At 19, Teymur Kutseba turned down investment in the US. Years later, he and his partners launched InnMedia in Azerbaijan, a start-up aiming to fully automate content marketing and newsroom operations.
For the Digital Business and Astana Hub project “100 Startup Stories from Central Eurasia”, Teymur spoke about how the platform works, why start-ups don’t need their own language models, how to create content for AI search, and how InnMedia plans to enter new markets.
“I could’ve been working in Silicon Valley at 19”
– Tell us about yourself. What did you do before InnMedia?
– I graduated from the Azerbaijan State Oil and Industry University in 2014 with a degree in biomedical engineering. I later earned an MBA and studied big data visualization online at the University of Waikato in New Zealand. But my career has always been at the intersection of technology, sales, marketing, and management.
I started out in sales at BeSmart, an online discounts and coupons platform. I then launched the Lalafo marketplace in Azerbaijan and worked as an analyst at Carlsberg.
For several years, I worked in fundraising, marketing, and digital products across different parts of the IFRC, the International Federation of Red Cross and Red Crescent Societies. My decision to join was more emotional than calculated, as the COVID-19 pandemic and the conflict in the region happened at the same time.
Alongside my day job, I was also building start-ups and businesses of my own. In fact, I gave my first pitch back in 2012 at Google’s office in Silicon Valley.
– What kind of start-ups were they, and how did you end up at Google?
– I came up with a way to charge a power bank without plugging it in, using the kinetic energy produced while you walk. The idea won several start-up competitions in Azerbaijan, and that led to an internship in the US. We visited Oracle and MIT, but spent most of our time at Google’s office in Silicon Valley. That’s where I gave my first pitch in English to a room full of investors.
– So that start-up never really took off?
– An American hardware company showed an interest in the project. They offered to move me to the US, cover my housing, relocation, and technical costs, and invest in developing the product. They wanted to buy the project outright and keep me on as CEO.
But like most first-time founders, I thought far too highly of my idea. I wanted millions of dollars, and they weren’t offering that. So I said I’d rather put the blueprints online than work for peanuts.
Looking back, I realise what an opportunity I missed. I could’ve been working in Silicon Valley at just 19. But I was a bit spoilt back then. Because of rowing, I was flying somewhere every month and a half and had already travelled across half the world. I’m a seven-time Azerbaijani champion, a European junior champion, and a World Championship finalist, so moving to the US just didn’t feel like such a big deal at the time.
“A journalist should still be a journalist, not a prompt-engineering expert”
– How did you come up with the idea for InnMedia?
– I started a marketing agency in 2020, and that’s when I noticed how absurd the situation was. Media outlets around the world were still relying on outdated workflows. Producing a single piece of content involved a whole chain of tools and specialists: one person researched the story, another checked the facts, a third wrote it, a fourth translated it, and then a designer created the visuals, a camera operator shot the video, an editor put it together, and so on. In the end, more time went into passing information from one person to the next than into the creative work itself.
And I thought: if car manufacturing, logistics, and financial operations can all be automated, why is content production still largely manual?
That’s how the idea for InnMedia came about in 2023.
– So, what does InnMedia actually do, and what problem is it trying to solve?
– A company might use one AI model for text, another for images, another for video, and another for voiceovers. Then there’s a CMS for publishing, an ERP system for storing documents, a CRM for managing clients, cloud storage for data, and yet more tools for analytics and distribution. Before long, you end up with a whole jumble of disconnected systems that are difficult to manage.
InnMedia pulls all of that into one place. It can spot a story, gather and analyse the information, check the sources, draft the content, create different text versions, images, audio, and video, tailor everything for different channels and languages, send it to a human for approval, and then publish it.
So the main problem we solve is the operational complexity of content production. Everything happens in a single dashboard. It also makes life easier for the finance team: instead of dealing with a dozen bills for different subscriptions, they just pay one monthly invoice.
– So, who actually needs a system like this?
– We originally built the product for media companies and publishers. But we quickly realised the potential audience was much broader: any organisation where content is an ongoing part of day-to-day operations. That includes newsrooms, communications agencies, marketing teams at large companies, super apps, government and quasi-government organisations, and others. Our product is aimed at medium-sized and large businesses. A small coffee shop that only needs two Instagram posts a week simply doesn’t need a system like ours. It would be like driving a tank to the supermarket.
– How can companies use InnMedia?
– Take retail, for example. Say a company adds several thousand new products to its catalogue. Each one needs a description, a product page, visuals, ad copy, short videos, and so on. InnMedia can take the raw product data and create all of that content according to the company’s own rules.
For advertising agencies working with several brands, InnMedia can create content that follows each brand’s tone of voice, visual guidelines, product specifics, and approval process.
For government bodies, InnMedia can take a formal document and turn it into a press release, an FAQ, social posts, an infographic, an audio version, and content in several languages.
– So what does the human actually do in all of this? Write prompts?
– We built the system so people could focus on the work they were hired to do. A journalist should be a journalist, not a prompt-engineering specialist. Users don’t chat to the AI or write prompts themselves. The system handles the prompts and AI workflows for them. They simply open the platform, review the content that’s already been prepared, make any changes, and approve it.
To make that work, though, the system first has to be tailored to the company. We gather its internal guidelines and learn its tone of voice, publishing requirements, data sources, and approval process. We then use that information to configure the AI models, data storage, processing workflows, and security settings.
We also plug in the information sources each organisation relies on. For a newsroom, that could be feeds from major news outlets, government websites, and legal databases. For a business, it might be internal documents, product catalogues, a CRM, or other company systems. The platform keeps an eye on those sources around the clock and starts the right workflow automatically. So when a major story breaks, it can prepare a draft for an editor. When a new product is added, it can create the description, visuals, and copy for each channel.
– But how good is the content in practice?
– The first and most important stage is fact-checking, which involves ten separate steps. The AI analyses the information and checks it against other sources and social media. If the details hold up, the story is marked as verified. If they don’t, it’s filtered out.
Only then does the system start generating the content. The text is created directly in the required languages and follows the company’s editorial policy, tone, and approved vocabulary. At the same stage, it is also optimized for visibility in AI-generated search results through GEO.
– Why are you focusing specifically on GEO?
– The way people consume content has changed. SEO is dead, organic traffic is falling, and that decline will continue as more people get their information directly from AI-generated answers. A lot of companies are trying to retrofit existing content for AI search, but that doesn’t really work. GEO has to be built into the text from the start. Here’s why: AI systems are designed to answer questions using as little computing power and as few tokens as possible. If a piece of content isn’t structured in a way the model can easily recognise and process, it may simply skip over it. At InnMedia, we generate content that naturally aligns with the patterns LLMs were trained on, making it more likely to appear in AI-generated results.
“There’s no commercial sense in building your own AI models”
– Who else is working on InnMedia?
– InnMedia has two other co-founders besides me, but they prefer to stay out of the public eye. We bring different areas of expertise to the table, which means we can cover the main sides of the business: one co-founder has 15 years of experience in IT, while the other is a finance professional from the oil industry.
We have a core team of 10, and during busy periods we bring in freelancers, taking the team up to around 20 people.
– Which LLMs do you use?
– We use models from the biggest names in AI, including OpenAI, Google, xAI, Anthropic, and DeepSeek. But InnMedia’s real value is in how those models work together: which one handles each task, when it comes in, what context it gets, and which sources it works from. We’re currently filing for an international patent in several Central Asian countries for the technology behind this. It’s an AI-orchestrated content production and processing pipeline built on Microsoft Azure, and the patent will help protect it from being copied.
– Why did you decide to build the product on existing LLMs rather than develop your own?
– Why would we? Building our own model would take tens of millions of dollars, a team of around 30 highly specialised experts, and several years of work. More importantly, it just doesn’t make commercial sense when the industry is moving this fast.
We use the best models available. To put it in car terms, the LLM is the engine, built by one of the world’s leading manufacturers. As that engine gets better, so does our car, without us having to spend money developing it ourselves.
– What was the hardest part of building it?
– Getting AI to write articles wasn’t the difficult part. The real challenge was building a system that works reliably in practice. How do you manage sources? Stop the same content being created twice? Handle several languages at once? Keep everything secure? What happens if one of the services goes down? And where does a person need to approve something? Those were the problems that took the most work to solve.
The other big challenge had nothing to do with building the product. It was helping potential clients understand that adopting AI means more than giving a few employees ChatGPT accounts. We often had to spell out the difference between using a standalone AI tool and actually weaving AI into the way a company works every day.
“At an early-stage start-up, the founder needs to speak to customers personally”
– Who’s already using InnMedia?
– We’re currently working with both private companies and public-sector organisations. Our contracts include strict NDAs, so I can’t name the clients. What I can say is that they include a government body in the security sector, a major super app, several well-known media outlets, and B2B and B2C companies that use media channels to shape public opinion around their business.
We also run TECHXƏBƏR, where we test the technology in a real-world setting. We use it as a showcase for what the platform can do.
We’re also in talks with a chain of medical clinics in Kazakhstan, companies in the alcohol industry, a dating platform, and a major e-commerce marketplace. InnMedia can be adapted to each of them, whether that means producing news stories, product listings, or dating profiles.
We’ve run the numbers on real projects. With InnMedia, the cost per piece of content is 62 times lower, while production is 24.6 times faster. Overall content costs fall by 70%, starting from $2.69 for one multi-format content asset, and customer acquisition costs come out two to four times lower.
– If you’ve built a content factory, why not use it to take over the media market yourselves?
– You can build one huge factory, or you can make the equipment that powers thousands of factories around the world. We’ve chosen the second route because the global market for content production, management, and distribution infrastructure is far bigger than the media market itself.
– How did you find your first clients, and which channels do you use now?
– Our first leads came through our network and direct sales. At this stage, I think that’s the right way to do it. Founders need to speak to customers themselves. Otherwise, too many layers start creeping in between the problem people actually have and the product you’re building.
These days, we use pretty much every channel available. We reach out to decision-makers through LinkedIn and Instagram, ask mutual contacts for introductions, and go to relevant industry conferences. Referrals are starting to bring in new business too. We’ve recently had companies from Kyrgyzstan and Azerbaijan approach us after being referred by others.
We’re also building partnerships with major IT systems integrators and distributors, giving them licences to sell InnMedia in specific countries. We’re currently in talks with one of the biggest global names in the industry.
“The platform works around the clock. It doesn’t call in sick, and it doesn’t quit”
– What’s your monetisation model?
– There are a few layers to it. First, clients pay for integration and implementation. After that, there’s a recurring usage-based fee. We can also offer InnMedia as a white-label solution, so companies can use the platform under their own brand. Another option is to provide it as infrastructure, with individual InnMedia features built into someone else’s product. And for organisations with stricter security and data-control requirements, we can deploy the platform on their own servers.
– How much does it typically cost to use the platform each month?
– Pricing is worked out individually and depends on traffic, server load, and security requirements. On average, the platform does the work of a 10-person team but costs roughly the same as four or five salaries. And it runs 24/7, doesn’t call in sick, doesn’t quit, and doesn’t need a desk.
In practical terms, B2B pricing starts at $2,000 a month.
– What’s your revenue at the moment?
– I can’t share the exact figure, but we’re still in the early stages of commercialising and scaling the business. A large share of our revenue is being reinvested in the product, the infrastructure, and expansion into new markets. Right now, our main focus is growing quickly.
– Are you profitable yet?
– If you set aside the initial R&D costs, we’re almost at break-even on an operational basis. Another two or three large contracts should get us there. The original R&D investment hasn’t paid for itself yet, though, so it’ll take longer before we’ve fully recouped what we put in.
“The start-up will become a unicorn and be acquired by Google, Microsoft, Amazon, NVIDIA, or a global media group”
– How much have you invested in the start-up?
– We’ve invested more than $85,000 in total. Around $35,000 came from my own money, and the remaining $50,000 came from angel investors.
We’ve also secured substantial infrastructure support. InnMedia received $150,000 in Azure credits through Microsoft for Startups and $25,000 in Cloudflare credits through Cloudflare for Startups.
– Are you looking to raise more investment at the moment?
– For the next stage, we’re planning to raise around $250,000 through a SAFE, a Simple Agreement for Future Equity, where an investor puts money into the start-up in exchange for the right to receive shares at a later date. We need the funding to scale faster. Right now, the priority is moving quickly and gaining ground in new markets.
– Which markets are you targeting?
– We’re already operating in Azerbaijan and are now entering the Kazakh market. We’re also planning to expand into Uzbekistan, Kyrgyzstan, Georgia, and Turkey. These are markets we understand both culturally and operationally. To prepare for that expansion and build relationships within the local start-up ecosystem, we’re currently taking part in the seventh cohort of Silkway Accelerator, an international accelerator programme run by Astana Hub and Google for Startups.
We’re giving ourselves six months to establish a strong regional presence. After that, we’ll start expanding into the wider international market, with the EU, China, and the US as our ultimate targets.
– Who are your main competitors?
– Most companies only compete with us in one part of the content workflow. Jasper, for example, is building an AI platform and agents for marketing teams. Arc XP is adding AI assistants to editorial and publishing workflows, while Brightspot is developing AI features within its CMS and digital publishing platforms.
In Central Asia, we’re the only start-up operating in this space. At the moment, I don’t see any direct regional equivalent to what we’re building.
– Where do you see the start-up in a few years’ time?
– As a unicorn, acquired by Google, Microsoft, Amazon, NVIDIA, or a global news group. We already have a clear exit strategy and a strong sense of what could make the company attractive to a potential buyer.