5 proven ways to leverage AI for healthcare data transformations

AI as Enhancer, Not Replacer: A Healthcare Data Consultant's Playbook

I often question if I’m using AI enough for hands on keyboard tasks. A common theme of business chats in the last 6 months is talking about how I’m using AI as a data consultant.

Two weeks ago, I attended a local Columbus tech talk about AI in day to day jobs. The speaker asked the audience who has used Model Context Protocols (MCPs) before. Only two people raised their hands: one other developer and myself. It caught me by surprise.

This past week, I had a conversation with a partner consulting firm about leveraging AI tools in my daily workflows as a healthcare data engineer.

I don’t consider myself an AI expert, but I definitely push myself to use these tools to speed up mundane tasks and ideation. These talks and conversations made me realize that I heavily use AI, especially compared to other demographics and skillsets in the workforce.

In this issue I’d like to highlight five ways I’m using AI for development and data transformation initiatives with a few of my healthcare clients. My goal is to get you thinking about how to better leverage AI to speed up your own workflows and tedious tasks, while complementing your unique skill sets to build ROI-centric data solutions.

Here we go…

The Big Five

For the record, all of what I’m mentioning has been achieved using Claude Pro & Copilot. I have not tried Claude Code yet, but want to look into it as I continue to perform more data warehousing development.

1. Create product roadmap outlines

I stepped into the product management role without any full time experience. Obviously as an analyst and engineer, I’ve worked under many product managers. So I know what they’re all about.

However, I’d never curated an end to end product roadmap myself. Claude Pro helped me do just that.

When I prompted Claude, I ensured that the scope was specific. It breaks down every phase of one data product that we’re scoped to deliver. A Proof of Concept, for example.

Here’s a real prompt I gave it to start:

I am starting a project with a midsized home medical equipment company tomorrow, And I have a data architect and a data engineer that are gonna be on the project driving a lot of the hands on technical implementation of our data modernization, but, essentially, we're responsible for migrating their current EHR data from their EHR database to Microsoft Fabric. So this is actually gonna be a snowflake to Microsoft Fabric mirror, most likely, and then we're gonna be setting up our own data models in fabric, data lakehouse, from a dimensional modeling standpoint, to replicate and then validate their current existing on prem net revenue report to our migrated one. And some advice I got from a former colleague and former leader that I worked under is that if I really want to come into my own as a data strategist and data product manager, I need to treat this opportunity as if I'm a data strategist or a data product manager role. So I'd like you to craft a product strategy roadmap for this home medical equipment company that's undergoing this data modernization from on-prem and EHR driven reporting to strictly cloud-based reporting sitting on Microsoft fabric. Please make the scope specific to this 4 week Proof of Concept net revenue report we’re building.

It followed the project Phase, Epic, User Story, Business Objectives and Business Outcomes framework beautifully. This helped me lead effectively throughout developing, helping to tie back everything done to the business outcome.

2. Kickstart conceptual data models

Specific to healthcare Electronic Medical Record (EMR) systems, I have prompted Claude to list and describe all of the entities associated with an EMR’s revenue cycle model. Its output is extremely helpful in getting a baseline. I then ask it to build an actual conceptual data model diagram to kickstart our data modeling process.

Here’s a real example prompt I’ve used:

I have a healthcare client using [Vendor Name] EHR. They're a medical equipment company for home equipment and services. They distribute and provide in-home services.

We are tasked with rebuilding their Net Revenue report that breaks down their product sales by Product and Branch. It contains KPIs such as Credit Adjustment, Net Revenue, Write Off Amount, etc. in table visualizations.

Anyways, can you give me a breakdown of what the core [Vendor Name] data model looks like when dealing with financial metrics like net revenue and beyond? Then build a conceptual data model diagram highlighting all of those objects and relationships.

This is a game changer for successful business data modeling sessions with clients.

3. Refactor SQL syntax and aliases

Vibe coding with Claude isn’t my recommendation. If you’re new to a programming language, it makes me uncomfortable to have an LLM do all the heavy lifting for me. Especially when I don’t really understand everything it’s doing.

I do think AI greatly accelerates development when it’s a specific “go do this” coding task. Hey Claude, write the syntax for a Python Pandas dataframe with this CSV file.

Or in my case, go change every alias to have underscores between each word like this:

SELECT

customer AS customer_name

Instead of this…

SELECT

customer AS customername

I’ve found being extra literal like the above helps it to find and change exactly what you ask. This has saved me HOURS combined over the course of projects when you’re dealing with hundreds to thousands of fields from the EMR tables.

4. Breakdown ROI of data products

Gone are the days of writing out arithmetic to calculate labor hours equivalent to dollar cost savings. I’ve used AI to help me calculate the ROI of a healthcare data product.

Be specific on the inputs you give it such as number of hours the business currently spends on a specific task, hourly rate equivalent, the costs to build the data product, and the time savings and estimated revenue opportunities resulting from it.

There are lots of ways to go about prompting this because you’ll find a blend of quantitative (hard) ROI and qualitative (soft) ROI useful. Furthermore, know exactly which financial metrics you want to calculate.

I’d recommend starting small with a time savings to cost savings calculation. Then taking that and moving into a payback period analysis is helpful. You could get creative by telling AI to account for a revenue opportunity too based on the data product/report/insights you’re automating and augmenting respective of the industry.

5. Debug and configure Microsoft Fabric Data Factory pipelines

A little known secret is that Copilot takes actions in Microsoft Fabric. If you tell it to build a Fabric Data Factory pipeline stored procedure, it will automatically create the Stored Procedure node in the UI.

If you specify the stored procedure you want used with the node, it will configure it accordingly. It’s slick!

Furthermore, if a Power BI dashboard errors out, leverage Copilot to narrow down where the source of the error may be. I’ve found it to be particularly useful in determining if the issue is located in the backend Fabric data pipelines or the Power BI dashboard logic itself.

Keep Brainstorming

There it is. Five key ways we’re leveraging AI strategically to save time and drive more valuable outcomes for our healthcare clients. I’d love to learn from all of you in the comments. What are some ways you’re leveraging AI in your day to day workflows as a healthcare data professional?

As you may have realized by now, I’m being very cautious to use AI as an enhancer rather than a replacer. I will always keep my own thoughts and uniquely human perspectives at the forefront of my work. AI tools are just helping me structure and execute more efficiently.

There are certain components of data strategy like the people side of the business that AI cannot account for. :)

Christian Steinert is the founder of Steinert Analytics, helping healthcare & roofing organizations turn data into actionable insights. Subscribe to Rooftop Insights for weekly perspectives on analytics and business intelligence in these industries.

Feel free to book a call with us here or reach out to Christian on LinkedIn. Thank you!