Configure existing software when it already supports the required property workflow. Integrate systems when the problem is disconnected information. Consider a proprietary proptech platform when a distinct, validated workflow cannot be served adequately by those options and the business can support the product after launch.
What problem are you actually trying to solve?
Write the operating problem without naming the technology. For example: property managers cannot see which inspections have outstanding actions across a portfolio. That brief is more useful than asking for an AI dashboard, because it leaves room for several ways to solve the problem.
Map who creates the record, who updates it, who acts on it and who needs visibility. Distinguish between an internal operational tool, a landlord or tenant portal and a product you intend to sell. Each has different users, support needs and commercial assumptions.
When is configuration the right answer?
Review the features and permissions in the systems already in use. A different intake form, a reporting configuration or a clearer approval workflow may address the problem without adding a new application. Test the actual requirement rather than comparing broad feature lists.
Configuration is not free of operating work. Someone must own the workflow, maintain reference data, train users and check exceptions. Include that effort when comparing it with an integration or a custom build.
When does integration make more sense than replacement?
Integration is a candidate when the right records exist but people repeatedly move them between systems. Before selecting an approach, confirm access to APIs or exports, the supplier's terms, rate limits and the identifiers used to match properties, units, tenancies and jobs.
Choose a source of truth for each record and decide what happens when systems disagree. Test failed updates, duplicate events, revoked access and records that do not match. A successful demonstration of one transfer does not establish a reliable production integration.
- Name the system that owns each property, job and financial record.
- Define which fields can be changed and which users may change them.
- Make failed transfers visible, with an owner and a recovery process.
- Confirm how data can be exported if a supplier or platform changes.
What must be true before you build a proprietary platform?
Validate that a specific group of users has a material problem and that a new product would improve the workflow. For an external product, also test who would buy it and why existing tools are insufficient. An internal process that works well for one team is not automatically a market-ready product.
Agree a small first release and the evidence needed to justify continuing. Budget for identity and permissions, hosting, backups, monitoring, support, security updates and data migration as well as visible features. Confirm ownership of the code, accounts and operational documentation.
If AI is part of the product, give it a bounded role with source access, evaluation examples and human checks. Do not make the business case depend on unmeasured savings or on a model always producing the correct answer.
What should you take to a first product discussion?
Bring a workflow map, a few de-identified examples, the current systems and their owners, recurring exceptions and the outcomes you want to measure. This allows a product partner to challenge whether the right answer is configuration, integration or a build before implementation is scoped.
Questions to consider
Is a custom proptech platform always better than existing software?
No. A custom platform adds product ownership and operational responsibilities. It should solve a validated gap that configuration or integration cannot address adequately.
What costs should be included beyond development?
Include hosting, identity, backups, monitoring, supplier usage, maintenance, support, security updates, migration and the internal time needed to operate the product.
Can an internal property tool become a commercial product?
Possibly, but it needs separate validation of buyers, user needs, data isolation, onboarding, support and the commercial model. Internal usefulness is a starting point, not proof of market demand.