I recently built an integration to check whether customer contacts exist on WhatsApp directly from Salesforce.
Starting out, I expected Salesforce and writing code in unfamiliar Apex to be the hard part. It turned out to be the exact opposite.
Coding in Apex with AI
I had never written Apex before. Usually, picking up a proprietary language on a massive enterprise platform means hours of reading setup guides and getting lost in unfamiliar menus.
With Claude as my pair programmer, that friction vanished. It showed me where everything lived in the Salesforce setup menu, and the Apex code itself was straightforward. It is standard object-oriented syntax, and AI writes it cleanly on the first try.
Writing the logic took under an hour. The rest of the week was a different story.
The Meta Configuration Maze
Around 90% of the project time went into Meta Developer settings.
Configuring the WhatsApp Business Cloud API meant dealing with an endless list of permissions, system user roles, access tokens, and app review requirements. The dashboard frequently threw random errors or reloaded pages without saving changes, leaving me to guess what actually broke.
And that was just in the test sandbox. Getting a full production app approved with live phone numbers is an even bigger administrative hurdle.
Key Takeaways
- AI makes language syntax irrelevant: Unfamiliar languages like Apex are no longer a barrier when you pair with modern AI assistants.
- Third-party platform friction is the real bottleneck: Always budget the majority of your integration time for platform permissions, app approvals, and token management rather than writing code.
- Test early in staging: Meta’s permissions behave differently across developer, sandbox, and live business accounts. Test access roles as early as possible.
ilf.studio — AI-native web studio, Gdansk, Poland.
