Anyone using AirOps in production? Real feedback after 2 months

After seeing the buzz around AirOps, I decided to shift a chunk of our content workflow over to it about two months ago. We were using a patchwork of Make, a bit of custom GPT, and some manual steps. The promise of "SQL for AI workflows" really appealed to my tinkering side.

I've been using it primarily for two production flows:

1. **Product Description Batch Generation:** We feed it a CSV of product specs (features, target audience, keywords) and use a custom template to generate a first draft. This then gets reviewed and pushed to our CMS via a webhook.
2. **Support Email Triage & Draft Reply:** A more complex workflow where incoming support emails to a specific mailbox are parsed, classified by urgency/topic using an AirOps "classification" module, and a draft response is generated with links to relevant docs.

The good stuff is real:
* The **template editor** is fantastic. Being able to lock certain variables, chain prompts, and test iterations quickly is a game-changer compared to juggling multiple ChatGPT tabs.
* **API reliability** has been solid. The webhook integrations have been firing without a hitch, which is more than I can say for some other services I've tested.
* The ability to use **your own OpenAI key** keeps costs predictable and transparent.

But, there have been friction points:
* **Debugging complex logic** can be tricky. When a workflow with multiple steps and conditional branches doesn't output what you expect, tracing the exact point of failure isn't as intuitive as in a traditional node-based editor like n8n.
* **Iteration speed vs. complexity:** For simple, linear prompts, it's incredibly fast. But when you need to add a loop or more advanced data manipulation, you sometimes have to get creative or drop into a code step, which breaks the low-code flow.
* **Cost model:** While using our own key is great, the platform cost itself adds up. You start wondering if you could rebuild the core "template + API" logic in a simple script for certain static workflows.

Here's a simplified snippet of a product description template I'm using:

```jinja
Product: {{product_name}}
Key Features:
{{features | join(', ')}}
Target Audience: {{audience}}

Write a 150-word product description that highlights {{primary_feature}} and incorporates the keywords "{{keyword_1}}" and "{{keyword_2}}". Use a {{tone}} tone.
```

The output is consistently structured, but almost always requires a human pass for brand voice and to trim overly generic phrases. It's a solid first draft machine, not a finished copywriter.

My verdict so far: It's a powerful tool for **standardized, repeatable AI writing tasks** where you have a clear template and input structure. For more dynamic, decision-heavy workflows that need to interact with many other APIs, I still find myself leaning back to n8n. I'm curious if others have hit similar walls or found clever ways to bend AirOps to more complex use cases.