The Case for Uploadable Form Data: A More Flexible Approach to Online Submissions

Anyone who has worked extensively with online submission systems will recognize a familiar frustration: you have done the hard work of gathering, drafting, and refining your content – often collaboratively, across multiple documents and tools – and now you face the tedious task of manually copying everything into a web form, field by field. The content is ready; the process of getting it into the system is not.

This challenge comes up repeatedly in our work with the Agri-food Data Canada (ADC) and the Climate-Smart Data Collaboration Centre (CS-DCC), and it is not unique to any one platform or domain. It reflects a structural gap in how most online forms are designed: they are built for data entry, not data transfer.

The Collaboration Problem

This tension is particularly well illustrated in the context of data management plans. As noted in a recent article from Upstream:

“Data management planning is often a collaborative process, involving researchers, librarians, and institutional support staff. External tools and shared documents make it easier to iterate on plans, incorporate guidance, and ensure alignment with institutional policies and available resources. When plans are created directly within submission systems, that collaborative process can become more difficult.”

The same dynamic applies across many submission workflows. Researchers, teams, and support staff work best when they can iterate freely in shared documents, draw on existing resources, and incorporate guidance from multiple sources. Forcing that process into a single submission interface introduces friction at exactly the wrong moment – when content is nearly complete and should be easy to finalize.

Why APIs Are Not Always the Answer

The conventional infrastructure response to interoperability problems is to connect systems via APIs. While APIs are powerful and appropriate in many contexts, they come with real constraints. Both parties must be ready and willing to build and maintain the connection. Integration work requires technical resources on both sides. Security and access management become more complex. And the result is a series of point-to-point connections rather than a flexible, open approach that any participant can use.

APIs are well suited for tightly coupled systems with dedicated integration teams. They are less well suited for the diverse, distributed ecosystems that characterize research infrastructure – where institutions, tools, and workflows vary enormously and where not every participant has the capacity to build custom integrations.

A Simpler Proposal: Publishable Schemas and Uploadable Data Files

At ADC, we have been exploring a different approach. The core idea is straightforward: if an online form publishes its expected data structure as a downloadable schema or sample data file, then users can prepare their submissions outside the system – collaboratively, using whatever tools work best for them – and upload a structured file (such as a .json file) when they are ready to submit.

The form receives the uploaded file, validates it against the expected format, and populates the interface with the pre-filled content. The user retains final control, reviewing and editing within the UI before submitting. This preserves the benefits of collaborative, tool-agnostic preparation while keeping the submission process human-centered and editable.

Consider the DMP Assistant, the Alliance’s data management planning tool. Currently, users draft their plans directly in the online interface. Under this model, a researcher could instead work with their existing lab documentation, institutional guidance, and an AI assistant to compile all the relevant information into a structured .json file, then upload it into the DMP Assistant to populate the form in a single step – arriving at the editing stage with a complete draft rather than a blank form.

We Already Do This

This is not a theoretical proposal. ADC already supports this kind of workflow in the Semantic Engine’s schema development tool. We provide a structured prompt that helps users draft their schema content with an AI assistant, then upload the resulting .json file directly into the Semantic Engine editor. The file is parsed, the fields are populated, and the user continues working from there – a draft version that may contain AI errors but offers the opportunity to correct and improve.

The pattern works. It reduces manual data entry, supports collaborative preparation, and lowers the barrier for users who want to work in familiar tools before engaging with a specialized system. We believe it is a model worth broader adoption, and one that submission systems of all kinds could implement without requiring significant infrastructure investment from either side.

Written by Carly Huitema