Instrument Workbench
An upstream system of record that turns the scattered work of building research instruments into a structured, searchable item bank, with provenance on every item and a recoverable record of every decision.
The challenge
Building a survey or assessment instrument is, in practice, a scattered process. Inputs arrive from everywhere: protocol documents, prior surveys, meeting notes, validated scales, email threads where collaborators weigh in. There is rarely one place where it all lives. Validated items get rebuilt instead of reused, pretesting is under-resourced, and the rationale behind each revision is usually gone by the time anyone needs it at manuscript review, IRB amendment, or reuse.
The solution
A system of record for the development process itself. It sits upstream of fielding tools, turning messy inputs into a structured item bank where every item carries provenance and every decision is tracked.
The deliverables
- Structured, searchable item bank
- Item provenance and versioned history
- Validated-item surfacing
- AI wording-issue flagging (human-reviewed)
- Structured pretesting loops
- Decision and revision tracking
- Platform-agnostic export (REDCap, Qualtrics)
Built for the work before fielding, not the fielding itself
Instrument Workbench is not a survey builder or a fielding platform. It handles the work that happens before an instrument is ready for REDCap or Qualtrics: gathering and structuring inputs, building and managing an item bank, surfacing candidate validated items, and tracking the decisions that shape the final instrument. It exports in formats those platforms can receive.
What it offers
What it gives research teams
- A searchable item bank instead of files scattered across inboxes
- Validated items you can find and reuse, with provenance attached
- AI that flags wording issues, with you making every call
- Pretesting loops that link participant responses to specific revisions
- Export in formats REDCap and Qualtrics can receive
What it preserves for later
- A documented account of how the instrument was built, not just the finished instrument
- Versioned item history with the reasoning behind keep, revise, drop, or split
- Revision rationale that holds up at manuscript review, IRB amendment, or reuse
- No silent edits: every AI suggestion is visible and reviewable
- No change to where you field surveys, only what arrives there
AI that supports judgment
AI assistance flags potential issues in item wording (ambiguity, double-barreled questions, inconsistent response formats) and suggests candidate items from relevant sources. Flagging is not deciding. Every suggestion is visible and goes through human review before anything changes. The system reduces friction on the analytical work; it does not remove the methodological judgment the work requires.
Where things stand
Instrument Workbench is at concept stage. The problem is well-defined and the solution concept is developed enough to test with people who work on it directly. The current focus is understanding real research workflows, validating the capabilities that would create the most value, and designing an early pilot.
Feedback from people who do this work is shaping the product direction now, before any significant build commitment is made. The trajectory is toward a focused pilot that validates the system-of-record concept against real research workflows. No general release timeline exists at this stage, and none is being offered.