The hidden cost of AI integration

I was listening to Ali Ghodsi, the CEO of Databricks, tell a story about building a connector himself.

He had used AI to put together a working proof of concept: an integration from Databricks into Salesforce, Workday and NetSuite. Then he went to his team and asked the question any vibe coding CEO would ask. Why would it take them nine months to do the same thing? If it is so easy to use AI to code, why is the work taking so long, and why can't we get to market quicker?

It is a fair question, and the answer is worth knowing if you work anywhere near a system of record.

What the prototype was missing

Some of the answer was the usual list. The code he had created worked, but it was not really scalable. Enterprise grade security was missing. That is typical of vibe coding: you prototype quickly, and the parts that make something production ready come later.

The bigger insight was somewhere else. Building the interaction with the external systems was never really about using the APIs or populating data into Salesforce or Workday. Anyone can use a publicly available API to put records into HubSpot or Salesforce. You could probably get that working in a day.

What held them up was understanding those systems properly, and the implications of what they were doing inside them. The team did not have that expertise.

We have seen exactly this at 1KE. We have similar integrations in our own product, and we have had to learn a lot about some complex external systems. Building the connection was the smaller job. The work was understanding what it did to the workflow and the value creation inside the other system.

What Databricks did about it

Their answer was to hire experts in Salesforce and in Workday, so they could understand the implications properly and make the right decisions about how Databricks would interact with those systems in a robust way. Those people ended up involved in the whole solution, not just the connector.

And this is the part I like. AI did help. Using it to code reduced the time. But the knowledge gap in the external systems was big enough that they did not get the full dividend until those experts were on the team. With them, they got to a point where they were still faster than the original estimate. Without them, they would never have got there.

What it means for the channel

A lot of channel partners are hearing the same thing from customers right now: we are going to build it ourselves, we have some internal capability. We hear it too.

If you have expertise in the core systems that still sit at the heart of most businesses, the systems of record, that expertise is going to be necessary for all of these projects. As we have said many times before, judgment, knowledge and experience are not replaced by AI. Effort is. Your experience with Workday, HubSpot or Oracle is just as useful and valuable as it ever was in these new agentic workflows.

Those core systems are not going anywhere. The data foundation is still fundamental. So if you are an expert in one of them, you have the opportunity to extend your capabilities into AI and agentic work. Perhaps you will do some of that work yourself. At the very least you will be necessary to every other integrator or internal team that wants to build on top of it.

The conversation to have now

Take the Databricks story at face value. They had strong engineers, and they were using AI, and they still needed to bring in consultants for knowledge those engineers did not have.

So whether your customer is working with internal builders or a third party that has come in to help, pay attention. Understand what they are trying to build, and what your expertise could do to help it. You may find an opportunity inside a project you had written off as work someone else was going to do.

Watch the full conversation with Sidney on YouTube: https://youtu.be/g-LrpRcCwR0

Subscribe to The AI Take for one AI lesson a week, and what it changes for your business.

About 1KE:

1KE is the presales autopilot for technical sales teams. It turns discovery calls, emails, and technical documents into structured requirements and customer-ready solutions in minutes, so teams can respond while the deal is still warm.

Learn more → https://1ke.ai

Michael → https://linkedin.com/in/michaelchanter/ · https://x.com/1keceo

Sidney → https://linkedin.com/in/sidneyminassian/

Follow 1KE → https://linkedin.com/company/1ke-ai/

‍