We are moving our professional services automation platform from ConnectWise to Halo. This is the first in a series about that migration, but I want to be clear up front: this is not a how-to for Halo. If you run an ERP, a CRM, an electronic medical record system, or any other heavy business application, the hurdles are the same. Only the logos change.
The first hurdle came before I signed anything. It was not which system to buy. It was how much of my data was coming with me.
You assume the answer is all of it. It almost never is.
We had a client move from one medical records platform to another. The new vendor promised a full migration. Then work started, and the story changed. There were terabytes of x-rays and clinical data, and suddenly the client was asked to pick a window, one to two years, and leave the rest behind. They had budgeted on the assumption that everything would be searchable and indexed in the new system. It was not.
Then came the second problem. The migration tooling only handled a narrow set of records. For anything else, the client had to hire a database administrator to read the data dictionary, decide what moved and where it landed, and build the plumbing to get it there. That is slow, expensive work, and the client paid for it twice: once in fees, once in disruption.
So I asked myself the question that engagement taught me to ask. Am I willing to pay for a single license to keep the old system running indefinitely, until I am confident that nothing will ever come up that I need to look up? Or am I confident that enough data made the trip?
Fourteen years in that platform. More than that, really, because when I bought it I imported years of history from before. Every client communication. Every quote. Every agreement. Every invoice. Anything that could ever be disputed lives in there. Losing access to that is not an inconvenience. It is a risk with very few exits.
Here is what I did instead. I connected Claude to the old system through its API. With a handful of commands, it built me a web-delivered application holding my full history. Every ticket. All seven hundred thousand-odd attachments. Categorized, so I know what is a time entry, what is an invoice, what is an agreement and the dates it covered. It did not rebuild the application. It gave me the data, searchable and indexed. That is all I needed.
It also handed me a script to run at final cutover, so I can sync everything up to that exact moment and then terminate the old agreement cleanly.
That used to be a nightmare. Practices keeping dead medical records systems alive for seven years or longer, usually longer, paying to secure an obsolete operating system just so somebody can log in twice a month to check a billing question. AI made that obsolete.
Once the archive is built, it runs anywhere. You can publish it as an executable against a SQLite database, a self-contained unit you copy from machine to machine. When Windows 11 ages out and whatever comes next arrives, you move the folder. No server. No legacy operating system. No unpatched attack surface sitting in your environment for the better part of a decade.
You can also write to it, carefully. Leave the underlying records immutable, but layer annotations on top. Say you are looking at a patient billing item that has since been paid. Tag it as paid, so the next person who opens it is not chasing something already handled. The history stays untouched. The usefulness goes up.
That was my first hurdle, and I cleared it before committing to Halo. I did not want to sign a contract and then discover that full migration would cost me $100,000 in DBA time. And honestly, I did not want the old data in the new system anyway. I wanted Halo clean, not stuffed with fourteen years of records mapped into roughly-right categories. I just wanted somewhere to go look things up.
The whole build took about 50 hours. But almost all of that was Claude working: downloading, transforming and indexing the data, and standing up the application itself. My own labor was three to four hours of conversation. Not $100,000, and not a weekend of my life either.
One note if you try this yourself. We used Claude Enterprise, which is HIPAA compliant, but you still need to be deliberate about what data you put in and where your boundaries are. Be conscious of that.
Using AI for work like this is not a small time and money saver. It can be massive.
If you want help thinking through your own migration, we are happy to talk. Call us at (918) 270-6600 or visit arclightgroup.com.
Brian Largent
Father to five, husband to one, founder, CEO, and all around swell fella (or so I'm told)
