Two tools. Two how-tos. One link.
Both tools you are about to see were built with AI from documents the firm already had. This page holds the working tools and the steps to build them, so you can bookmark one URL and come back to it. Move with the arrow keys or the tabs above. Each how-to follows the same four moves.
Gather
Collect the documents you already have, scattered as they are.
Write it down
State the structure or the rules once, in plain text.
Let AI assemble
AI reads the documents and builds a single working file.
Host and maintain
Put it where you control it. Feed it updates to keep it current.
Example 1: The benefits portal.
An employee-facing triage portal. The front door routes a person by their situation, the office selector filters plans, and search reads across every section. Click around inside it. Set the office in its top bar and open the Life Events page.
Generic demonstration build. Fictional firm, sample rates throughout.
How to build the portal.
- PDF Benefits 2026 FINAL.pdf, buried in an email thread
- PDF SBC_carrier_v2.pdf on a shared drive nobody opens
- PDF 50-plus carrier documents across folders, no index
- ? Employees email HR: "which plan covers my kid's braces?"
Build it in a Project
A Project keeps your source documents in one place so every chat can read them. On a paid plan it scales to a large document set automatically.
- Create a Project at claude.ai and name it for the plan year.
- Add your benefit PDFs and a short brand note (colors, font, logo) to the project knowledge.
- Ask Claude to generate the single-file portal from those documents.
- Review on screen, ask for changes, download the finished HTML file.
Reading your PDFs, organizing them into your structure, and writing the HTML, CSS, and JavaScript so the portal is one file with no server.
ChatGPT or Gemini can generate similar HTML. The real difference is how each reads a stack of PDFs. Pick the one your firm already trusts with this data.
Static file. SharePoint, an internal IIS server, or Azure. HR keeps write access and updates it. See the IT spec for the handoff.
Example 2: The licensing tracker.
An HR-facing compliance dashboard. It reads each state's CLE rule against each attorney's earned credits and computes status. Expand a row to see the requirement read against the credits and the exact gap. Look at Sofia Marchetti and Priya Anand.
Generic demonstration build. Sample attorneys and simplified rules, verify with each jurisdiction.
How to build the tracker.
- XLS CLE_tracking_v7_FINAL.xlsx, one person maintains it
- PDF Loose CLE certificates in scattered folders
- ? The rules live in someone's memory and old bar emails
- ! A multi-state attorney's second state quietly slips
Give it the rules and the transcripts
The interpretation is the whole point, so the inputs are two things: the rule for each state, and each attorney's earned credits.
- Write one rule file per state: total credits, cycle length, and each subcategory such as ethics. Claude can draft these from the bar's site for you to verify.
- Export each attorney's CLE transcript from the bar or your tracker as a CSV or PDF.
- Ask Claude to build the dashboard that reads the rules against the transcripts and computes status.
- Review the flagged attorneys, then download the file.
Matching each attorney's earned credits to each state's total and subcategory requirements, then computing what remains, the days to deadline, and the status.
ChatGPT or Gemini can build the same dashboard. Whatever you choose, keep the attorney data on a tool your firm controls.
Drop in a new transcript export and the status recomputes. Change a rule file when a state changes its requirement.
Your takeaway.
Both tools came from the same four moves: gather the documents, write the structure or rules down once, let AI read and assemble a single file, then host it where you control it and feed it updates. The hard part was never the code. It was deciding what the tool should know.
The portal
Employee-facing triage. Click in and try it any time.
Portal how-to
Structure, tools, and the two prompts to copy.
The tracker
HR-facing compliance. Expand a row to see the read.
Tracker how-to
Rule files, transcripts, and the two prompts.
Next in the series.
This was Session 4 of 5. One session remains, and it ties the whole series together: the IT foundation that lets everything you have seen actually run inside a firm.