The AI gold rush is in full swing. Business leaders across Norway, from tech startups to established companies, are racing to adopt or embed artificial intelligence to stay competitive. Over 73% of organisations globally use or pilot AI in core functions, and 78% reported using AI in at least one business function in 2024 — a significant rise from 55% in 2023.
The dangerous gap between AI ambition and infrastructure reality
A recent survey from IDC (International Data Corporation), conducted in January 2024, shows that 23% of organisations are now directing budgets towards GenAI projects. That focus, however, can lead them to overlook the critical work of modernising and streamlining their application portfolios.
— IDC, January 2024This misalignment of priorities creates what we call the “AI infrastructure gap” — a chasm between ambitious AI goals and the underlying systems’ ability to support them.
For years, core and ERP systems have been the single source of truth for how businesses store and manage information. AI fundamentally challenges that model.
Why is this especially acute in traditional industries?
The problem is particularly acute in several traditional industries, where nearly three quarters of banks globally still run on legacy core banking systems that cannot deliver on transformation goals. New reports from International Data Corporation (IDC), commissioned by Thought Machine, reveal that 98% of banks plan to upgrade their core systems. If banks — with large IT budgets and regulatory requirements — struggle with this balance, the challenge is even greater for Norwegian SMBs.
Why core systems matter more than ever in the AI era
Your core systems are the organisation’s nervous system; they handle the most critical functions: customer data, financial transactions, inventory management and operational workflows. When these systems are outdated, unstable or poorly integrated, they become a bottleneck that constrains every other digital initiative — especially AI.
Modern AI applications require data in consistent formats, accessible via APIs, processed in real time and secured to current standards. When your core systems store data in proprietary formats, lack modern integration capabilities or operate with outdated security protocols, every AI initiative becomes exponentially more complex and costly — or impossible altogether.
A strategic approach to balancing AI and core modernisation
The solution is not to abandon AI initiatives or postpone them indefinitely while every system is updated. Successful companies take a strategic approach that addresses both needs at the same time.
Start with a comprehensive audit of your current systems
Identify which core systems directly affect your AI goals, and which can remain unchanged for now. This assessment should evaluate data quality, integration capabilities, security posture and scalability potential.
Modernisation strategy
Run a phased modernisation strategy aligned with your AI roadmap. Focus first on systems that will directly support or enable planned AI applications. That may mean modernising the customer database before implementing AI-driven personalisation, or updating the inventory system before rolling out predictive analytics.
Consider hybrid approaches that can deliver immediate benefits while laying the groundwork for long-term modernisation. API encapsulation can make legacy systems more accessible to modern AI tools without requiring full replacement, but it does not remove data-quality challenges and can increase the complexity of the solution itself. Data lakes or data warehouses can consolidate information from multiple legacy systems and give AI applications the clean, integrated datasets they need.
Invest strategically, not in frequent, uncoordinated moves
Invest in integration platforms that can bridge old and new systems during the transition. These platforms can handle the complexity of connecting disparate systems while providing a foundation for future AI implementations.
Building AI-ready foundations
At Nettkonsult, we understand that successful AI implementation starts with solid foundations. Our approach focuses on creating system architectures that support both current operational needs and future AI ambitions.
We begin every project with a thorough assessment of existing systems and their ability to support planned AI initiatives. This is not just a technical audit — it is a strategic evaluation that considers business goals, budget constraints and timelines.
Our modernisation strategies are designed to be AI-enabling from the start. We build with modern data architectures, API-first design and cloud-native approaches that provide the scalability and flexibility AI applications require. That means you are not only solving today’s problems — you are creating the foundation for tomorrow’s opportunities.
We implement security and compliance frameworks that meet both current regulatory requirements and expected AI governance standards. This proactive approach ensures your systems are ready for AI implementation without needing further security retrofitting.
Our integration expertise lets us build bridges between legacy and modern systems during transition periods. That means you can start realising value from AI while gradually modernising underlying systems, instead of waiting for a complete infrastructure overhaul.
What is the real cost of AI without a solid foundation
The numbers speak for themselves when we look at organisations that have tried to deliver AI solutions on unstable core systems. One study shows that 70% of AI implementations fail or deliver significantly lower returns than expected, and the main cause is often inadequate data infrastructure and legacy issues.
Norwegian case: when the CRM system slows the AI project
For Norwegian companies, the consequences can be especially costly. A mid-sized Norwegian business invested about EUR 220,000 in an AI-driven customer service solution, only to discover that a 12-year-old CRM system could not deliver data in the right format. The result was a further about EUR 160,000 in integration and system upgrade costs, and the project took 18 months longer than planned.
This pattern repeats across industries. Retailers trying to implement AI-driven inventory management discover that old ERP systems cannot handle real-time data flows. Manufacturing companies aiming for predictive maintenance find that operational data is scattered across incompatible systems from different decades.
The hidden cost does not only come from delayed projects and budget overruns. AI systems built on weak foundations perform worse, produce inaccurate predictions and create a false sense of security that can lead to costly decisions based on flawed data.
How modern architecture enables AI success
Companies that have succeeded with AI transformation share one thing: they have invested in modern, integrated system architectures that support AI applications naturally. This means more than upgrading servers or migrating to the cloud — it requires a fundamental rethink of how data flows through the organisation.
Microservices architecture has become the standard for AI-ready systems because it allows independent scaling of different components. When your AI analytics engine needs more compute power, it can scale separately without affecting other critical business systems. That is difficult with monolithic legacy systems where all components are tightly coupled.
API-first design ensures that all system components can communicate effectively with AI tools. Instead of building custom integrations for every new AI application, a well-designed API architecture provides standardised ways for AI systems to access and manipulate business data.
Cloud-native solutions provide the elastic scalability AI applications require. Machine learning models may need intense compute power for short periods, followed by minimal resource needs. Traditional on-premise infrastructure cannot adapt to these fluctuating demands cost-effectively.
Containerisation with Docker and orchestration platforms such as Kubernetes provide the portability and consistency that complex AI workflows need. This makes it possible to experiment with different AI tools and approaches without locking into specific vendors or technology stacks.
The security implications of AI on legacy systems
Legacy systems often lack the granular access control needed to secure AI applications. Traditional “all or nothing” patterns fit poorly with AI systems that may need access to specific datasets based on context and user roles.
GDPR and privacy with AI on legacy systems
GDPR and other privacy regulations add further complexity. AI systems must be able to track data selection and processing to fulfil the “right to be forgotten” and other privacy rights. Legacy systems that were not built with this traceability in mind can make compliance extremely difficult.
Audit trails and monitoring become critical when AI systems make automated decisions that affect business outcomes. Modern system architectures build in comprehensive logging and monitoring, while legacy systems often provide limited insight into system operations.
Zero Trust security models, which are increasingly necessary for AI-enabled organisations, require every system interaction to be verified and authorised. That is difficult to implement retrospectively on legacy systems designed around “trust but verify” principles.
Organisational changes that come with AI enablement
Implementing AI on modern core systems requires not only technological change, but also organisational adaptations that affect everything from role structures to decision-making processes.
DevOps culture becomes even more important when AI is introduced into the organisation.
DevOps, DataOps and new roles
Continuous integration and delivery (CI/CD) must expand to include machine learning models, which have different lifecycles and quality requirements than traditional applications.
DataOps is emerging as a discipline focused on optimising data flow and quality for AI applications. This requires cross-functional collaboration between data engineers, data scientists and business analysts.
New roles such as ML Engineers, AI Ethics Officers and Data Stewards become necessary to manage the complexity of AI-enabled organisations. Legacy organisational structures must adapt to support these new functions and responsibilities.
Decision governance must be updated to handle AI-generated recommendations and automated decisions.
Guidelines and change management
This includes establishing guidelines for when AI recommendations can be followed automatically, and when human override is required.
Change management becomes critical because AI can alter how employees perform their day-to-day work. Without proper preparation and training, even the best AI implementation can face internal resistance that undermines success.
Conclusion: Building tomorrow’s success on today’s foundations
The AI revolution is real, and the potential to transform businesses is enormous. But as with any major technological shift, success depends not only on adopting new tools, but on having the infrastructure to support them effectively.
— NettkonsultThe question is not whether to invest in AI or core systems — it is how to do both strategically. Companies that find this balance will not only keep pace with the AI revolution, but lead it in their markets.
Norwegian SMBs have an opportunity to build competitive advantage through thoughtful integration of AI capabilities and modernisation of core systems. The key is to avoid the trap of chasing shiny new technologies while overlooking the foundations that make them possible.
Your core systems are the foundation of your digital future. Make sure they are strong enough to support the AI-driven business you want to become.
At Nettkonsult, we specialise in building AI-ready foundations through strategic modernisation of core systems. Our expertise in creating scalable, integrated and secure architectures ensures your AI investments deliver maximum value.
Ready to make your core systems AI-ready?
Contact us to discuss how we can help you balance immediate AI opportunities with long-term infrastructure needs.