems designed to create and execute programs that make programming possible even for those without deep knowledge of the field.

Cursor is an AI code editor that resembles the familiar VS Code editor, but has a unique feature: a built-in assistant. This assistant can independently generate and correct code, as well as create scripts, websites, and extensions.

Claude Code is a tool developed by Google that allows users to build and deploy projects in the cloud, as well as test and automate various processes directly from the browser.

Both services allow people without programming experience to easily create code - just formulate the task in words, and the neural network will do the rest.

How we use these tools. Thanks to them, several of my colleagues and I unexpectedly became programmers. Now we have the opportunity to automate various processes, develop internal services, launch new products, and even create extensions for Google Docs.

For example, our project, which previously had no programming experience, developed a proofreading bot for Google Docs. To reduce the cost of proofreading services, he used Cursor, created an extension, and now we have a neural network that automatically corrects errors in texts.

I work on similar projects. Currently, for example, I'm developing a tool for analyzing company mentions in search results. We load the company name into the neural network, which then runs a variety of queries, such as "recommend a good company in this field," "name the top ten in the industry," and "what can you say about this company." All the resulting data is processed and compiled into an analysis table, which indicates: "Your company is mentioned 67% of the time, ranking 4th on average, while competitors X and Y are ahead of you thanks to such-and-such mentions." The result is a nearly complete audit, and I was able to implement all of this without any programming experience.

Using Cursor, I've also developed several products that operate independently of the agency:

• Broken Link Checker is a tool for detecting broken links. It currently has close to 1,000 installs and attracts around 500 active users weekly. Additionally, the app is featured on Product Hunt and is featured as a "Google Featured" app. All aspects of the development were completed independently, without coding, using only a neural network.
• "Finspoved" is an independent media outlet operating within a Telegram channel. Users fill out questionnaires, after which the neural network generates posts, verifies them, and automatically posts them to Telegram. The entire process occurs without any human intervention.
• A Telegram bot called "Emoji Picture" can break an image into a combination of emoji. It accepts payments for its services and also has the ability to display ads.

Previously, such things would cost us thousands of rubles and require a lot of waiting time. Now I can build the tool myself, test it, and launch it - in just one evening.

Podcastle: Achievements in Creating High-Quality Audio Without Significant Production Costs

Another of the neural networks we use is Podcastle.

It's a platform designed for creating and editing podcasts, equipped with a neural network that allows you to improve sound quality.

The basic idea is that you can create a podcast or video on any available device - be it a smartphone, AirPods, inexpensive headphones, or a camera without a built-in microphone. After that, just upload the audio file to Podcastle. The neural network automatically processes the audio, eliminating background noise, removing crackling and artifacts, equalizing the volume, and adding spatiality.

How we use it. I started working with Podcastle last year. Now the process of recording videos, webinars, or podcasts has become a hundred times easier. Previously, to achieve good quality, you had to find equipment, go to a studio, and rent a space. Now, all you need is a phone and a pair of headphones—the neural network does the rest.

Of course, this doesn't meet the standards of a professional studio with a microphone costing 50,000 rubles and a sound engineer, but for recording at home or in the office, it's just perfect.

Observable transformations in the workflow thanks to the implementation of AI

The main thing that catches the eye is that a significant proportion of contractors have lost their relevance. What was previously performed by humans is now being successfully replaced by neural networks, which operate faster, more efficiently, and without interruption. This situation is not temporary, but rather a stable trend.

I am convinced that in the future, artificial intelligence will increasingly replace specialists, and agencies will be organized according to a different principle. They will still include projects that manage various neural networks, interact with clients, formulate tasks, and aggregate the results obtained using AI tools.

There will be specialized specialists who create unique products that require manual labor or a creative approach. While everything that can be classified as mass production and average quality is already being performed by artificial intelligence.

It never hurts to learn more about neural networks: we present to you several more materials from Skillbox Media dedicated to this topic.

• How artificial intelligence collided with content authors and failed: 5 real stories.
• The introduction of artificial intelligence into work processes made it possible to achieve impressive results.
• Job search using neural networks: recommendations and example queries with response images.
• How to effectively interact with neural networks: recommendations from experienced users who regularly communicate with artificial intelligence.
• Without causing a feeling of awkwardness, but on the contrary, creating a “wow” effect: how brands should develop advertising using artificial intelligence so that it is effective and does not irritate.