Projects in production. Measurable results. Real clients.
Generative AI is creating real value in real companies. But there’s a significant gap between the enthusiasm generated by demos and the reality of what’s being done with measurable results in concrete organizations.
Much more information can be gleaned from the MIT study that found 95% of AI prototypes never reach production: https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
The three cases we present below are not prototypes or lab pilots. They are projects we’ve implemented with our clients, in production, with verifiable metrics. In all of them, the same pattern emerges: the differentiating factor isn’t the AI model itself, but rather the data preparation, the clarity of the problem, and the management of organizational change.
These real-world examples of generative AI applied to business demonstrate that successful projects share a common pattern: they start with a specific problem, prepare the data before automation, integrate the solution into workflows, and ensure the client maintains technical control. AI delivers value when it reduces friction, accelerates processes, improves governance, and enables the scaling of data products without creating black boxes.
The conversation about generative AI in companies usually begins with very attractive demos: assistants that answer questions, tools that summarize documents or agents capable of automating repetitive tasks.
But a demo does not represent the complexity of a real company.
More difficult questions arise in production: where does the data come from? Who validates the answers? How is the information updated? What happens if the model is wrong? What team maintains the solution? How is the system audited?
Therefore, many AI projects do not fail due to lack of technological capacity. They fail because there is no solid foundation of data governance, clear processes or internal ownership.
At Galde, the approach is based on a simple idea: generative AI applied to business only works when it is designed as part of a real transformation, not as a superficial layer on top of messy data.
Case 1 · Digital Platforms / Employment
documentation
time
faster
onboarding
documented
assets
projected annual
savings
InfoJobs, Spain’s leading job portal (part of the Adevinta group), had grown exponentially. Its data infrastructure had scaled at the same rate, but governance, documentation, and standardized practices had lagged far behind.
The timing was particularly critical: InfoJobs was about to become an independent legal entity, which required translating Adevinta’s global governance framework to its own operational reality. Previous attempts to implement governance had met with limited acceptance due to organizational resistance; teams perceived it as bureaucracy, not as a value-adding structure.
The differentiating factor wasn’t the technology itself, but the methodology. Instead of imposing a top-down technical framework, we used generative AI as a lever for adoption and knowledge transfer:
Galde used generative AI as a lever for knowledge adoption and transfer. The key was not to impose a technical framework from above, but to reduce friction within teams when working on knowledge tasks.
The work included:
“Exceptional in every aspect. The AI-accelerated data governance transformation is technically rigorous and culturally sensitive. The 4x improvement in onboarding and 60% adoption of the governance demonstrate a real transformation.”
— AI Transformation Award Jury Member · Malt, 3rd Place Finalist out of 161 projects
Key takeaway: Organizational resistance (not technology) is the primary reason governance initiatives fail. When AI is used to eliminate friction in knowledge work, without replacing human judgment, adoption accelerates dramatically.
Case 2 · Reinsurance / Financial Services
MVP in
production
multi-country
since day 1
black
boxes
internal ownership for
SCOR
SCOR SE, in the APAC (Asia Pacific) region, needed to launch a new credit-scoring product to support B2B partners in India and extend it to multiple countries in the APAC region. The challenges were simultaneous: speed of validation (weeks, not months), regulatory heterogeneity between countries, operational scalability of the product, and complete control of the model by the internal team, without vendor lock-in.
Galde applied its MVP60™ co-creation framework, designed for data products in regulated industries. The project was structured around three pillars:
The project was executed with SCOR, not for SCOR: internal teams participated in the development to ensure full visibility and ownership.
“Thanks to Galde, we launched a credit-scoring prototype in record time. It allows us to start operations in India and scale it easily and autonomously across APAC.”
— Adeline Chua, Head of Product & Innovation, APAC (SCOR)
Key takeaway: Speed wasn’t achieved by cutting corners on governance, but by building it into the process from day one. A transparent and auditable model is faster to validate with regulators and easier to scale across countries.
Case 3 · Reinsurance / Sustainability
of its kind
in the world
consortium with
AXA (2026)
biomes and
geographies
international
ecological standard
SCOR SE set out to build a first-generation insurance product to capture the risks of ecological restoration projects, a market that had never been insured in this way. The challenges were exceptionally complex: no existing price benchmarks, apart from a few pilot programs by other large reinsurers; processes that develop over decades; and the need to simultaneously align actuaries, legal teams, environmental experts, and global innovation teams.
The product had to be, from the outset, a scalable platform, not a one-off, capable of evolving across different project types, geographies, and risk profiles.
Galde joined SCOR SE’s Product & Innovation team to co-develop the technical and commercial foundations of the NatReCo solution. The work was structured around three pillars:
“Galde consistently brought strategic judgment and cross-functional collaboration to complex discussions involving pricing models, multi-stakeholder data gathering, and product design. Their contributions were fundamental in aligning technical feasibility, regulatory requirements, and commercial viability within a highly innovative and nascent product area. The impact of Galde’s work has been tangible: these initiatives have grown into industry collaborations, including the launch of a Lloyd’s of London consortium with AXA.”
— Henri Douche, Head of Product & Innovation (SCOR SE)
Key takeaway: The unprecedented complexity of the domain, lacking price benchmarks and involving multiple simultaneous stakeholders, could only be addressed with a methodology that prioritized technical clarity, regulatory feasibility, and governance from day one. The product wasn’t built for SCOR; it was built with SCOR.
Analyzing these three cases, consistent patterns emerge that distinguish them from projects that don’t reach production or are abandoned early on:
Generative AI is creating real value in real companies. Not in every company or every project, but certainly in those that have identified a specific problem, properly prepared the underlying data, and managed the implementation with the same seriousness as any other transformation project.
The InfoJobs, SCOR APAC, and SCOR SE + AXA cases don’t share technology, sector, or geography. They do share a methodology: clarity about the problem, data preparation as a prerequisite, not a consequence, and a genuine transfer of capabilities to internal teams.
The enthusiasm of demos must be contrasted with the discipline of production. And the discipline of production almost always begins with the data.
Galde works with companies that need to move from idea to execution, building data and AI solutions that can operate in real-world contexts.
Through its generative AI service, Galde helps identify viable use cases, design secure solutions, and build production-ready systems.
Through data governance, it supports organizations in defining ownership, documentation, traceability, and quality processes.
And through data platforms, it builds the necessary technical foundation to scale data products, advanced analytics, and AI.
You can see more projects in Galde’s success stories section or explore other content in Thought Leadership.
Generative AI applied to business involves using models capable of generating responses, documentation, analysis, or automation within real business processes. Its goal is not to create a demo, but to improve processes, reduce friction, and generate measurable results.
Many generative AI projects fail to reach production because they are developed without prepared data, governance, clear metrics, or internal ownership. The technology may work in a test, but fail when integrated into real-world processes.
Data governance enables AI to work with reliable, documented, traceable, and up-to-date information. Without governance, models can generate seemingly coherent responses based on incorrect or disorganized data.
Some real-world examples include the use of generative AI for data governance at InfoJobs, the development of a scalable credit-scoring product for SCOR APAC, and the creation of insurance products for ecological restoration with SCOR SE and AXA.
Before applying generative AI, a company needs to define a specific use case, prepare its data, establish responsibilities, design a secure architecture, measure the impact, and ensure that the internal team can maintain the solution.
