API Integrations
The integration problem people bring me sounds operational and is really about ownership: “our systems don’t talk to each other, so we re-key the same data by hand, and the one tool that connects them charges us per task to do it.” Connecting systems so they share data cleanly is straightforward senior work. The decision that actually matters — and the one I’m really hired for — is whether you own that connection or rent it back from a platform that prices up the moment it becomes load-bearing.
The problems I solve here
The most common: two or more systems that should share data and don’t. The CRM doesn’t know what the billing system knows; the website orders don’t flow into the operations tool; someone copies numbers between them by hand every day. The job is to wire them together reliably so the data moves itself, accurately, without a human in the loop re-typing it.
The second: the integration that’s quietly become a single point of failure. A connection that breaks silently, drops records when an API is slow, or duplicates data when something retries — and nobody notices until the numbers are wrong downstream. Integrations are where data quietly corrupts, and most of them are built without the error handling to catch it. I’m brought in to make the connection trustworthy, not just present.
The third: the per-task connector bill. A business running its core data flows through a hosted automation platform, and as the volume grew so did the invoice — now paying meaningfully every month for connections that, built directly, would be a fixed, one-off piece of code they’d own. The honest answer is usually to own the integration outright.
How I think about API integrations
The first principle is that an integration is only as good as its failure handling. The happy path — both systems up, data clean, API responding — is the easy 80%. The value is entirely in the other 20%: what happens when the other end times out, returns garbage, rate-limits you, or goes down mid-transfer. An integration without proper retries, idempotency, and error visibility isn’t an integration, it’s a future data-corruption incident with a delay on it. That unglamorous robustness is most of the real work and most of what gets skipped.
The second is the build-don’t-rent call, and it’s sharper here than almost anywhere. Connector platforms are genuinely useful for quick, low-stakes, low-volume links — and a genuine trap when a core business data flow runs through them. The moment an integration becomes essential, renting it per task or per seat is renting access to your own operation. Build it directly and it’s a fixed cost you own, that you can debug, extend, and trust. The line I draw: rent it while it’s an experiment, own it the moment it’s load-bearing.
The third is that integration is where ownership of your data is won or lost. Done right, integrations are how your systems stay yours — your data flowing between systems you control, on your schema, available for whatever you build next. Done lazily, every connection becomes another vendor holding a piece of your operation hostage. I treat the integration layer as part of owning your data, not as plumbing to outsource and forget.
How I’d actually do it
I’d build the integration directly against the APIs, owned by you, with the failure handling treated as the main event rather than an afterthought. That means designing it to be idempotent so a retry can’t double-charge or duplicate a record, queuing the work so a slow third party doesn’t take your system down with it, and making failures loud — logged and visible — instead of silent. A boring, observable integration that you own and can fix beats a clever one inside a platform you can only file a support ticket against.
I’d also build it to the one flow that matters first. Rather than integrating everything with everything, I’d wire the single connection that’s costing the most manual work or the most risk, prove it solid, and add the next from there. Each one a fixed asset you own, replacing a recurring manual cost or a metered connector fee. The cash saved starts compounding immediately, which is the whole point of building rather than renting.
Where this connects
This is one of the specialisms I’m hired for. Integrations are how data gets somewhere useful — most often into the dashboards that make it actionable — and they sit right alongside automations, which act on the data once it’s flowing. The engagement this powers most directly is technical advisory when the question is which systems to connect and how, or an MVP development when it’s time to actually build the owned integration layer.
When you need a team instead
A large, sprawling integration estate across many systems, maintained continuously as APIs change underneath you, is ongoing team work. That’s my agency, Digital Royalty — the crew and process to build and keep a big integration layer healthy at scale. My lane is the senior architecture call and the lean, owned build. I’ll tell you which one your situation needs.
Who this isn’t for
If you want the fastest possible click-together connection for a core business flow and don’t care that you’re renting it, a no-code platform will do that and I’m not the right call. I’m worth hiring when the connection matters enough that owning it and trusting it is worth doing properly.
Talk to me
If you’re re-keying data by hand, or watching a connector bill climb on flows you should own, tell me about your problem and I’ll tell you which connections are worth building directly — and which genuinely aren’t.