Thousands of users needed to search across large volumes of documents, models, and metadata without requiring deep knowledge of ISO 19650 folder structures or document naming conventions, a barrier that slowed delivery and increased rework.
.png){kind=small}
As capital infrastructure programmes grow in scale and complexity, the volume of project information becomes a liability rather than an asset, unless teams can find, interrogate, and act on the right data at the right time. A major UK public-sector infrastructure client engaged VE3 to demonstrate how AI-augmented information management could reduce friction for designers, contractors, and operations teams across a multi-organisation supply chain. VE3 designed and deployed PromptX, our proprietary AI search and assistant platform embedded within the VE3-XIL layer as a practical, governed capability to transform how project teams interact with information held in the CDE.
Thousands of users needed to search across large volumes of documents, models, and metadata without requiring deep knowledge of ISO 19650 folder structures or document naming conventions, a barrier that slowed delivery and increased rework.
The client needed AI-assisted answers that cited their sources and linked back to authoritative CDE records not generative responses that could not be verified, traced, or relied upon in a regulated infrastructure environment.
Technical Submission Assessments were managed through unstructured spreadsheets with no KPI tracking, no formal workflow, and no traceable record of assessment decisions, creating governance risk on a programme where submission quality directly affected programme schedule.
Requirements and specification inputs from the supply chain were free-form, inconsistently structured, and difficult to validate, generating errors downstream that required rework at later design stages.
The client required strict assurance that AI capabilities would operate within their data residency, access control, and information security requirements, with no model training on client data without explicit approval.
VE3 deployed PromptX using Retrieval-Augmented Generation (RAG) over CDE metadata and approved content in the Azure Lakehouse. Every answer cites its source and links back to the authoritative document in Aconex, with per-record ACLs ensuring users only see content they are permitted to access.
A purpose-built Technical Submission Assessment application replaced spreadsheet-based processes with a governed digital workflow, providing KPI tracking, guided assessment inputs, and a fully traceable record of every submission decision aligned to the programme governance framework.
The Spec Builder tool guides supply-chain users through structured requirements capture, outputting validated PDFs and JSON records that are ingested directly into the CDE, reducing free-form errors and improving consistency across all contributing organisations.
PromptX assistants are accessible within Microsoft Teams and Slack, allowing project teams to query CDE content from the collaboration tools they already use, eliminating context-switching and increasing adoption across the supply chain.
All AI features operate under strict controls: no model training on client data without explicit approval, PII minimisation through tokenisation and redaction, feature flags enabling per-project on/off control, and full audit logs of all prompts, sources, and decisions.
.png)
PromptX delivers AI-assisted search with citation-first answers and full audit trails — giving infrastructure teams the speed of AI without compromising the governance that public infrastructure demands.
.png)
