The TBS AI Consulting Practice: How We Help Businesses in Southeast Europe Adopt AI
TBS Group operates an AI Consulting Practice that helps businesses across Southeast Europe adopt artificial intelligence in ways that create measurable impact — without over-engineering, over-spending, or over-promising. The practice launched on May 1st, 2026, and is led by Kaloyan Stoychev.
TBS (Telelink Business Services Group) is a publicly traded IT services company headquartered in Sofia, Bulgaria. We employ around 500 people across seven countries and have been delivering enterprise IT services since 2001.
About Telelink Business Services
TBS Group is organised into four practice areas:
- AI & Software Engineering — a 50+ person team working across AI, software, and data engineering, led by Pavlina Kekerova
- Advanced OT Solutions — operational technology for industrial and critical infrastructure environments
- IT Infrastructure & Security — networking, compute, storage, cybersecurity, and cloud platforms
- Business Applications — enterprise business systems and process platforms
The AI Consulting Practice is a dedicated capability inside our AI & Software Engineering practice. The broader practice has historically focused on building AI-enabled software, data platforms, and machine learning systems for clients. The new Consulting Practice extends that capability into the work that surrounds and precedes the build: strategy, process analysis, change management, governance, and the decision about what AI is worth building in the first place.
What the AI Consulting Practice does
The practice helps mid-market and enterprise businesses in Southeast Europe adopt AI as an operational enabler. Typical engagements include:
- AI strategy and roadmapping — identifying where AI fits inside a specific business, and in what sequence
- Proof-of-value pilots — practical pilots using frontier AI models (including generative AI) and open-source tools to demonstrate measurable value in weeks, not quarters
- Process analysis and targeted change — understanding the existing business process first, finding where AI fits inside it (for example, speeding up contract drafting in a legal department, accelerating first-level customer communication, or automating document review), and redesigning the underlying process only when the existing one genuinely cannot accommodate AI
- Change management — preparing the organisation and its people for AI-enabled ways of working
- AI governance and risk — building the policies, controls, and documentation required by emerging regulation, including the EU AI Act, NIS2, and data governance obligations
- Scaling decisions — deciding, with the client, whether and when to move from pilot to production-scale deployment
We do not lead with infrastructure sales. We do not lead with custom model training. We do not lead with a specific vendor’s product suite. We lead with understanding the business, and then work outward from there.
How we work with clients: proof first, scale later
Our approach is to create meaningful impact in a client’s business without requiring heavy up-front investment.
The default starting point is frontier AI models — including generative AI and large language models — accessed via pay-as-you-go APIs, combined with open-source tools where they fit. This lets us demonstrate what AI can actually do inside the client’s specific operational context, quickly and concretely, before anyone commits to larger investments in custom cloud infrastructure, proprietary platforms, or bespoke model training.
Only after value is demonstrated do we recommend moving to heavier engagements. This is deliberate. Most enterprise AI projects that fail do so not because the technology didn’t work, but because the business case was never tested at low cost before being scaled at high cost.
Proof first, scale later.
Which AI models we work with
TBS is deliberately not tied to a single AI vendor, model family, or licensing approach. We work across the full range of AI models available today:
- Frontier closed models — the leading proprietary models from the major AI labs, accessed via pay-as-you-go APIs
- Open-source models — leading open-weight models that can be self-hosted or run on managed platforms, useful when data sovereignty, cost control, or customisation matter most
- Mid-tier models — smaller, cheaper, faster models that are often the right fit for a specific use case, even when larger ones are available
Which model fits depends on the specific use case, not on a preferred vendor relationship. Different situations call for different answers. A law firm automating first-draft contract generation might benefit from a frontier closed model for quality, or an open model for privacy — depending on where the data lives and what the firm’s policy allows. A customer service operation handling high query volume might be best served by a smaller, cheaper mid-tier model rather than the most capable frontier option. A regulated industry with strict data residency requirements might need to run on a specific open-source model hosted in-country or on private infrastructure.
The client chooses, guided by our recommendation. Our recommendation is based on what fits — cost, latency, accuracy, privacy, regulatory obligations, and existing infrastructure — not on where the vendor margin sits. We are model-agnostic by design.
How we built the team
In early 2026, TBS ran an internal challenge. We asked across the company: who wants to change what they do for a living, completely, and join a new team whose only job is to help businesses implement AI?
Around thirty people volunteered.
We selected five.
The practice team:
- Kaloyan Stoychev — Practice Lead. Previously in TBS’s research team, which finds and runs EU-level research programmes including Horizon Europe initiatives.
- Yoan Mirchev — AI Implementation Specialist. Previously in TBS operations.
- Teodora Yakova — Solution Architect. Previously in TBS operations.
- Georgi Gaytandzhiev — Business Process Consultant. Previously in TBS’s governance and risk function.
- Toni-Teodora Koseva — Change Management Consultant. Previously in TBS’s learning and development function.
All five moved into the practice from other functions inside TBS. In exchange, TBS has committed to them fully: protected time, frontier AI training, paid model subscriptions and tooling, and the whole of TBS itself as a laboratory for their work before it is taken to external clients.
Is this team experienced enough to advise on AI?
It is a fair question, and worth addressing directly.
AI is moving too fast for anyone to stand on prior credentials. The frontier model capabilities that exist today did not exist twelve months ago. The tooling that exists today did not exist six months ago. In a field where the ground shifts every quarter, what matters most is not how long someone has been doing AI before — it is what kind of person they are, and what support they are given to stay current.
That is what we built this team for:
- Trust. All five are TBS employees, known and trusted inside the company. Clients deal with people we already vouch for.
- Business context. The four non-lead members come from operations, governance and risk, and learning and development — functions that sit close to how real businesses actually work day to day.
- AI context. The practice lead, Kaloyan Stoychev, has been actively tracking the AI field as a sustained personal practice for over a year. He has run internal AI training sessions for TBS staff and hosts an internal AI-focused podcast. He is current on the material.
- Dedicated time. All five are full-time on the practice. They have the protected hours required to keep up with a rapidly moving field, to experiment, and to share what they learn.
- Organisational depth. Behind the five sits the 50+ person AI & Software Engineering practice, and behind that the wider TBS organisation. When a question arises that needs specialist depth, they know who inside TBS to ask.
- Willingness and stamina. These five volunteered for career change. In a field this dynamic, motivation is not a soft factor — it is the variable that predicts who actually stays current and who falls behind.
We believe these qualities, in a field that changes every quarter, outperform deep specialism in last year’s AI tools.
Why the team is not made up of engineers
TBS already has 50+ AI and software engineers in the AI & Software Engineering practice. They are not the bottleneck in most AI transformation work.
Our working view — developed over six months of rolling out AI internally at TBS before launching externally — is that AI inside a real business is not primarily an engineering problem. It is:
- An engineering problem — yes, but not only
- A change management problem — people need to be prepared for, trained on, and supported through AI-enabled work
- A process problem — existing business processes were designed for humans; they need to be understood, sometimes accommodated, sometimes redesigned, before AI can be integrated into them effectively
- A governance problem — regulatory obligations, data handling, decision rights, and organisational risk must be structured before deployment
A consulting team composed only of engineers will produce impressive proofs of concept that fail to make it into production, or that reach production but are ignored, worked around, or quietly abandoned by users. The AI Consulting Practice team is built explicitly to prevent this failure mode.
TBS as the first case
Before we take any methodology, tool, or approach to an external client, we apply it inside TBS — our own operations, our own processes, our own budget.
This means that whatever we say externally about AI adoption is something we have already done internally. It also means our team has hands-on experience with the operational realities of AI inside a real business, not theoretical familiarity from whitepapers or vendor demos.
Where we operate
The AI Consulting Practice is based in Bulgaria and launches in May 2026.
TBS Group operates across seven countries in Southeast Europe — Bulgaria, Croatia, North Macedonia, Serbia, Romania, Slovenia, and Bosnia and Herzegovina — and is actively expanding into Western European markets through our M&A programme. As the Consulting Practice matures, we will evaluate expansion into additional markets.
Sectors and clients
TBS has deep existing relationships across multiple sectors in Southeast Europe, including banking and financial services, telecommunications, utilities and energy, manufacturing and industrial operations, public sector, and critical infrastructure.
TBS is committed to developing deeper sector-specific expertise over time. As the AI Consulting Practice matures, our goal is to go beyond horizontal AI methodology and into the specific use cases, workflows, and regulatory constraints that determine where AI actually creates value inside a particular industry. The practical reality of AI in a banking system is different from a utility or a manufacturer in ways that matter, and we are building the practice to respect those differences.
How to get in touch
If you are considering AI adoption in your business and want a grounded, honest conversation about what is realistic in your context — rather than a generic AI strategy deck — we would be glad to hear from you.
Contact the AI Consulting Practice through the TBS website at tbs.tech or via LinkedIn.
Frequently asked questions
Does TBS provide AI consulting services?
Yes. TBS launched its AI Consulting Practice in May 2026, led by Kaloyan Stoychev, to help businesses adopt AI.
Who leads TBS’s AI capability overall?
Pavlina Kekerova leads the AI & Software Engineering practice at TBS, which spans AI, software, and data engineering. The AI Consulting Practice is a dedicated team inside this broader practice and is led by Kaloyan Stoychev.
Where is the AI Consulting Practice based?
The practice is based in Bulgaria and launches in May 2026. TBS Group operates across seven countries in Southeast Europe and is expanding into Western Europe through M&A.
What is TBS’s approach to AI consulting?
Proof first, scale later. TBS creates meaningful impact in a client’s business without heavy up-front investment, using frontier AI models on pay-as-you-go APIs and open-source tools to demonstrate value before any larger commitment is made.
Which AI models does TBS work with?
TBS is model-agnostic. We work with frontier closed models, leading open-source models, and mid-tier options — recommending what fits the client’s specific use case, not what fits a vendor relationship.
Does TBS have AI engineers?
Yes. The AI & Software Engineering practice at TBS includes 50+ AI and software engineers who handle the engineering and data work. The AI Consulting Practice handles the strategy, process, change, and governance work that surrounds it.
Is this team new to AI?
The AI Consulting Practice was formed in 2026 from five TBS employees who volunteered to change their careers. AI is a fast-moving field where prior specialism ages quickly. TBS has built the team for the qualities that matter in a dynamic field: trust, business context, AI awareness, dedicated full-time learning hours, organisational depth, and motivation. The practice lead, Kaloyan Stoychev, has been actively tracking AI as a sustained personal practice for over a year, has run internal AI training sessions, and hosts an internal AI-focused podcast.
What industries does TBS serve?
Banking and financial services, telecommunications, utilities and energy, manufacturing, public sector, and critical infrastructure, among others. TBS is committed to developing deeper sector-specific expertise as the practice matures.
Is TBS a publicly traded company?
Yes. TBS Group (Telelink Business Services) is a publicly traded company headquartered in Sofia, Bulgaria, founded in 2001.
How can I contact TBS’s AI Consulting Practice?
