JTC bids goodbye to the paper-heavy world of construction tenders
The Evaluation Virtual Assistant marks a technological breakthrough in a traditionally conservative sector
Add BT as a preferred source
DeeperDive is a beta AI feature. Refer to full articles for the facts.
[SINGAPORE] Public-sector procurement is not typically where one usually finds technological innovation. The processes are deliberate, the governance requirements demanding, and the consequences of error in a high-stakes construction award can be potentially costly for years downstream.
But JTC Corporation’s generative artificial intelligence (AI) tool, Evaluation Virtual Assistant (EVA), is changing that. The tool is designed to modernise one of the most document-intensive processes in government contracting.
Each construction tender that JTC processes involves roughly eight evaluators, approximately 14 man-days of work per person, and submissions that can run to some 30 documents for each tenderer.
With around 40 construction tenders evaluated annually, the cumulative toll is estimated at approximately 34,000 man-hours a year.
EVA addresses this by deploying large language models to perform the extraction and initial scoring work that currently falls to evaluators. Tender documents are uploaded into the system, which reads and interprets the unstructured submissions by identifying technical, contractual, and performance-related details, then mapping them against JTC’s evaluation criteria.
The system then generates a scored output, with its reasoning made explicit, allowing evaluators to review the logic rather than reconstruct it from scratch.
Navigate Asia in
a new global order
Get the insights delivered to your inbox.
In one documented example, EVA correctly flagged that a tenderer’s stated performance ratings were not substantiated by the supporting documents provided. That is precisely the kind of inconsistency that manual review under time pressure risks missing.
Importantly, the tool is designed around a human-in-the-loop architecture. Evaluators retain full authority over scores and final recommendations; the AI structures, surfaces and highlights, but does not decide.
This is a deliberate governance choice, rooted in the accountability requirements of public-sector procurement, and one that also shaped the interface design.
SEE ALSO
The step-by-step workflow also mirrors the natural cognitive sequence of an evaluator: establishing context first, confirming who is being assessed, validating that the evaluation criteria have been correctly interpreted, and only then proceeding to the submissions themselves.
The design process drew on direct observation of EVA committee members navigating their existing workflows, and the solution was co-created with project managers, contract managers and procurement officers rather than developed in isolation by a technology team.
The projected impact extends beyond efficiency gains.
JTC estimates that faster evaluations could shorten overall project schedules by two to three months, enabling the earlier delivery of industrial facilities that businesses are waiting to occupy.
The methodology is also designed to be transferable. The same architecture could, with limited reconfiguration, be applied to consultancy and IT tenders, and potentially extended to qualitative evaluation criteria in future iterations.
Decoding Asia newsletter: your guide to navigating Asia in a new global order. Sign up here to get Decoding Asia newsletter. Delivered to your inbox. Free.
Copyright SPH Media. All rights reserved.
