Dependency on Salesforce and HubSpot APIs

This risk is the tax Ampersand pays for going deep where the money is. Ampersand is built for enterprise grade Salesforce and HubSpot integrations that do more than pull a few standard records, they read and write large volumes, subscribe to events, handle custom objects and fields, and work across many customer specific tenants. That makes the product more valuable than a shallow unified API, but it also means Ampersand has to track platform limits, schema drift, version changes, permissions, and partner policies much more closely.

• Salesforce and HubSpot are not static pipes. Salesforce enforces org level API and bulk usage limits, and HubSpot publishes breaking changes and versioning rules with formal notice windows for GA APIs. For an integration layer sitting in the middle, every platform change can trigger connector updates, retries, remapping, or customer support work.
  2 salesforce 3 salesforce 4 hubspot 5 hubspot
• Ampersand chose this tradeoff on purpose. The product mirrors underlying APIs and turns customer specific integration logic into configuration, which is why it can support things like 50 million contact syncs, many custom objects, and tenant by tenant Salesforce setups. That depth is exactly what makes upstream API change more consequential than it is for a lowest common denominator connector.
  1 sacra 6 sacra
• This is a known pattern across middleware. Unified API founders describe the work as constant gardening around deprecated endpoints and changed fields, and they emphasize that long term stability improves when the integration company has real partner relationships with the underlying platforms. In practice, the strongest moat is not just code, it is monitoring, operational tooling, and ecosystem trust.
  6 sacra 7 sacra

The market is moving toward fewer, deeper platform relationships and more operational software around them. As HubSpot rolls out date based versioning and Salesforce keeps tightening usage governance, the winning integration layers will look less like simple connector catalogs and more like reliability infrastructure for the biggest systems of record.