An IT Ticket Takes Six Weeks. This Took Forty Minutes.

Most practice managers and revenue cycle leads have a running list of things they need but haven't been able to build. A denial tracking dashboard. A consolidated AR view by payer and aging bucket. A simple database to track prior auth submissions from open to close. A quick way to check whether a specific code combination is going to create a problem before the claim goes out.

The list sits because the path to getting those things built is either expensive, slow, or both. Hiring an analyst isn't in the budget for most independent and small-to-mid practices. Putting in an IT request means waiting weeks to get scoped, longer to get prioritized, and longer still to get built, by which point the operational moment that created the need has usually passed.

So the list grows.

What's changed is that general-purpose AI tools, primarily Claude and ChatGPT, have become genuinely capable of building the infrastructure side of these problems in a single conversation, without requiring any technical background from the person asking. Not perfectly. Not automatically. But well enough that the starting point you get in forty minutes is better than the blank page you've been staring at for six months.

This is the first issue of a three-part series. This issue covers what's possible. The next two go deeper on data infrastructure (dashboards, Excel, databases) and revenue cycle intelligence (coding lookups, denial pattern research, prior auth documentation).

Front-office and administrative tasks consume between 30 and 40 percent of staff time in small and mid-size practices, according to a 2026 analysis from Inside Out Medicine. That's not surprising to anyone running one of these operations. What is worth naming: almost none of those practices have a dedicated data analyst. The same leader managing prior auth volume, payer escalations, and staffing gaps is also expected to produce reports, track denial trends, and build the operational visibility that larger health systems pay full-time analysts to maintain.

The structural assumption embedded in most practices is that analytical tools get built by IT or outside consultants. That assumption made sense when building those tools actually required technical skill. It's increasingly out of date.

None of what follows requires coding knowledge. It requires the ability to describe what you need clearly. The example prompts below are intentionally generic. They are starting points, not solutions. Every one of them needs to be adapted to your specific systems, your payer mix, your column names, and your workflow. Treat them as a starting framework, not a script.

Dashboard shells. Ask AI to build the structure of a KPI dashboard, which you then populate manually or connect to your own exports. A starting point might look like:

"Build me an Excel dashboard shell to track prior authorization performance for a medical practice. Include columns for submission date, payer, CPT code, decision date, outcome (approved/denied/pending), appeal filed (Y/N), and appeal outcome. Add a summary section that calculates approval rate, denial rate, and average days to decision."

The output won't connect to your system automatically. You'll review it, rename columns to match your actual data, and adjust the calculations to reflect how your practice categorizes things. But the structure is there in minutes instead of weeks.

Power Query builds. Describe your data to AI and ask it to write the M-code that consolidates or cleans it. You paste that code into Excel's Power Query editor. No technical skill required. A starting point:

"Write Power Query M code that takes three monthly denial report tabs, each with columns for claim number, date of service, CPT code, payer, denial reason, and amount, and consolidates them into one table. Remove duplicates based on claim number and add a column that categorizes denial reasons as: eligibility, authorization, coding, timely filing, or other."

Your column names will be different. Your denial categories may be different. The code AI generates is a draft you adapt, not a finished product you deploy unchanged. Excel University has documented Claude building complete Power Query pipelines for multi-tab data consolidation, with evidence that it reasons through the logic rather than just producing something that looks right. That matters when you're working with revenue cycle data.

CPT and ICD-10 research. This one requires the clearest framing of what AI is and isn't doing. AI can surface published policy information: NCCI edit conflicts for specific code combinations, common denial triggers, CMS Local Coverage Determination requirements, documentation checklists. What it cannot do is account for your specific payer contracts, plan-level variation, or guarantee how a given claim will adjudicate. A starting point:

"What are the common NCCI edit conflicts for CPT 99213 when billed with CPT 93000? What documentation should be in place to support billing both on the same date of service?"

Treat the output as a research accelerator, not a coverage determination. Verify against actual payer policy before acting on it. Organizations using AI-assisted coding review have seen first-pass claim approval rates improve from 72 to 94 percent in acute care settings and AR days drop from 52 to 31 in multi-specialty groups, according to HFMA's revenue cycle research. Those numbers reflect purpose-built tools with dedicated implementation. The underlying capability, surfacing published coding policy quickly, is available now without a vendor contract.

A few things worth knowing before you start.

Don't include patient data. General-purpose AI tools are not HIPAA-covered by default. Work with de-identified data, structure-only descriptions, or column headers without actual records. HIPAA-covered use requires an enterprise-tier tool with a Business Associate Agreement in place.

The output is a draft. Every Power Query, every dashboard schema, every coding summary AI produces needs to be reviewed by someone who knows your environment. AI doesn't know your EHR's export format, your payer contract terms, or how your team has historically categorized denial reasons. That context lives with you.

You don't need to know how to code to use the Power Query output. You paste it into the Power Query editor under Data, Get Data, Launch Power Query Editor. AI can walk you through exactly where to put it if you ask.

The meaningful shift isn't that AI is doing the analytical work of healthcare operations. It isn't. The shift is that the blank-page problem, the gap between knowing what you need and having a starting point to work from, is no longer the barrier it was.

A denial tracking dashboard that would have taken three weeks to get scoped and six more to get built is now something you can have the structure of by end of day. It still needs your data, your review, and your operational judgment. But the starting point exists, and that matters when you're running a lean operation without a bench of analysts behind you.

Issues two and three get into the specifics. Issue two covers dashboards, Excel builds, and database scaffolding in depth. Issue three covers the revenue cycle intelligence side: coding lookups, denial pattern research, and prior auth support documentation.

What's on your analytical wish list that's been waiting on an IT ticket that never got answered?

Which of your team's reporting tasks still live in someone's head instead of a dashboard because there was never time to build the infrastructure?

If the blank-page problem weren't the barrier, what would you build first?

Sources: Inside Out Medicine (2026), HFMA AI and Automation in Revenue Cycle Research, Excel University / Claude Power Query documentation, Codify by AAPC, CWS Health Healthcare Staffing Crisis Report (2026)