At the recent Agentforce World Tour London, our CTO, Angelica Buffa, took the stage to outline an all-too-familiar scenario in hopes of helping audience members see just how architects can approach Data Cloud integration with strategy, structure, and long-term success in mind. Her session with PeerNova’s Mehmet Gökmen Orun, ‘An Architect’s Guide to Data Cloud Integration Patterns,’ followed Luna, a fictional enterprise architect facing the real-world challenges of connecting legacy systems, cloud platforms, and evolving business needs.
Let’s take a look through Luna’s journey and the architectural choices that can make or break your Data Cloud strategy.
The challenge of inheriting a complex landscape
Our architect, Luna, took on an enterprise ecosystem with multiple existing systems. This ecosystem included:
- Salesforce CRM: Accounts, contacts, opportunities, support cases
- HubSpot: Leads and marketing engagement data
- ERP System: Financial supply chain data
- Snowflake: Historical and transactional customer data
Luna wanted to integrate all this data in a way that’s efficient, scalable, and actually useful.
Understanding that not all data needs to be integrated
In her presentation, Angelica noted a simple truth that not everything needs to be connected by default. A default mindset aiming to connect every system and bring in all data results in high costs, unclear value, and timelines that stretch forever.
On the other hand, a strategic mindset to connect what matters and scope integrations around business outcomes is a faster, more cost-effective approach that focuses on delivering measurable results.
Strategic architecture design
Angelica shared a four-step framework that helps enterprise architects like Luna avoid chaos.
It starts with understanding what outcomes the teams are after and which data will actually support those goals. From there, integration patterns and systems can be chosen to best serve those outcomes. Then, architects can build, validate, and adjust in small phases. And finally, data quality, system performance, and ROI should be continuously evaluated.
But effective architecture isn’t just copying data. Architects need to be careful choosing what to copy, when, and how. It’s important to think about integrating data that actually makes an impact. It’s also good to keep in mind that clean data is the foundation of AI and personalization, and that poor data quality will lead you to poor outcomes. Choose patterns that can grow with your needs, and remember that not all data needs to be real-time.
Smart scoping and strategic archiving
Rather than ingesting data from every system, Luna started with just Salesforce CRM and Snowflake as they held the most business-relevant, high-quality data.
She made the important discovery that 90% of old support cases hadn’t been touched in over two years, totaling 2M records. And with that information, she decided to archive all but the most recent 6 months of support cases, retaining access via external storage.
Her decisions resulted in lower costs, faster processing, and no data loss.
Selecting the right integration pattern
For the presentation scenario, Luna opted for batch ingestion, which kept costs down and supported her performance needs. If necessary, real-time could come later.
Luna also made use of standard connectors and zero-copy access wherever possible to reuse existing infrastructure, accelerate delivery, and avoid redundant transformation logic. Angelica and Mehemet emphasized in their presentation that zero-copy isn’t always the cheapest long-term strategy, especially for frequently queried datasets. In those cases, ingestion may be more cost-effective.
Normalize, normalize, normalize
A major focus of Angelica and Mehmet’s session was data normalization. Just because two systems have a field called ‘email’ doesn’t mean the data inside is consistent or useful.
Luna introduced Interim DLOs (Data Lake Objects) to:
- Clean and reshape incoming data
- Normalize field values
- Control the timing of identity resolution
This approach guaranteed that the final mapped data in Data Cloud was clean, contextual, and activation-ready.
Summary
The presentation at Agentforce World Tour London showed us Data Cloud can be even more powerful when it’s paired with a thoughtful architecture. It’s good to start small, scale smart, and avoid moving data just because you can. Every integration should have a reason, a result, and a return.
Modelit helps companies like yours design smarter Data Cloud architectures that prioritize business value. Whether you’re considering your first integration or simply looking to improve your existing setup, our certified team is here to help. Reach out today, and let’s build something scalable together.
