Your PSA Data Is Ready for AI. Are You?

AI can already do meaningful work on the data your PSA system holds. Whether that work is any good still comes down to your people, your process, and you.

I use AI frequently. Across my work and plenty outside of it, and not as a novelty or reluctantly. It has changed how fast I move through requirements documentation, data migration, the parts of this work that used to eat whole afternoons. So when I say what I am about to say, I am not the person in the corner warning you off the new thing. I am the person telling you it is further along than you think, and that most services organizations are leaving value on the table.

Here is the thing I keep running into: people assume AI cannot touch their PSA data because their platform does not have a built-in connection to it. That is the wrong conclusion from a true premise.

The connection question

Some platforms are building native AI integrations directly into the product. NetSuite, for instance, is moving in that direction, toward the kind of standardized connection the industry is starting to call MCP. SuiteProjects Pro does not have that native integration today. But "no native connector" does not mean "no path." It means a different path.

Your SuiteProjects Pro data can come out of the system through several well established routes: the analytics connector and its OData feed, scheduled report exports, structured data exports. None of this is exotic; it is standard platform capability that operations teams already use for reporting. Once that data is out and landed somewhere a model can work with it, AI can do a great deal with it.

If your organization has the development bandwidth, it's worth exploring building that bridge yourself. The data routes already covered here, the analytics connector, the OData feed, the structured exports, give a homegrown MCP connector a solid foundation from which to work. It's an option worth knowing about, even if you're not ready to act on it yet.

The barrier most people imagine is a connector. The real barrier is knowing what to do once the data is out.

Where you land that data matters, and this is where a lot of organizations are headed: a lakehouse.

A lakehouse, such as the MS Fabric Lakehouse, brings structured and unstructured data together in OneLake. The data can be transformed in the lakehouse using the medallion architecture into a “Gold” business level data standard.

Once your data is there, an AI layer on top of it can surface trends across projects, flag anomalies you would not have caught by eye, draft the narrative for a reporting cycle, and produce the visualizations that used to take someone half a day. That is not theoretical. It is available now, and it is improving at a rapid pace.

So the technology question is mostly settled. The data can get out. AI can work it. That part is here.

The part that is not settled is everything around it.

People, process, and now AI

I have always believed that a services organization runs on three things working together: its people, its processes, and its technology. AI does not remove any of the three. It pours fuel on all of them, which means it makes a strong foundation stronger and a weak one fail faster.

Technology is the part AI changes most, and most visibly. The analysis that used to require a person and a spreadsheet and a quiet afternoon can increasingly be done in minutes. I am not romantic about this; the acceleration is significant and I lean into it. If you are not finding ways to put AI to work on your operational data, you will be at a disadvantage to the organizations that are.

Process is the part AI cannot fix for you. This is the one I want services leaders to sit with. AI will analyze whatever process you hand it, faithfully and confidently, including a broken one. If your project setup is inconsistent, if your billing and recognition rules are applied differently from one project to the next, if your bookings are not maintained, AI will not catch that the foundation is wrong. It will give you a beautifully formatted, completely unreliable answer, and it will sound just as sure as it would if the data were clean. The process work still has to happen first, and it is still human work.

People are the part that decides whether any of this is worth anything. AI does not know what your organization is trying to learn. It does not know which metrics actually drive your decisions, what “good” looks like in your business, or which question matters this quarter. Someone has to bring that. Someone has to know enough to look at a confident answer and recognize when it is the easy answer instead of the right one. That judgment is not getting less valuable as AI gets better. It is getting more valuable, because the cost of asking the wrong question well has gone up.

AI raises the floor on what anyone can pull from their data. It raises the ceiling for the people who know what to ask of it.

What this actually means for you

If you are running a services organization, the takeaway is not “wait for your platform to build a connector.” The data is reachable now. The takeaway is also not “AI will handle the analysis, so we can think less about our setup.” The opposite is true; the better your tools get at producing answers, the more it matters that your process is sound and your people know what they are looking for.

The organizations that will get the most out of this are the ones that do the unglamorous work alongside the exciting part. Clean, consistent setup. Clear definitions of what matters. People who understand both the business and the system well enough to tell a good answer from a plausible one.

AI is ready to do meaningful work on your PSA data. The question worth asking is not whether the technology is there. It is whether your people and your process are ready to make that work mean something.

I would like to hear where you are with this. Have you started pulling your PSA data somewhere AI can reason over it, or is the setup underneath it the thing still holding you back? Tell me in the comments.