From a coder's perspective: the best AI solutions are always based on high-quality data.
06 | 2025 Antti Rintala, Data Architect, partner Kipinä
There is a lot of talk about AI solutions, but less about what they are actually built on. The answer is simple: data. And not just any data, but high-quality, reliable and business-driven data.
In a blog series, Kipinä consultant Antti Rintala has summarised the kind of foundation required to use AI. In this compilation, we bring together key perspectives that help build a solid foundation not only for AI solutions - but also for sustainable knowledge management in business.
1. Data quality matters - not quantity
The success of an AI solution starts with the data available actually reflecting reality - accurately, timely and logically. Data quantity is an attractive metric, but without quality it can mislead or distort analysis.
Key quality dimensions include:
Timeliness: is the data real-time or is a quarterly update sufficient?
Accuracy and integrity: does the data really reflect what it should?
Consistency and accuracy: is there conflicting information on the same subject in different systems?
Contextual validity: is the data relevant for the purpose for which it is used?
The most important thing is to be aware that quality is not a single technical indicator, but a shared goal of the organisation. Without it, AI models may well make decisions - but from the wrong starting point.
2. Trust is built on sources, transparency and documentation
High-quality data is one thing - but reliable data is quite another. The reliability of data is built on many factors: where it comes from, how it is processed and how openly it is managed and documented.
The main points of view are:
Credibility of the source: do you know the origin of the data, the reputation and expertise of the producer?
Data lineage: can you trace how data has moved and changed at different stages?
Transparency: is there sufficient background information on data processing and decision-making available to non-technical experts?
By being transparent about the data handling process, an organisation can not only develop its own skills, but also demonstrate accountability to the outside world - an increasingly important competitive advantage.
3. Good governance combines strategy and data
Often, data is treated in isolation from business objectives. Data management is best aligned with the strategic priorities of the business: what to measure, why to measure it, and under what conditions.
Information policies should reflect business strategy - not just comply with regulation.
Enforcement can mean technical means, not just guidelines.
Relevance does not come from the data itself, but from business needs: what supports decision-making, what doesn't?
Data management is not just about managing risk - it's about building competitive advantage. When business and data are aligned, a common language and better decision making emerge.
4. Data security and regulation - plan ahead, not afterwards
Data security and data protection are not things that can be added at the end of a project. They need to be taken into account when building the data and systems foundation - otherwise technical debt and business risk will be created.
Key observations:
Cybersecurity is more than a firewall - it's about continuous planning and risk management.
Data protection legislation (such as GDPR) is constantly changing - good documentation and clear policies reduce risks.
Cooperation between the legal department and the data team must not be haphazard.
Introducing AI can mean handling sensitive data in a new way - which is why it's critical to be prepared even before the first line of code is written.
5. AI is never finished - it is constantly evolving
AI is not a "set it and forget it" solution. Models change, data quality varies, and the business environment evolves. Without constant monitoring and maintenance, systems will start to draw the wrong conclusions.
Why is this important?
The performance of models will degrade over time (e.g. model drift) if they are not updated.
Ethics and transparency require human involvement in decision-making - including when AI makes proposals.
Bias and discrimination can occur unnoticed if the data is not balanced or if testing is missing.
Managing AI is not just a technical issue, but a business and ethical responsibility.
Building the real value of an AI solution before the code
Companies that invest in a high-quality, transparent and strategically aligned data platform are not only getting more out of AI - they are also preparing for the future. Data quality, management and reliability are not isolated steps, but part of an ongoing strategic effort.
A good AI solution doesn't start with code - it starts with the question: do we trust our own data?
Antti Rintala, Data architecht & partner Kipinä
Antti Rintala is a software developer and solution architect with nearly 20 years of experience in cloud services, IoT solutions and diverse technologies. He combines strong technical expertise with business-oriented thinking and specialises in Azure environments, performance and user experience. Antt is driven by a desire to solve complex business challenges through technology - in a practical and impactful way.