Cases

Case 01 · AI agent

Product Compliance Agent

The agent answers questions about regulatory requirements, and it takes every answer from real documents. Right now the document set covers the laws of Spain, Italy and Germany, plus an EU-wide layer on data minimisation.

Every answer comes with a link to the exact version of the document and the exact section. If the database has nothing for a particular country and situation, the agent says "this is not in the document set" instead of writing a believable answer. This is built into its rules, and it cannot get around them.

What business problem this solves

Take a guest data management system for hotels. One of its features is legal compliance: collect guest data and send it to the police. "Collect and send" means something different in every country. One country asks for one set of fields, another asks for a different set. Some require an electronic signature, some do not.

The most expensive part is proving why. The police come to a hotel and issue a fine: you collected too much data, the passport number is enough, you do not need to ask for citizenship. The hotel goes to the software provider. Someone there sits down and collects the paragraphs: another part of the same law says you must not collect data on the country's own citizens, only on foreign guests — and to find out whether a guest is foreign, you have to ask for citizenship. Everyone calms down. The next day the same thing happens with another hotel and another country.

Each case like this costs hours of work from a person who keeps the whole picture in their head. And it is the same person for the whole company.

The other half of the problem is sales and support. The product is large: features, integrations, specifications, the laws of several countries. Even when all the documentation is written and available to everyone, colleagues still go to that one expert, because a person finds the right paragraph faster. In a client meeting the company pays for this with its reputation: the client asks about the integration with service X, and the salesperson asks what that is.

The same kind of bot could work here, built on the same principle but with a different document set: product documentation, specifications, integration descriptions, pricing and release history instead of legislation. The rule about sources stays the same — the answer says which document it came from, and when there is nothing in the document set, the bot says so. Sales people need this most of all: a confident wrong answer to a client costs more than "let me check and come back to you". Two main scenarios:

• I am getting ready for a client meeting: collect what could be relevant for this client and explain those services.
• A request came in from a client: check whether we cover this or not.

What could be improved

A database built and maintained by professionals, instead of a web search. Even the most expensive models with web search build an answer out of random pages from the search results, and for regulatory work that is not good enough. What is needed is a document set with document versions and the dates they came into force, with a real lawyer responsible for the content.

Query logs. If you save what people ask about, after a month you can see what is unclear to users, where they get stuck again and again, and which parts of the documentation are missing. On top of the main product this gives you a second one: analytics for the person who is responsible for sales training and for the documentation itself.

I did something similar in the early days of GPT assistants — a sales assistant chatbot for online shops. It worked, but the token costs were too high, so we did not launch it. Since then the economics have changed and the problem is still there. I would like to keep working on tasks that sit between business goals, content and AI.

Case 02 · AI workflow

AI content factory: a content production pipeline

A process for producing text for a brand: 15 steps, 4 checkpoints, 4 types of participant, and one parallel branch for social media. The scheme works for different industries and formats — articles, reference pages, product pages.

Why it is needed

AI writes readable text quickly and cheaply. It invents a believable fact just as confidently. So the pipeline is built around a chain of checks: several checks by different models, always from different providers, human approval at key points, and a final read by an editor. Every human decision is saved with the name of the person and the date. If there is no evidence for a claim, the work goes back.

What is built in as ready-made skills

An SEO/GEO brief generator. It goes through the top search results, finds the patterns that repeat — structure, search intent, level of detail, images, FAQ blocks — and turns them into a working brief.

Removal of AI writing traces. It cleans out the words and sentence patterns that make a text easy to recognise as model-written.

A tone of voice check. It compares a draft with the editorial guidelines, marks problems with tone and clarity, and suggests fixes before publication.

The same idea, already in production

I have already done something like this on a real project, and it made the work about 10 times faster. It was a Claude workflow for affiliate content: one skill picks partner products that fit the topic of the article and ranks them by the size of the bonus our company gets when someone buys through the affiliate link. Then it pulls the referral links and selects images by visual criteria — clean background, horizontal format, little visual noise. After that come the draft, the editing and the layout. A full article package took half a day instead of several days: text, links, images, proofreading and layout.

Case 03 · Content strategy

Content strategy for REES46, built from scratch

A B2B marketing automation platform for e-commerce, with 40+ tools for retention and repeat sales.

I joined when content was produced ad hoc: texts were written whenever someone asked. I built a system — an editorial calendar, channel logic, briefs, repurposing rules, and planning around sales funnel stages and SEO priorities.

I also rebuilt the way we worked with case studies. This is a classic B2B problem: the client will not give you numbers, or the numbers are not impressive, so the case study never gets written at all. I moved the focus from metrics to how well the problem was described, the expert context and the practical advice a reader could use — and stories without strong numbers started to work.

Skolkovo

Skolkovo is Russia's main technology park, and resident status there is a sign that a tech company is taken seriously. I prepared the pitch, the presentation and the supporting materials for the application. The company got resident status. This is a rare case in content where the result is binary and easy to check: either you are accepted or you are not.

A repurposing system

One case study, one expert interview or one product story turns into an article, social posts, a newsletter, slides, a lead magnet, an angle for an external publication and a sales support document.

A real example is the launch of Stories for online shops. We had to tell clients about a new tool they did not know about, and answer the objections we had already heard from them: "Stories are hard to design", "you need a separate team for that", "it is an Instagram thing, I do not see how you sell products through those circles". We made a case study showing what the tool can do, two blog articles, posts on the company's social media and on the CEO's, a newsletter, and a lead magnet based on the same texts.

Lead magnets and lead generation

A library of checklists, guides and research: a map of trigger email scenarios, a checklist for reviewing on-site search, a checklist for testing hypotheses, a white paper on product recommendations. We handed them out by QR code at CEO talks and trade shows, and sent them to decision makers at target companies. At one trade show we collected 42 leads in a minute and another 200 over two days. I later repackaged every lead magnet into an article and social posts.

Social selling for company leaders

I ran the presence of the CEO and other experts on LinkedIn and Facebook: ideas, post structure, presentations. I also prepared talks and webinars end to end — topic, structure, script, slides, lead magnets.

Case 04 · Employee communications

Employee communications at large companies

Twelve years of work with corporate media at large companies — what is now called employer branding and internal comms.

Relaunch of the Cherkizovo Group magazine

Cherkizovo is one of the largest meat producers in Russia and a vertically integrated agricultural group: its own feed mills, poultry farms, pig farms, meat processing and distribution. Tens of thousands of employees, a public company, production sites all over the country.

That is where the editorial task came from. Before, it was a basic factory newsletter. It became a colour magazine of 28 pages, four issues a year, where the news of every business division lives together in one issue. A poultry farm in one region, a meat plant in another and the head office are three audiences with different language and different ideas about what counts as news at all.

The magazine had to read like one publication about one company, not like three brochures under one cover.

Rebranding of the PepsiCo corporate newspaper

I rebuilt the newspaper of the Russian PepsiCo division around the updated brand book, mission and values — both visually and editorially.

Gazprom Neft

One of the largest Russian oil companies. I wrote for internal communications: publications in corporate media and material for the company's internal messengers.

Besides that — corporate publications for Rosseti (the national power grid operator), Dobroflot (a fishing group in the Russian Far East), Transoil (rail transport of oil products), and also for OBI and VTB. Plus gift books, brochures, calendars and texts for a book about the museums of the "Russia — My History" historical parks.

Case 05 · Point of view

What content can solve

I believe content helps you:

• sell directly and indirectly — through texts, presentations and talks;
• collect data and understand what the audience is really interested in;
• reduce the load on the call centre;
• work better with different audience segments, speaking to each one in its own language;
• justify a high price — for example, with an educational webinar run together with an outside expert or partner;
• localise marketing materials for different countries.