← All posts

Journal

Shipping enterprise data pipelines in three weeks

IguanAI Team · April 7, 2026

We are often asked how long a given piece of data or AI work should take. The honest answer is that most of the timelines we have seen quoted in our careers were three to five times longer than the work actually required. The difference was almost never the engineering. The difference was the coordination.

A recent engagement is representative. A large organization needed a production data pipeline to consolidate transactional records across several source systems, apply a handful of business rules, and make the result available to internal tools and reporting. The initial estimates from internal stakeholders ranged from four to six months. We agreed to deliver it in three weeks.

We did not compress the timeline by skipping steps. We compressed it by removing the ceremony. Week one was source-system discovery and a working end-to-end skeleton in a staging environment, deliberately simple, deliberately visible. Week two was filling in the transformation logic with the business owner in the room, reviewing sample outputs daily against a handful of reference cases the business had already validated. Week three was cutover: production credentials, monitoring, documentation, a runbook, and a handover session with the team that would own it.

What made this possible is not a trade secret. It is that we had one decision-maker on the client side, a narrow scope written down on the first day, and a willingness to ship an obviously incomplete version early so everyone could see the shape of the thing. The same discipline that makes a startup move fast makes enterprise work move fast, once you are allowed to apply it.

There are good reasons to take longer on some projects. Regulatory workflows need careful validation. Customer-facing systems need real load testing. Data with lineage implications deserves a slower, more documented treatment. But the default assumption that serious work must take quarters is, in our experience, a coordination habit, not a technical reality.

For the executives we work with, the practical takeaway is this: before you budget six months for an integration, ask what the work would look like if it had to be done in three weeks. If the answer is "unsafe," take the longer timeline. If the answer is "we would need to stop arguing about scope," the scope is probably the work.

— The IguanAI Team