Business process automation case study: clearer healthcare provider credentialing for providers and administrators

The challenge: keeping providers and administrators informed

A healthcare organisation credentialing its providers manages many files at once across many practices, and each provider has different information to send. Before the build, the work moved through repeated emails and personal follow-ups, with each provider's details spread across several tools. This made the process slower and made missing information harder to spot early.

Three parts of the experience caused most delays:

• Incomplete forms. A form submitted with missing information could delay an approval by 3 to 7 business days.
• Invitation issues. Links expired, providers sometimes received duplicate invitations, and emails could be missed, each of which could pause a file.
• Repeated follow-ups. In the previous process, administrators spent 20 to 30% of their time requesting pending forms and missing documents or sharing status updates.

Whatever replaced the emails had two more conditions to meet. With a portal for each side, the provider's and the practice's details had to match in both, because even a small mismatch could mean another check, another message and a slower approval. And credentialing rules change, so the system had to scale and be simple to adjust when the rules or the back-end requirements moved.

What we built: workflow automation with no AI model in it

There's no AI model in this build, and it didn't need one. Every routine step that took the administrators' time had a rule-shaped answer: is the form complete, has the invitation been used, and whose response is needed next. The automation therefore uses rules, tracked state and notifications across two portals that share one set of data.

Both portals, one for providers and one for administrators, are built with Next.js and TypeScript, so they share code and the same business logic, and both are hosted on Vercel. Supabase, on PostgreSQL, holds the data, sign-ins and file storage, and structured migrations keep the system stable as the schema changes.

Four pieces now provide the structure and updates that previously depended on email:

• Forms in steps, with validation. A provider fills in the submission step by step, and each step has its own checks, so a gap is caught while the provider is still there to fill it. Incomplete submissions fell by about 25 to 35%.
• One set of data models behind both portals. Admins and providers see the same profile and practice details, so there's no second copy to drift.
• Invitations as tracked links. Each invitation is a secure link carrying a token: tracked, set to expire, and resent only under controlled rules.
• Live notifications. Each side hears when an action is pending, when a detail is updated and when a status changes, so fewer people write in to ask. Status-check messages fell by around 35 to 50%.

Where the automation stops, and who takes over

The platform collects information, checks entries, sends reminders and shows where each file stands. The review belongs to the credentialing team.

Before the build, administrators handled the follow-ups between steps by email. Now more of their time goes to thoughtful review, and the file in front of them is more often complete.

Results

Since launch, the process has run in set steps, and it's easier to manage: each file's status shows as it moves. None of these gains needed a model. They came from catching gaps on the form, keeping one copy of the data, and telling people what's waiting before they have to ask.

Technologies and tools

Timeline

The build took ten months.

If your own process depends on reminder emails, our page on business process automation services explains how we'd support it, with AI only on steps that need judgement and plain code on the rest. For new customers rather than providers, customer onboarding automation covers the same collection and follow-up work.