Skip to main content

How much does a custom internal operations platform cost?

A custom internal operations platform for a document-heavy firm costs $40,000 to $150,000 for the first production release, and most firms then keep a small engineering team on retainer to extend it. Smaller builds that replace one spreadsheet-and-email workflow land at $15,000 to $40,000. Platforms that run a regulated, multi-step process end to end (intake, review, approvals, payouts, reporting) run $150,000 to $400,000 over the first year.

Those are the ranges we see across the lending, credentialing and healthcare-practice systems we’ve built. The rest of this page explains what moves a project between them, and what to build first so the first number stays small.

What "internal operations platform" means here

A system of record your team works in all day: the place where a loan file, a credentialing application, a patient case or a rental application lives from arrival to close. It usually replaces a CRM that was bent out of shape, three spreadsheets, a shared inbox and someone’s memory. It typically includes:

  • Intake from email, scans, portals and forms, with documents attached to the right record
  • A pipeline with stages, owners, due dates and approvals
  • Integrations with the systems that already hold truth: Salesforce or HubSpot, your practice or loan-management tool, accounting, e-signature
  • Reporting that leadership opens without asking an analyst
  • Audit trails, roles and permissions, and the compliance logging your industry expects

The five things that set the price

1. How many systems have to agree. Every integration is a project inside the project: authentication, field mapping, sync direction, deduplication, retries, and what happens when the other side is down. One integration adds $8,000–$20,000. A connector framework that makes the fourth integration cheaper than the first is worth it once you pass three; we built one for Fundz that carries signals into Gmail, HubSpot and Salesforce from the same code.

2. How much of the work is documents. If intake arrives as PDFs, scans or handwriting, you are paying for OCR, document classification, field extraction and a human review step for the exceptions. Done well, this is where AI earns its place: a model reads the application, a person checks only what the model flagged. Budget $20,000–$50,000 for a document-AI intake pipeline with the review queue, more if the document types vary widely.

3. Compliance and auditability. Lending, healthcare and anything touching PII needs an audit trail on every status change, field-level permissions, retention rules and sometimes third-party due-diligence calls that must be traceable later. For Samson Funding’s working-capital lending platform, the due-diligence API calls were built for retries and traceability from day one, because a funded loan has to show why it was approved. This adds 15–30% to a build and is the worst thing to retrofit.

4. Reporting and the data model under it. Dashboards are cheap; the data model that makes them true is not. Hoopes Vision had years of surgical data in spreadsheets; the cost was in normalising it into one database with ingestion, transformation and reporting kept separate, so a new data source never breaks an existing report. Expect $15,000–$40,000 for a reporting layer with a clean model, less if your data is already structured.

5. Who has to use it. An internal tool for twelve staff needs less polish than a client-facing portal. nCred’s credentialing portal shows live progress pulled from Salesforce and lets clients upload documents themselves so files stop arriving by email. Client-facing adds design, onboarding, self-serve flows and support tooling, roughly 30–50% on top of the internal version.

A realistic first-year budget

Phase What it covers Typical range
Scoping One call, one written scope, architecture and risk list Included
First release (8–16 weeks) Core pipeline, one or two integrations, basic reporting, roles $40,000–$150,000
Retainer (monthly) Dedicated engineers who own the platform, ship weekly, absorb integration and model updates $6,000–$25,000 / month
Year-one total First release plus 9 months of retainer $95,000–$375,000

The retainer is not maintenance. It is where the second integration, the document-AI step and the reporting you didn’t know you needed get built, at a lower cost per feature than the first release because the team already knows the system.

What to build first so the number stays small

Start with the one workflow that hurts most, built end to end, on a data model designed for the rest. For most firms that is intake: get documents and requests landing in one place with a stage, an owner and a due date. Skip the dashboard for the first release; a filtered list view that leadership trusts is worth more than a chart they don’t. Leave the third integration for the retainer.

This is how we run every engagement: a fixed-scope first project that is the entry point, not the whole relationship, then an ongoing engineering retainer. Most clients convert to the retainer and stay for years.

When you should not build custom

  • Your process is close to standard and a vertical SaaS covers 80% of it. Buy it and integrate around the gaps.
  • Fewer than five people will use it daily and nothing about it is regulated.
  • You cannot name an owner inside the firm who will make decisions weekly during the build.
  • The pain is one report. Build the report, not the platform.

If two or more of those are true, a custom platform will cost more than it saves. If none are, the spreadsheet is already costing you more than the build.

Frequently asked questions

Why not just customise Salesforce or HubSpot?
Often you should, and we do: nCred’s portal runs on Salesforce data. The cost of customising a CRM climbs when the process is document-heavy, multi-party or regulated. At that point the CRM becomes one integration into a system of record built around your workflow, instead of the other way around.

How long does a first release take?
Eight to sixteen weeks from signed scope to production, depending on integrations and document volume. Anything quoted under six weeks for a multi-integration platform is a prototype.

Where does AI fit, and what does it add to cost?
In document intake, classification and extraction, with a human checking exceptions. Models are the cheap part; the integration, the review queue and the monitoring that keeps accuracy up are the cost. A document-AI intake step adds $20,000–$50,000 to a first release.

What does the monthly retainer include?
A small dedicated team that owns the platform: weekly releases, new integrations, model and dependency updates, monitoring and fixes. Hours are scoped per month and roll into a written plan you approve.

Do we own the code?
Yes. The platform runs in your cloud account, under your repositories, with documentation written for the next engineer, whether that is us or someone else.

Related answers

Send us the workflow that hurts most

We’ll tell you what we’d automate first, where a model belongs, where a human check stays, and what the first release would cost. One call, one written scope. Start a conversation.