Our CEO, Anthony Gyursanszky, recently sat down with a leading Swedish CIO media Techtidningen at the Google Cloud Summit Nordics to discuss the massive shift happening in enterprise AI—and why the current ”bottom-up” approach is leading straight to chaos.
Here you can read the English translation of the original Swedish article:
”Codento CEO Anthony Gyursanszky has a vision for a company where AI agents make important strategic decisions. His advice to CIOs looking to succeed with their AI initiatives is to start with a clear target vision and establish control over agents from the outset.
AI agents can deliver enormous value to businesses, but they also require governance, strategy, and scalability. That is the message from Anthony Gyursanszky, CEO of Codento, when Techtidningen meets him at Google Cloud Summit Nordics.
Codento describes itself as a Nordic AI and data consultancy with an exclusive focus on Google Cloud. The company has grown rapidly in recent years and recently acquired Sweden-based Dataedge. Its goal is to become a full-service Nordic Google Cloud consultancy, with AI and data at its core.
– We are a fairly unique player in the market because we believe in scalable AI agent solutions that are planned top-down by the executive leadership team. This enables the CIO to properly address compliance requirements while also helping the business succeed.
The company’s largest customers are primarily in the telecommunications sector. Codento works with all Finnish telecom operators as well as a global telecommunications provider. Its customer base also includes banks and major retail companies.
Codento has chosen Google Cloud as the strategic foundation for building AI solutions at scale. Anthony Gyursanszky sees particular value in Google’s end-to-end approach, where everything from hardware to language models and agent platforms is part of the same ecosystem.
– They differ from other IT vendors that are constrained by their existing product portfolios and the traditional, 30-year-old rule-based world in which they are trying to implement AI. Google can truly focus on helping customers move toward the future, and over the past year I think they have demonstrated that their solutions accelerate innovation.
There remains, however, a significant gap between what the technology can do and companies’ ability to translate those capabilities into real business value. According to Anthony Gyursanszky, the vast majority of Nordic companies have yet to achieve substantial impact from tactical AI deployments. In his view, success requires not only new technology but also a new way of thinking about business, organisation, and governance.
– It requires a major shift in mindset and probably a cultural transformation as well. We need to be more humble about rethinking both business and technology in the AI era.
When Codento engages with a new client, the work often begins with the executive team. For the collaboration to be truly successful, there need to be committed leaders who understand AI’s potential. That person may be the CIO, but it could just as easily be someone else in the leadership team who has both the mandate and the willingness to drive change.
The process typically starts with workshops designed to reshape the company’s strategy through the use of AI agents. This allows leadership teams to experience firsthand how AI agents work in practice.
– In this way, AI capabilities become integrated into the strategy process while we demonstrate the power of AI agents, ensuring that the entire leadership team becomes engaged and understands what can be achieved with AI.
– We then simulate how roles within the organisation will change. What new skills will be needed, what new roles will emerge, how existing roles will evolve, and likely which roles will become obsolete. This helps them design the agent system in a way that reflects its future potential.
Codento also works with customers who choose a narrower entry point into the world of AI. This may involve integrating AI into a specific function such as sales, marketing, or customer service. In those cases, Codento builds a scalable agent capability on Google Cloud that solves a specific business challenge while being designed to scale horizontally and upward within the same architecture. A key part of Codento’s philosophy is ensuring that AI projects do not become yet another isolated solution but instead fit into a broader enterprise-wide framework.
When asked where AI agents are particularly effective, Anthony Gyursanszky points to the strategy process. It is especially interesting because it does not immediately involve real-time production data, yet it influences a company’s most important decisions.
– AI agents can provide the right information for making decisions about market conditions, strategic options, and can even simulate competitors.
Other areas seeing rapid progress include inbound and outbound operations. He highlights outbound sales, customer service, and marketing as functions where AI agents are already delivering significant results. Operations is another area, where AI can quickly identify anomalies in vast amounts of data across industries such as telecommunications and manufacturing.
According to Anthony Gyursanszky, however, the real AI transformation occurs when these areas are connected. Today, strategy, outbound, and inbound functions are often treated as separate domains. The greatest impact is achieved when agents operating across different layers can communicate with one another.
One of the more advanced examples involves competitor simulation. Anthony Gyursanszky describes a process in which a strategy is first developed with the help of an AI agent that challenges and refines it from multiple perspectives and across different market scenarios. The model then generates simulated strategies for competitors.
These strategies can subsequently be tested in a kind of wind tunnel, where different actions and responses are simulated to evaluate the resilience of the company’s own strategy.
– You discover where the weaknesses are and can prepare accordingly. It sounds like science fiction, but in many cases we are doing this with executive teams right now.
At the same time, Codento sees clear problems in the way many organisations are approaching AI. Anthony Gyursanszky describes how machine learning was previously a more mature and disciplined field, often closely tied to operational and strategic processes. When generative AI emerged, however, it was largely categorized as a personal productivity tool.
– When you apply that mindset across an organisation, it becomes chaos. Everyone controls their own agents, there is no guidance from strategy or AI governance, and agents cannot be reused across teams and individuals.
He argues that it is essential to design AI strategy and governance in parallel with the technical orchestration of AI agents. If the work does not begin at the top, organisations risk ending up with a growing landscape of unmanaged AI solutions, leaving the CIO to regain control after the fact.
At Google Summit Nordics, Codento launched a concept it calls the “Self-driving Enterprise.” The framework is designed to address many of the common challenges that arise when organisations lack a cohesive AI strategy.
He describes the Self-driving Enterprise model as a way to first define the rules and then allow a central governance function to determine how agents are permitted to operate.
– The steering wheel is the place where, at either the enterprise or functional level, you decide how agents should be instructed. When the agents are implemented, they do not contain that logic internally. Instead, they ask the steering wheel: am I allowed to do this? And once they take action, the steering wheel also verifies that it was done correctly.
The idea is that companies can build AI in a controlled manner from the outset. Existing agents can be connected into the domain without every new agent becoming yet another component in an increasingly difficult-to-manage AI landscape.
Codento has created ten different modules covering areas such as finance, sales, marketing, customer service, operations, executive management, and IT. Companies can choose where they want to begin and connect that function into the model.
For Swedish CIOs, the pressure to succeed with AI transformation is already high. Many are expected to drive innovation, manage costs, and ensure compliance simultaneously. Anthony Gyursanszky believes success depends not only on which new tools organisations choose to adopt, but also on whether leadership teams are willing to challenge their existing ways of working.
– The common denominator among successful companies that are proactively preparing for AI is that they challenge their own thinking and remain humble enough to keep learning. It is not easy for CIOs to get the organisation into this mindset, but if you believe that what worked in the past will also be the right approach in the AI era, it will not lead to success.”
Read the interview in Techtidningen (in Swedish): Techtidningen