RockPort Cos. / AI capability
A team that can
build its own systems.
RockPort Cos. is an Austin-based energy company across oil and gas. Their teams were already using AI. What they did not have was a shared way to use it across the company.
- 01Understand
- 02Structure
- 03Enable
- 04Build
The starting point
Adoption had moved faster than structure.
RockPort’s people were already putting AI to work on real company problems. Operations, land, engineering, and finance had each started using it in different ways.
The methods, prompts, and context behind that work stayed with the individual who created them. Similar problems were being solved repeatedly, with no shared approach to how AI should be used or what it could access.
01 Understand
We started with how the company was already working.
Discovery spanned RockPort’s operating companies and the external IT partner that manages their identity environment. We interviewed people across land, reservoir engineering, operations, and finance, mapping the tools already in use and the information they touched.
We understood the operation before designing the AI layer.
02 Structure
We designed a shared way to work with AI.
One operating framework gave the company a common foundation while giving individuals room to work. At the company level, a shared repository for business context, standards, reusable skills, prompts, templates, and governance. At the individual level, a structured workspace organized from defining the problem through execution and finished output.
Approaches that prove out get named, documented, and moved into the shared repository, where the rest of the company can use them.
03 Enable
We turned the framework into a working capability.
The framework was implemented inside RockPort’s own environment and was in use before the engagement ended. We trained employees against live RockPort work through two workshops, both recorded so the material could keep supporting onboarding.
A method adapted from software engineering carried the work: every piece of work begins with a specification, a design document, and a task list that makes progress visible.
- Financial actions require human approval.
- Destructive changes require confirmation.
- Uncertainty is stated rather than guessed past.
Governance was built into the same environment: an AI use policy, data classification on RockPort’s own information, identity-based access through the existing directory, and an offboarding process tied to controls already in use.
The result
They build their
own tools now.
The change that mattered was in what RockPort’s people became capable of doing.
CohesionLabs did not simply deploy AI for RockPort. We helped the company develop the internal capability to keep building with it. Team members with limited software development experience now specify, build, and run internal tools around their own operational needs.
RockPort’s land manager, with no software development background, built a working version of the company’s land administration system himself. Working from his own understanding of how land administration actually gets done, he had an initial version running in about thirty minutes, then kept extending it.
On the first asset he loaded, the system holds that asset’s leases, extracts and summarizes each lease’s provisions, and places every extraction into a review queue where a person approves or corrects it. It also calculates the ownership interests that feed accounting. In one test it reviewed a roughly 275-page title opinion and surfaced three issues in about three minutes; the calculations were worked by hand to verify them, and counsel confirmed all three needed correcting.
This first version proves it: a working version of a function RockPort pays a five-figure monthly retainer to run was built in-house for about $100 in API usage, by someone with no software development background. That recurring cost is now something the company can begin bringing in-house.
The goal was not to make RockPort dependent on CohesionLabs.
It was to build enough structure, understanding, and internal capability for the team to keep operating and building on its own.
The decision
Software decisions were constrained by what existing products could accommodate.
RockPort can buy when the problem is commodity and build when the workflow is strategically distinctive.
Build or buy is now a real decision.
CohesionLabs has been a great partner in helping us develop our AI capabilities. As a group with limited coding experience, instead of learning to ride a bike on our own, with the trips and falls along the way, they gave us training wheels and got us pointed in the right direction quickly. Now we are developing our own tools the way a software engineer would, and in the way AI systems want to be worked with.
What came next
The foundation opened the next stage.
With a shared operating model and governance in place, RockPort could begin connecting AI to the systems the business already runs on: a governed platform layer that connects operational systems and supports production workflows against company data.
Technology consult
Start with a conversation.
If your organization is trying to introduce AI, strengthen how your systems work together, or build technology around the way the business actually operates, start with the problem.