Building the Data+AI Advantage

We’ve all heard it a million times now, “data is the moat for AI.” But what does that actually mean? What would it look like to actually harness the power of data with AI? 

That’s exactly what we’ll break down in this post.

The Reality of Today's AI Landscape

The AI revolution is well underway. Companies are transforming themselves to become AI-native, in how they build products, run the business, and leverage data for insights and automation. 

Companies are racing to build agentic workflows - automated support, document processing, sales assistants, and more. We’re getting pretty far with zero-shot, basic prompt engineering, and RAG pipelines, but we’re quickly approaching a plateau where the novelty of LLMs is wearing off - especially with user-facing features. 

Meanwhile, OpenAI and Anthropic are playing a different game. They're not just improving models - they're building data moats. OpenAI is creating connectors to every enterprise tool. Anthropic launched MCP to democratize context sharing. They know that whoever holds the most context wins.

As Brian Balfour notes in the next great distribution shift,  

“OpenAI has identified their moat, and it's not the model quality. It's context and memory.

Think about what makes an AI assistant truly valuable. It's not just answering questions—it's understanding your specific situation, remembering your past conversations, knowing your preferences, and building on previous interactions. The AI that knows you've been working on a product launch for three months and can reference your previous discussions is infinitely more valuable than one starting from scratch each time.”

If your company has missed the importance of this, it’s not too late. While most companies won’t be building their own foundation models, there’s a very real opportunity to build a data moat by mastering the intersection of 3 competencies:

1. Connect data to the AI-powered apps and workflows you’re already building
   More data means more context, which means better AI results
   

2. Expand your data foundation with frontier data and agentic workflows
   Build a 10x data platform using agents, unlocking the ability to bring in frontier data
   

3. Build next gen data products using AI to level up your analytics and data science
   Leverage AI to help you munge and explore data more deeply, delivering more robust ML models, data products, and analytics

1) Applications and Workflows

More data → more context → better AI products

Let’s start with the work that most teams are focused on today - building AI-powered apps and workflows to improve customer service or manual processes. These typically look like AI chatbots or agentic workflows built off MCP servers that aim to automate tedious tasks. 

In building these, teams are pulling basic customer data from their data warehouse or productionDBs, storing them in vector databases and providing predefined, yet limited context via RAG. But can we do more? 

This is the breakthrough for your AI-powered applications. When your support chatbot understands how to respond like your best support agents (because you’ve collected that data), it can deliver superior service over the generic LLM chatbot. When your model knows everything about your users’ preferences (because you’ve been accumulating data about them), it can better predict what they need for a delightful experience. 

More data means more context, and more context means AI that actually understands your specific business, your users, and their needs. 

So, how do you make that happen? Well, this takes us back to data foundations and your strategy around collecting and cleaning datasets. 

2) Data Foundations

Frontier data collection and agentic data pipelines 

This is the platform that collects, cleans, and serves all your data - structured and unstructured. In yesteryear, this was a data warehouse primarily for analytics. Today, it’s the cornerstone of your AI advantage. 

You’ll want strong fundamentals - a modern data stack that reliably manages your data pipelines with quality and compliance in mind. 

From there, it’s important to start thinking about frontier data collection - what data isn’t accessible today that would be valuable context for your AI efforts?

This could be data from your enterprise that you aren’t yet harnessing. Can you bring in chat data, CRM data, or even documents that might have valuable context?

This can also be data that you collect by building features that capture user intent data (ex: goals setting) or data from partnerships. Focus on data that tells you more about your users that isn’t captured right now. 

But as your platform scales to more datasets, especially unstructured ones, the maintenance cost goes up too… unless you build agentic workflows to automate the heavy lifting for data ingestion, normalization, cleansing, and quality checks. 

Imagine having an agent check for new datasets and automatically normalizing them and adding them compliantly to your core tables. In the age of AI, you can utilize agents to deliver a data platform experience that rivals the Mag7 with a fraction of the team. 

With a wider array of data at your fingerprints, this opens up opportunities to get more out of your analytics and data products - the final leg of your AI data advantage.

3) Data Products

Using AI to build better models, better analytics, better datasets

Imagine your existing ML models and analytics workflows, supercharged with deeper insights and agentic automation. 

Instead of a team of data scientists working for 6 months to deliver a churn prediction model that’s “fine,” you can have a smaller data team deploy agents to handle the data munging, exploration, and have them parallelize lots of tests to see which models perform the best. Then, add the rich data you’ve collected in your foundation layer – feed it support sentiment, usage patterns, and billing signals instead of just demographics. In a fraction of the time and cost, you can have a more powerful model with deeper insight. 

Your resulting models, analytics, and commercialized datasets can be supported with higher confidence and a lower maintenance cost. 

• Context-aware AI can help your analysts perform deeper analyses faster than before by parallelizing analyses, fine tuning parameters, and handling pre and post processing for you. 

• With performance marketing, you can finally manage and understand your return on ad spend (ROAS) and customer acquisition costs (CAC) to the level of detail you want. 

• Deliver more robust data science solutions such as churn models, personalization models, even experimentation with a fast and flexible agentic data science platform

What’s more, LLMs are great at translating technical SQL into human readable semantic knowledge. This can finally make self-service analytics a reality by translating SQL pipelines to comprehensible business logic that empowers end users to trust data and do their own analyses.

This final leg of the AI data advantage helps you change the way you understand and run your business. It’s not to be missed as part of your moat. 

The Path Forward

When we put this entire picture together – an agentic data platform that houses deep and rich context about your business and users, a high powered analytics and modeling arm, and differentiated AI user experiences powered by your rich data library – we can really see what it looks like when you create a flywheel between data and AI. 

Realizing this entire vision requires a whole library of new agentic data enablement tools and a long-term strategy that prioritizes collecting the right data that drives long-term differentiation. 

But there are a multitude of short term wins along the way. Here are a few examples: 

• Build a compelling product feature that better captures users’ thoughts as they use your product - this helps capture not just what happened, but why

• Bring in unstructured data (CRM, chat, documents) and make it joinable to the rest of your customer data 

• Establish an agentic data quality and governance layer that checks for data defects and potential vulnerabilities so that you can use your data for AI and analytics with peace of mind 

• Provide your analytics end users with an “AI translator” that looks at your data catalog and SQL to help less technical users understand the logic behind metrics

• Cursor-style analytics tool that 10x’s your analysts and enables your business users to build their own analyses

At Moda Labs, we help companies build this data-to-AI flywheel. We assess where you are, identify quick wins, and deliver production software that turn your data into competitive advantages. 

We’ll continue with deep dives into this framework as we continue exploring the intersection of data and AI. We’re excited to have you join us on this journey 

Reach out to luke@modalabs.ai to learn more