Inside the Model Factory: Nihan Yami on Scaling AI with Business Impact
What does it take to move AI from isolated experiments to enterprise-wide impact? At the Business Analysis Conference Europe 2025, Nihan Yami—Data Scientist at SAS Institute—will explore just that. Her session, “Building a Model Factory,” focuses on how organisations can streamline machine learning using tools like AutoML, feature stores, and proactive monitoring to drive real business value.
Ahead of the conference, we caught up with Nihan to discuss the rise of Revenue Operations (RevOps), the realities of operationalising AI, and the evolving role of business analysts in data-driven environments. From churn prediction to model drift, she shares actionable insights for any analyst or BA navigating today’s fast-changing AI landscape.
1. Your session focuses on RevOps and AI — can you briefly explain what Revenue Operations is and how it intersects with business analysis and data science?
Revenue Operations (RevOps) is about aligning all revenue-generating functions: sales, marketing, and customer success from the ground up to eliminate silos and drive smarter, faster decisions. Analytical teams at RevOps provide the C-suite with insights using both internal and external data streams from CRM, ERP, and sales systems. This creates a unified data view, which is ideal for data science projects.
We work across the entire sales lifecycle from optimising lead generation to identifying upsell and cross-sell opportunities post-sale. Depending on the complexity, we may use anything from simple linear models to advanced architectures like RNNs. What excites me most is that these models truly influence strategic decisions.
2. You’ll be introducing the concept of a “Model Factory” — what does that mean in practical terms, and why is it important for scaling AI?
A “Model Factory” is a practical implementation of MLOps that enables mid-to-large organisations to scale their AI efforts efficiently. As the number of models grows, so does the complexity of managing them both technically and operationally.
The Model Factory approach focuses on three key pillars:
- Feature Store: Creating reusable, centralised engineered features
- AutoML: Accelerating model development with automation
- Monitoring: Continuously tracking performance to detect and address model drift
This approach boosts repeatability, reduces manual effort, and makes AI production-ready.
3. Your session includes real-world examples like churn prediction and customer segmentation. Which of these use cases has had the most impact in your experience?
A model’s impact depends less on its technical performance and more on how well it aligns with the business need. I always ask:
- Have we clearly defined the business problem?
- Does the model solve it meaningfully?
- Can we explain the results to a non-technical audience?
- How are the results being used and by whom?
For example, churn prediction is a powerful use case, but only if customer success teams can act on the insights. Business impact is the true benchmark of model success.
4. What are some common challenges organisations face when trying to operationalise machine learning in business functions and how does your approach help overcome them?
Three key challenges I’ve observed:
- Low reusability – Many teams end up reinventing the wheel. Our Feature Store addresses this by promoting the reuse of validated features across multiple projects.
- Poor collaboration – Shared repositories for both code and documentation, along with version control tools like GitHub, ensure better teamwork and transparency. Using accessible spaces like SharePoint also helps business stakeholders stay in the loop.
- Lack of documentation – Not just the code, but the thinking, assumptions, and logic behind the model. With thorough documentation, projects can be handed over or revisited much more smoothly.
5. How do you see the role of business analysts evolving in AI-driven environments like the one you’re building at SAS?
Analysts should keep a critical, business-oriented mindset. The tools and platforms are getting better each day, but the real value comes from understanding the business challenges.
Analysts are vital in bridging the gap, translating analytical findings into clear, actionable narratives. That’s how AI becomes relevant and trustworthy. A compelling story is more powerful than just a model output.
6. You mention proactive monitoring of data and model drift. Why is this such a critical part of the AI lifecycle?
Building models is exciting, but deployment and monitoring are just as essential. Deployment means making the results usable via dashboards, automated actions, or other applications.
Once a model is in production, we need to monitor it to ensure it still performs well. As the world changes, so does the data, and models can become outdated. We track performance metrics and set up alerts to catch degradation early. This helps us retrain or rebuild proactively, keeping the model relevant and trustworthy.
7. What’s one tip you’d give to BAs or analysts looking to build stronger bridges with data science or AI teams?
We live in a fast-paced world of continuous innovation. It’s easy to feel overwhelmed, but instead of treating learning like a sprint, think of it as a long-term habit.
Stay curious, develop a sustainable system to keep up with trends, and connect with others in the field. Collaborating, sharing ideas, and learning from experienced peers can open doors to new ideas and accelerate progress.
8. What excites you most about how AI is being integrated into day-to-day business decision-making?
I’m excited about how AI is becoming more accessible. While complex environments still pose challenges such as explainability and accountability, we’ve made significant progress.
Today, you don’t need a PhD to build impactful models. With low-code tools, explainability techniques, and large language models, more people can participate in the innovation process. This democratisation of AI means decision-making can become more data-driven, inclusive, and creative. It’s a thrilling time to be in this space.
Join Nihan at BA2025 and discover how to make AI not just possible, but practical.
🗓 Tuesday, 16 September 2025 | ⏰ 10:25 AM – 11:10 AM BST
📍 Convene 133 Houndsditch, London EC3A 7DB
🎤 Building a Model Factory: Streamlining Machine Learning for Business Success
