How we use AI in our work: tools of web studio 12ia
You don’t have to read any more articles about “AI is changing the world.” The world has already changed. The question is: who benefits from this right now, and who continues to do everything manually. We at the 12ia web studio have switched to the side of the former – and now we will show what exactly this means. Not a theory, not a forecast. Tools that work for us every day.
Claude Code: AI agent that writes code
The main tool that changed our development is Claude Code. This is not auto-completion in the editor and not a chat that needs to be explained from scratch. This is a full-fledged AI agent that works with the project code base: it sees files, understands the structure, makes changes, runs commands.
What Claude Code does in our studio:
- Writes and refactors code for websites and web applications
- Configures server infrastructure – from nginx configs to deployment
- Creates integrations between services via API
- Analyzes bugs: reads logs, finds the cause, suggests and applies a fix
- Generates SQL queries, parsers, data migration scripts
The speed of development has increased by 3–4 times. Not because the code has become worse or simpler. Because AI in web development takes care of the routine: template code, standard integrations, documentation, tests. The developer deals with architecture and non-standard solutions—those where human experience is needed.
Important point: Claude Code does not replace the developer. He works as a junior specialist with perfect memory and zero time for development. He does not need to delve into the project – he is already in the context.
AI chatbots for our clients’ clients
The second area is development of AI bots for business. We make chatbots that work in Telegram, WhatsApp and on the website. But not those bots that respond using the script “press 1 if you want to know the price.” And those who understand free speech and conduct meaningful dialogue.
The difference is fundamental. A classic bot is a decision tree. An AI bot is an employee who knows everything about the product (learn how an AI bot differs from a chatbot), doesn’t get tired, isn’t rude, and works at 3 a.m.
Real problems that our AI bots solve:
- Consultation on assortment and prices – the bot knows the catalog and answers questions in free form
- Lead qualification – the bot finds out the client’s needs and sends the manager a “warm” application with context
- First line technical support – a bot solves standard issues and translates non-standard ones to a human
- Registration for services – the bot checks available slots and books time
One bot replaces 1–2 operators on incoming requests. See a concrete example: the AI bot we built for a customs broker. Not “theoretically it can replace” – it actually replaces, with measurable results in terms of conversion and response speed. Artificial intelligence for business is not about the future, but about specific savings right now.
n8n: automation that glues everything together
AI tools don’t work in a vacuum. They need a connection with CRM, instant messengers, databases, mail, telephony. For this we use n8n – an open-source platform for workflow automation.
n8n is a visual business process designer. But unlike Zapier or Make, it is deployed on our server. Customer data does not go to third parties, there are no limits on the number of transactions, and we fully control the logic.
Typical scenarios we build on n8n:
- Request from the website → AI processing → creating a deal in CRM → notification to the manager
- New order → document generation → sending to client → payment control
- Missed call → AI bot calls back → qualifies → transfers to manager
- Data from Google Analytics → AI analysis → weekly report to the manager
n8n + AI is a combination that turns business process automation from a set of individual scripts into a single system. Processes are not just automated – they are smart. We cover the most impactful use cases in 5 business tasks to automate. The system makes decisions based on context rather than rigid rules.
AI for content: blog, SEO, texts
We are also writing this article using AI. Honestly. But “with the help” is the key word. AI generates a draft, structures the information, and selects wording. The final edition, facts and tone are controlled by a person.
What we automated in content production:
- SEO analysis of competitors and selection of topics for articles
- Generation of structure and drafts of articles
- Writing meta tags, descriptions, snippets
- Adaptation of texts for different platforms – website, social networks, newsletters
Result: instead of one article per week, we publish three or four. Without loss of quality, because AI takes care of the mechanical part, and the expertise remains with the author. Content production is no longer a bottleneck.
AI analytics: data that no one has looked at before
Every business has data. Metrics, CRM, sales tables, call logs. The problem is not the data – the problem is that no one has time to analyze it. A marketer looks at the dashboard once a week and sees “traffic has increased.” Why did you grow up, where did you come from, and what to do with it – this takes time, which you don’t have.
AI tools for the studio solve this problem differently. We set up automatic data collection from several sources, AI analyzes it and makes specific recommendations. Not graphs, but text: “form conversion on page X dropped by 30% after the March 15 update – likely cause: field reorder.”
It’s not magic. This is pattern matching on large amounts of data, which a person physically cannot do on a daily basis. But AI can.
Voice bots: Yandex SpeechKit + AI
A separate story is AI voice bots. We use Yandex SpeechKit for speech recognition and synthesis, and AI models for understanding context and generating responses. The result is a bot that talks on the phone almost like a human.
Where it works:
- Calling the customer base – appointment reminder, order confirmation, collecting feedback
- Incoming line – the bot accepts the call, answers questions, transfers to a specialist if necessary
- Qualification by phone – the bot conducts a short interview and transmits the result to the manager
A voice bot is several times cheaper than a call center. And, more importantly, it scales: 10 simultaneous calls or 100 – it doesn’t matter. There is no need to hire people.
What does this add up to
Each tool individually is an optimization of one process. Together is a different level of studio work. We don’t just “use AI” for marketing buzzwords. We are working on it. Daily.
Specific numbers:
- Development speed – x3–x4 due to Claude Code
- Content production – x3 due to AI generation
- The cost of customer support is minus 50–70% due to AI bots
- Time for routine tasks – minus 60% due to n8n automation
And here’s what’s interesting: this is not the final point. The tools are getting better every month. What six months ago required complex crutches now works out of the box. How AI helps in work is a question whose answer becomes outdated faster than we can write about it.
We continue to test new tools, break old processes and put them together into something that works faster. If you want to implement something described in your business – start with an AI bot or process automation. These are the fastest entry points with measurable results.