The Analytium founder spent years delivering data projects alongside enterprise players. With AIgentic, he is working on the premise that mid-sized firms need education before they need automation.

Vasilij Nevlev has been working with artificial intelligence since 2012, long before ChatGPT made the term a household word. Today he is founder of Analytium, a multi-million pound data analytics consultancy, and of AIgentic, a newer business focused on AI agents for SMEs. His story matters now because mid-sized businesses are keen to adopt AI, and many of them are not ready. 

From enterprise analytics to AI agents for SMEs 

Analytium was built as a classic consultancy, serving large organisations with data engineering and analytics work. Nevlev says the lesson that held across that period was adaptability. Corporate demands shift with technology and analyst recommendations, so a consultancy has to keep watching the horizon and be ready for what clients will ask for next. 

What clients ask for has changed. Requests used to centre on data solutions such as self-service BI. Now, Nevlev says, they tend to fall into two camps. Some are driven by fear of missing out and want to know where to start. Others are executives who believe their business can be transformed quickly and want to move ahead of competitors. 

His decision to target the mid-market came down to structure. Selling directly into enterprise is hard for a business of Analytium’s size: sales cycles are long, projects are large, and the firm previously delivered bigger work through partners such as Capgemini. Mid-sized projects, he argues, are too small for the large consultancies, which leaves room for a firm like his to deliver efficiently and still make a margin. He describes it as medium-sized businesses buying from medium-sized businesses. 

The readiness problem behind AI agents for SMEs 

The difficulty is readiness. Nevlev says some mid-sized companies ask for AI agents before they are clear on their business strategy, their route to market or their marketing approach. In his view, AI belongs in processes that are data-rich, repeatable and frequent, the parts of a business where many people already do the same tasks. 

He gives a simple example. If a company has five salespeople who each sell in a completely different way, there is no defined process to automate. Larger corporates have governance and quality controls in place because of their scale. Many SMEs, working with smaller teams, never had the resources to define their processes in the same way until now. 

He is also direct about risk. He has published material on cases where an AI-automated shared mailbox sent out confidential documents, and where AI platforms from newly launched startups ran production servers without any security protocols in place. 

For one charity client handling medical information, his team is bringing workflows and language models in-house so that data stays within the organisation’s own network. That approach costs more, he says, but it gives the client full control.

Why AI agents for SMEs start with people and data
 

Where he sees the most traction right now, Nevlev says, is employee training. When a business announces AI as a way to cut costs, staff can hear it as being asked to train their replacements. Failing to bring employees on board is, in his view, one of the reasons adoption is slow. 

The second challenge is older than ChatGPT. Before the current wave, he says, a typical data scientist would spend 80% of their time wrangling data before training a machine learning model. He says that has not changed. Businesses still have to integrate systems and make the right data available to the AI in a secure way. 

Building an audience before a sales pipeline 

Those constraints have reshaped how AIgentic operates. Nevlev says the business currently resembles a media company more than a consultancy, and its approach to AI agents for SMEs starts with education. It publishes short-form video across social platforms and longer content on YouTube and its training academy, including a masterclass covering basics such as tokens, context and hallucinations. One of its most viral pieces explained that choosing a different OpenAI model could mean a 120-fold difference in cost per token. 

Content decisions are data-led. The team researches search volume and competition before scripting, and Nevlev uses AI agents built on written customer personas to review scripts before production. Leads are scored as they move through the funnel, because the firm needs clients with some digital maturity in place. The YouTube channel was at around 1,270 subscribers at the time of speaking. 

He frames the approach as being useful first rather than selling first. He also admits he felt a little insecure appearing on camera at the start, until he realised viewers care about getting something useful, not about him. 

Internally, AIgentic was set up as AI-first from scratch, using tools such as Notion AI and Google Gemini. That meant consciously letting go of how Analytium had worked. 

Nevlev is open that AIgentic has not yet found product-market fit, and he believes no one has fully solved AI for SMEs. Over the next 12 to 18 months he expects to finish experimenting with face-to-face and online training, packaged agents and several niches, then focus on one or two and build teams to serve them. He credits what has kept him going to a pairing of long-term optimism and short-term paranoia.