Stay Ahead & Go Beyond
Youwe shares the latest projects and industry news to help you shape your version of digital success
Stay Ahead & Go Beyond
Youwe shares the latest projects and industry news to help you shape your version of digital success
Averting the Invisible AI Crisis: What Every Leader Needs to Know Today
When Nobody Knows What's Running
Here's a scenario: your marketing team has been using a copywriting tool for months, your sales team uploads client data to a chatbot they found online, and your operations department built a pricing algorithm using an API they discovered. The problem isn't that these tools are being applied, but rather that there is no clarity, visibility or related documentation on where, what and how. This pattern appears consistently when we audit organizations. Systems multiply invisibly across departments while most companies remain unaware until a problem occurs. What looks like innovation at the department level often creates serious vulnerabilities at the organizational level, particularly when these tools process sensitive data without any oversight.
Why this Creates Real Business Problems
Invisible systems create two distinct problems that aggravate over time. Firstly, the opportunities are missed with the solutions that could benefit your entire organization remain siloed and underutilized. Secondly, and far more seriously, you create real risk when unapproved systems process sensitive data without any security review, risking regulatory fine and reputational damage. "For example, if your product is selling high-quality advice, and your system gives bad advice, it can ruin your business," explains Sebastiaan den Boer, Director of Data Science at Youwe. "The issue isn't technical complexity, but rather awareness, since most companies never establish a process to ask whether a tool needs review before someone starts using it."
The Competitive Advantage
Procurement teams at major enterprises now include governance in vendor evaluation scorecards, which means companies without documentation are being eliminated during due diligence, regardless of product quality. Many B2B customers now require governance documentation before signing contracts, particularly in regulated industries, fundamentally changing the competitive landscape.
This creates a clear divide where companies that move early on compliance win contracts that competitors can't even bid for, build trust through transparency about how they manage systems, and identify operational improvements while conducting compliance reviews since visibility reveals both risks and opportunities that were previously hidden.
What You Need to Do Now
The fastest path to compliance visibility requires one clear mandate: inventory every system currently operating, including tools employees adopted informally outside official procurement processes. Once you have that inventory, classify each system by risk level, begin documentation for high-risk tools, and build explainability into how automated decisions are made.
The EU AI Act rewards organizations that build responsible systems from the beginning rather than slowing innovation. The question isn't whether to address visibility and governance, but whether you'll do it proactively as a competitive advantage or reactively under regulatory pressure when your options become more limited.
Fragmented Data is Killing Customer Experience
Organizations today face a critical problem of siloed data and content. When Product Information Management (PIM), Master Data Management (MDM), Digital Asset Management (DAM), Content Management Systems (CMS), and commerce platforms operate separately, the consequences can be quite costly.
Duplication and version drift create confusion about which data is the "system of truth," while data inconsistencies lead to inaccurate product content across channels, eroding customer trust. Teams struggle to gather, optimize, and distribute content quickly, resulting in slower product launches and missed market opportunities. When customers encounter conflicting or incomplete information, conversion rates suffer, and brand dilution occurs from inconsistent messaging and experiences across touchpoints.
0%
organizational data is unknown or inaccessible to those who need it
Omnichannel Excellence Through PXM
Today's customers interact with brands across multiple channels, and they expect seamless consistency throughout their journey. PXM ensures that whether a customer engages via social media, browses your website, shops on a marketplace like Amazon, or visits a physical store, they encounter unified, accurate product information tailored to that specific channel's requirements.
This omnichannel approach addresses one of the most significant challenges retailers face: maintaining consistent product information across platforms with unique technical specifications, content formats, and customer expectations.
As an illustration, you can check the case study of the Chadwicks Group, Ireland’s number one builders' merchant, which faced a critical challenge in managing its vast product catalogue: over 200,000 SKUs sourced from multiple suppliers. Chadwicks implemented Youwe Intelligence, a suite of AI-powered solutions designed to optimize product data to improve business efficiency. The first model rolled out was the Product Spider, focused on product data collection, competitor analysis, and overall data enrichment.
The Results
0%
Marketable Catalog Uplift
>0h
Saved manhours
Getting Started with PXM
Implementing PXM doesn't require a complete technology overhaul, you can start small and scale strategically. The option is always to commence with a pilot; launch PXM for a high-performing product category, demonstrate value, then scale organization-wide.
Evaluate current systems
Assess existing PIM, DAM, CMS, and commerce platforms to identify integration opportunities
Build your tech stack
Integrate PXM components that connect your existing systems and break down silos
Consider AI-powered solutions
Tools like SPIDER can accelerate product enrichment and eliminate manual bottlenecks
Master Your Toolkit with Shopify and Dotdigital
Breaking Free from Legacy System Constraints
The first episode tackles one of the most pressing challenges for eCommerce businesses: being trapped by outdated platforms that drain resources and stifle innovation. Many brands find themselves caught in a cycle of "keeping the lights on," spending valuable time and budget on maintenance rather than growth initiatives. High total cost of ownership, lack of agility, and innovation bottlenecks plague businesses stuck with monolithic architectures.
Composable architecture offers an escape route here: With low-maintenance infrastructure that automatically scales, merchants can redirect resources toward customer experience and business growth. Tools like Shopify Flow can help your teams build sophisticated automations without deep technical knowledge, while seamless integrations with partners like Dotdigital mean marketing teams can work faster and smarter.
Solving the Fragmented Data Problem
Episode two addresses a challenge of fragmented customer data, scattered across disconnected systems. When customer information, purchase history, and behavioral data live in silos, you can't deliver the coherent experiences modern consumers expect. The consequences are rea, for example sending abandoned cart reminders to customers who already purchased, retargeting with products they just bought, or missing critical signals about customer preferences.
This session reveals how a unified data foundation where information flows in real-time can be created with Shopify and Dotdigital's SaaS-to-SaaS integration. With RFM modeling and predictive analytics, you can identify high-value customer segments and tailor engagement strategies accordingly. The unified approach extends beyond marketing, to improved inventory visibility, order management, and fulfillment.
Delivering Personalization at Scale
The third episode confronts a challenge of meeting high consumer expectations for personalization while respecting privacy regulations and maintaining operational efficiency. Today's customers expect seamless experiences across every touchpoint, from online browsing to in-store purchases, with promotions, preferences, and loyalty rewards following accurately every step of their journey.
This session explores how unified data from Shopify powers Dotdigital's automation and personalization capabilities. Learn more on how behavioral triggers through Shopify Flow enable sophisticated automations like replenishment reminders and first-purchase campaigns that respond to real-time customer actions, and how a complete customer lifecycle journeys, from welcome series to reactivation programs, can be designed with Dotdigital. The emphasis on first-party data collection ensures you're building sustainable personalization strategies that comply with constantly evolving privacy regulations.
Future-Proofing Your Commerce Stack
In the final episode, you can see how to build a stack that grows with your business rather than holding it back. Scalability isn't just about surviving Black Friday, it's about having the flexibility to expand into new channels, markets, and business models as opportunities arise.
This session showcases Shopify's global infrastructure with over 300 points of presence, dynamically allocating bandwidth based on customer location and traffic patterns, and Dotdigital's modular approach that impliers you can start small and bolt on additional channels, increase data capacity, and integrate APIs as your needs evolve. With predictive analytics becoming more accessible and actionable, brands can shift from reactive to proactive marketing strategies that strengthen customer relationships at every stage.
Building a Bespoke AI Roadmap
What's Important
When a technology becomes ten times more available and ten times cheaper, demand explodes in ways we cannot predict. Consider what happened when pizza restaurants started offering delivery apps. Suddenly, people who might have cooked at home were ordering pizza three times a week because the friction disappeared. The overall pizza market didn't shrink because it became easier to deliver, but rather expanded dramatically.
The same thing is happening right now with AI capabilities. Tasks that used to require hiring specialized consultants or building custom software can now be performed in an afternoon with the right AI tools, and companies are identifying dozens of potential applications for AI across their operations. The challenge lies in coming up with the ideas that have the greatest impact on the business, that one has the data and capabilities to support, and the most sensible sequence given the current position.
Doing it Alone
We see a consistent pattern when companies attempt to develop AI strategies without structured guidance. They can identify plenty of potential use cases, but the problem is almost always prioritization. Without a framework for evaluating which use cases will deliver the most value, companies either pick the flashiest option or try to do everything without meaningful progress on anything.
Some use cases seem valuable on the surface but require data the company doesn't actually have, or technical capabilities they'd need to build from scratch. A company might want an AI assistant that helps sales prepare for client meetings by summarizing past interactions, but that can't happen until they've cleaned up how customer data flows between their CRM and other systems. Other use cases might be simpler to implement and deliver outsized returns as they eliminate repetitive, time-consuming work, but these opportunities often get overlooked because they're not as exciting or visible as customer-facing applications.
Phase 1.1. Current Diagnostics
The first phase involves understanding where you are right now. We map your current IT landscape and identify which AI tools are already in use across your organization. This number is almost always higher than leadership expects, and if nobody has visibility into the full scope of AI solutions used across the company, nobody can identify redundancies or opportunities to consolidate around better solutions.
Phase 1.2. Assessing Data Readiness
We also assess your data readiness during this phase, as AI systems are only as good as the data they can access. If your customer information is fragmented across multiple systems that don't talk to each other, or if your product data is incomplete and inconsistent, those data problems will limit what you can accomplish with AI, regardless of how sophisticated your tools are.
Phase 2 Possibilities
In the second phase, we introduce you to the new possibilities. This involves: sharing what we've seen work for other clients in similar situations, presenting frameworks for thinking about AI architecture, and exploring ideas your own team has been considering.
During this phase, we typically use AI tools ourselves to help structure the discovery process. Instead of filling out lengthy surveys about your business processes and pain points, we conduct AI-powered interviews. By the end of this discovery phase, we've typically identified anywhere from ten to a hundred different potential AI use cases for your business. These range from simple applications, like using AI for translation if you have international operations, to more complex applications, such as building AI assistants that help employees navigate your internal knowledge base, to sophisticated use cases, as using AI to optimize your supply chain or personalize customer experiences at scale.
Phase 3 Prioritization
The third phase focuses on structured prioritization. We evaluate each potential use case based on multiple dimensions: business impact, technical feasibility, data requirements, strategic value, and implementation risk. We're also identifying which projects’ sequence will compound value over time. Some ideas might deliver quick wins that build momentum and demonstrate value within months. Others might be longer-term plays that require foundational work first but unlock more sophisticated capabilities down the line.
This prioritization work also involves honest conversations about what you're not going to do, at least not yet. A use case might seem valuable, but if it requires technical capabilities you'd need to spend six months building, and there are other opportunities that deliver similar value with capabilities you already have, we help you understand why the simpler path makes more sense for now.
Phase 4 Detailed Plan
In the fourth phase, we develop detailed plans for the top three to five use cases. They're project proposals with clear technical requirements, resource needs, realistic timelines, specific success metrics and project sequencing.
Phase 5 Delivery
The final phase is delivery and go-live planning. We present the complete roadmap to leadership and help them understand not just what to do, but why this sequence makes sense for your specific situation.
From Strategy to Results
Companies that work through our roadmap development process typically gain several things that make the investment worthwhile.
First, they gain clarity about where to invest their AI efforts and how to sequence those investments over time. They understand the technical foundations they need to build and why those foundations matter. When someone proposes adding a new AI capability, they can evaluate whether they have the necessary components in place or whether they'd need to build additional infrastructure first.
They can articulate their AI strategy to stakeholders and explain how it connects to business objectives. Most importantly, they shift from scattered, reactive AI adoption to intentional, strategic implementation. Instead of different departments independently using whatever AI tools and creating disconnected islands of automation that can't share data or compound value, the organization develops a cohesive approach where earlier projects create capabilities that make later projects easier.
Governance frameworks scale across the business instead of being reinvented for each new initiative. You establish guidelines once about how AI should handle customer data, then apply those guidelines consistently as you roll out different AI applications. You build monitoring systems that work across all your AI use cases rather than each team creating their own monitoring in isolation.
This transformation doesn't happen overnight, and the roadmap provides a realistic, phased approach that helps companies make steady progress without becoming overwhelmed by the scope of what's possible with AI.
In Practice
A recent client had the ambitious goal of reducing office work by forty per cent within five years. When we mapped their current state, we discovered they were using various AI features across different software platforms, without full visibility. They were spending money on AI capabilities across multiple tools without knowing their total AI spend or whether they were getting value from those investments.
During our discovery workshops, we identified over sixty different potential AI use cases across their operations. Some were simple, like using AI to translate internal documentation for their international team members. Others were sophisticated, like building an AI system that could predict which customer inquiries would likely escalate to complex issues requiring senior staff attention.
Through structured prioritization, we helped them see which use cases would deliver the most value relative to their complexity and data requirements. We identified three quick wins they could implement in the first six months: automating routine data entry between their systems, implementing AI translation for their growing international team, and rolling out AI writing assistance for their marketing team with centralized prompt management to ensure brand consistency.
These projects would save over three thousand hours of employee time annually, reduce errors from manual data entry, accelerate their international expansion, and give their marketing team more time for strategic work instead of routine content creation. Just as importantly, these projects would build foundational capabilities in prompt management, workflow automation, and AI governance that would make their more sophisticated use cases possible later.
When Everything Connects, but Nothing Works Smoothly
This complexity impacts the business in several ways that directly affect the bottom line and your team's ability to deliver value. Projects that should take weeks take months because developers have to unravel existing integrations before they can build something new, meaning your time-to-market suffers while your competitors move faster. Operational costs rise steadily as teams spend more time maintaining fragile connections than building new features, and when things break, as they inevitably do, troubleshooting becomes a time-consuming detective exercise as the relationships between systems often exist only in the minds of a few key people.
Perhaps most worryingly, integration debt quietly erodes organizational agility, making it increasingly difficult to introduce new technologies, respond to market changes or retire outdated systems because everything has become so entangled that change feels impossibly risky. Technological complexity also creates security and compliance vulnerabilities, as data moves through channels that may not be properly monitored or controlled, creating potentially serious exposure.
The Warning Signs
“The warning signs usually appear when simple changes take disproportionately long to implement,” explains Erik Poolman, Software Developer (Back-end) at Youwe. "When a client asks you to change something very small, like adjusting how data flows between systems, and that change takes two days instead of two hours, that's a very big sign that integration debt has accumulated."
You should also pay attention if your team frequently discovers "zombie" integrations, which are data flows or connections that nobody owns, and yet everyone is afraid to turn off because of what might break. When multiple teams have built their own separate integrations to the same system because they didn't know about each other's work, that duplication signals a lack of visibility and coordination that typically indicates deeper integration debt. This also reveals that institutional knowledge about your integrations has degraded to a dangerous level.
Similarly, if you're experiencing frequent integration failures, especially ones that seem to cascade across multiple systems, that fragility suggests your architecture has become brittle through accumulated debt. Finally, when onboarding new systems or vendors takes substantially longer than expected because of the need to navigate existing integrations, you're seeing integration debt actively impeding your organization's ability to evolve.
Measuring the Real Cost
Understanding the true cost of integration debt requires looking beyond the obvious expenses to capture its full impact on your organization's effectiveness and efficiency. Start by creating an inventory of all your integrations, mapping out how many connections exist, what types they are, and how quickly that number has grown over recent months or years, since rapid, uncontrolled growth often correlates with accumulating debt.
You'll also want to measure your maintenance burden by tracking how much time your teams spend fixing integration failures, updating broken connections, and dealing with cascading issues when one system changes, then compare that time investment against effort that goes into building new capabilities that actually create business value. This comparison often reveals a sobering truth about where your resources are really going.
Change velocity provides another crucial metric, as you should measure how long it takes to integrate a new system into your landscape or to modify an existing integration significantly.
Look for redundancy and overlap as well, auditing whether you have duplicate data flows. Additionally, assess your governance and ownership situation by determining what percentage of your integrations have clearly assigned owners, up-to-date documentation, and active monitoring, because integrations without these elements are ticking time bombs waiting to cause problems.
How to Survive the Peak Season This Year & How to Thrive Next Year
- 01
No New Features Launches
Many organizations feel the pressure to squeeze in last-minute improvements, but as Arjan notes, this often leads to unnecessary risk. “We’ve seen clients introduce big changes just before the season starts, and it often causes more harm than good. Stability is everything when traffic hits its highest point.” Instead of new development, focus on performance, marketing alignment, and readiness.
Instead of pushing new development during this critical period, take the time now to reflect on which features you wish you had and which improvements would have made a difference. Add them to your roadmap for the next cycle, scheduled well before the autumn rush, so they can be properly tested, refined, and deployed with confidence prior to the next peak season. - 02
Schedule Load Test Early (and Together)
If this year’s load testing felt rushed or revealed unexpected weak spots, those learnings are invaluable for your planning. Arjan explains. “Load test is about finding weak links between systems; your website might perform well, but the connection to your PIM or other externally connected systems (tracking, socials) could be the bottleneck.”
By documenting where performance slowed down or systems strained under traffic, you can build a prioritized plan for next year.
Run your load tests earlier in the year—ideally during summer—so you have time to address issues calmly, allocate additional server resources if needed, and avoid emergency fixes in Q4. - 03
More Data, Better Planning
Your data from this season is your future roadmap. Now is the time to explore when traffic surges occurred, how customers responded to promotions, and which products drove the highest demand. Look closely at the relationship between marketing actions, such as email sends or ads, and resulting spikes in sessions and orders.
One client, Arjan notes, treats their annual campaign like a continuous optimization process:
“They prepare two months in advance, analyze every metric afterward, and use those insights to refine their next campaign. The results speak for themselves as each year, they break new records.” - 04
Think Beyond the Website
While your website takes center stage, operations often define the customer experience during peak season. If this year revealed stock shortages, overselling, delayed fulfilment, or tense moments with delivery partners, it’s crucial to document exactly what happened and why. The challenge might not be technical at all, but rather it might be forecasting accuracy, warehouse processes, or communication gaps. By capturing these operational insights, you can plan safeguards and improvements for next year, from stronger stock allocation strategies to clearer communication workflows for handling delays or unexpected surges in demand.
- 05
Strengthen Marketing–Tech Collaboration for Next Year
If this season brought chaotic moments caused by sudden traffic spikes or unexpected campaign impacts, it’s a sign that better alignment between marketing and tech teams is needed. Even small adjustments, such as sharing campaign timelines earlier or spreading email sends instead of releasing large volumes at once, can significantly improve site performance. Document where communication broke down this year and turn it into a more predictable collaboration model for next season. When marketing and tech move in sync, peak season becomes far more controlled and successful.
- 06
Continuous Improvement
Peak season can be chaotic, but with preparation, it can also be your most rewarding time of year. By planning early, testing thoroughly, and learning continuously, you’ll not only prevent issues, but set yourself up for record-breaking results.
Capture everything, translate it into improvements, and start building next year’s strategy while the insights are still vivid. “Every year is a chance to get better,” Arjan says. “If you treat peak season as a strategic project, not just a sales sprint, you’ll see long-term gains, in both performance and customer satisfaction.”
Youwe's experts are always happy to help you navigate the data and learnings collected during the peak season.
AIGrunn
Groningen is quickly becoming one of Europe’s most exciting AI hotspots.
In November, it hosted the latest edition of the AIGRunn event, with innovators, policymakers, and digital pioneers coming together, showcasing the northern region's bold vision for an AI-powered future. Youwe was a proud Gold Sponsor of this event, and two of our experts took the stage with their speaking sessions, covering the topics of local AI and the ethical questions of it.
If you are interested in the AI topics, join our AI Café, an initiative for enthusiasts to share tools, experiments, demos, and real-world challenges.
The Foundations of Trust in AI
Trust in AI must therefore be carefully designed and cultivated. The AI system must perform its intended task accurately, safely, and in a way that users can rely on. But even a highly competent system will fail to earn trust if it behaves unpredictably, which is why consistency is the second essential element. The final element, care, is what gives AI systems a human-centered foundation. Care means the system protects the dignity, safety, and well-being of the people who interact with it. Thus, competence, consistency, and care form the basis of trusted AI, and they map onto the expectations users have around accountability, reliability, privacy, and fairness. If even one of these expectations is violated, the entire trust relationship becomes fragile.
Designing Values
AI systems do not come with an inherent understanding of human norms, culture, or context. They reflect whatever is intentionally, or unintentionally, built into them. This means engineers and designers must consciously embed values into the system. These values come from several sources: your own ethical position, the values your clients hold, the expectations of end users, and the needs of indirect stakeholders who may be impacted without directly interacting with the system. When these perspectives are ignored, organisations quickly accumulate what can be described as ethical debt, the growing risk that an AI system will behave in harmful, biased, or legally problematic ways. Ethical debt, like technical debt, becomes more difficult and expensive to fix the longer it remains unchecked.
Ethical AI
Ethical AI requires moving beyond high-level principles and embedding responsibility into every layer of the system’s architecture. De Vreede explains this, “At the presentation layer, this includes designing interfaces that provide explainable outputs and clear user guardrails, giving people insight into how and why the AI arrived at a recommendation. In the business layer, teams must incorporate feedback mechanisms that detect bias and maintain responsible handling of prompts and decision flows. Meanwhile, the data layer must uphold transparency and accountability through detailed logging, robust security measures, meticulous tracking of data origin, and strong governance practices.”
Team Effort
De Vreede highlights that creating responsible AI is not the job of a single specialist, but a collaborative commitment of the entire product team. UX designers ensure that interactions are understandable and respectful. Developers and AI engineers implement safety mechanisms, explainability tools, and bias detection. ML Ops engineers monitor models after deployment, ensuring they continue to behave responsibly. Lead architects define the structures that prevent ethical faults from cascading through the system. Product owners make decisions that balance user expectations, business value, and societal impact. QA teams test the system not only for functionality, but also for fairness, transparency, and reliability. Together, these roles form a shared ethical ecosystem around the AI.
Image after Image
Erik told us how he started experimenting with early models for image generation, first with simple instructions and then with creative tests aimed at 'cracking' the models. Along the way, Erik also trained the models on images of his own face, producing surreal results like an 80-year-old version of himself. A learning point in there: be careful what data you feed your models, as even a small error in the training data can have big consequences on the model you train.
Video, Audio, Local AI
Erik’s talk also demonstrated what’s already possible in video, audio, and what you can run locally on a personal machine.
Video animation: With the new Wan 2.2 Animate feature, you can animate any character using just a single image and your own movement. It works for different styles (e.g., anime, robots).
Real-time face swapping: Local Face-Swapping in Live Video": Deep Live Cam, using a webcam and one uploaded face image, you can transform your live video feed to look like someone else, all running locally.
Audio + video generation: Tools like Sora and V3 can generate video with matching audio. A local tool called OVI can also do both on your own device.
Voice cloning: Local Face-Swapping in Live Video": For Dutch: Coqui/xTTS-v2, For Emotion: Zonos, From a short voice clip, you can recreate someone’s voice and make it say anything, including adding emotions.
Local AI tools for coding and text: Terminal-based “agentic” tools let you chat with your files. Tools like Ollama, Opencode and LM Studio make it easy to run and test AI models locally.
Boosting small local models: Smaller models know less, but with strong system prompts and access to tools (like web search via DuckDuckGo), they can produce accurate, up-to-date results.