A Marketplace Where Every Purchase Decision Carries Its Evidence
The Integrated Procurement Marketplace is being built as the buying and paying system for infrastructure delivery. Its stated design covers a catalog of pre-approved suppliers, an online ordering system, invoicing and payment capabilities, and analytics and reporting that give insight into procurement performance. It is designed with an open API architecture so it can sit alongside the ERP, CRM and supply chain management systems a company already runs. Behind it sits Resource.AI, a self-improving knowledge system that connects documents, site conditions, resources and decisions into one operational picture, so a decision arrives with the records that justify it.
This is a proposition for capital partners. One marketplace for buying, paying and reporting on supply, and one knowledge layer under it that turns what actually happened into the next, better decision. The commercial development is disciplined on purpose: a practical entry point, measurable results, and expansion through evidence.
Construction Has An Execution Problem And A Memory Problem
Every project generates thousands of documents, decisions, warnings and corrective actions. Most of that experience is scattered across inboxes, spreadsheets and the heads of individual people, rarely connected to the live project, and rarely surviving handover to the next one. Project managers carry the burden of connecting the dots, and the next project starts largely from zero.
The scale is not in dispute. McKinsey’s Global Infrastructure Report 2026 puts the cumulative global infrastructure investment need at $106 trillion through 2040. Infrastructure Australia recorded that 43 per cent of road projects exceed their initial estimated costs, and a 90-day review of the federal Infrastructure Investment Program in September 2023 found that known cost overruns across all Infrastructure Investment Program projects equal 41 per cent of the total budget for that program. Large federally funded highway projects can take from nine to 19 years from planning through completion, on Federal Highway Administration estimates cited by Pew in 2026, and those durations are maximum estimates, so actual timelines may be shorter. The demand is long-dated, the overruns are real, and the decisions sit with people who are not being supported well enough.
01The Execution Gap
The missing piece is an intelligence layer that continuously connects what the project requires, what the documents actually say, what is happening on site, what has changed since the last decision, which activities and resources are affected, what is likely to happen next, and what action should be taken now.
02The Memory Gap
Lessons learned are written once, filed, and seldom retrieved. The same RFI delays, the same defect patterns and the same subcontractor warning signs are rediscovered project after project. The industry pays repeatedly for knowledge it already owns.
03Where The Gap Shows Up
A delayed RFI looks administrative until it blocks a critical activity. A subcontractor who repeatedly misses prerequisites goes unnoticed until a delivery window is lost.
One Marketplace From Supplier Catalog Through To Payment And Reporting
The Integrated Procurement Marketplace is intended to streamline the procurement process for businesses. Its stated design integrates with a company’s existing procurement systems, and is accessible to all stakeholders within the company, including procurement, finance and operations teams. What follows is the marketplace’s own stated design, set out plainly. None of it is a claim that the marketplace is live, in use or processing transactions.
A Catalog Of Pre-Approved Suppliers
The design starts from a catalog of pre-approved suppliers, so a buyer chooses from a set the company has already approved rather than building one from scratch for every package. The source names the category and nothing more: no supplier counts, categories, geographies or qualification standard are claimed here.
Evidence: a recommendation identifies the documents, observations, dates and assumptions behind it.
Ordering, Invoicing And Payment
Online ordering sits alongside invoicing and payment capabilities. The stated design stops there. No payment rails, credit terms, ERP posting or reconciliation behaviour is claimed, because none is in the source.
Evidence: procurement records are ingested as authorised evidence and linked to the planned work.
Analytics And Reporting On Procurement Performance
The marketplace is designed to include advanced analytics and reporting capabilities, providing data-driven insight into procurement performance and identifying areas for improvement. No dashboard metrics, cycle times or savings figures are attached to them.
Evidence: the design automates information gathering and routine workflows.
An Open API Architecture
The marketplace would be built with an open API architecture that enables easy integration with other systems, such as ERP, CRM and supply chain management systems. Integration is qualified, as the source qualifies it: it happens where technically feasible, and where authorised integrations exist.
Guardrail: integration is where technically feasible, and where authorised integrations exist.
One System, Three Teams
Procurement, finance and operations teams are the intended users, and the stated design also provides a platform for companies to collaborate with their suppliers, partners and customers, to drive innovation, reduce risks and increase speed to market. The access described is within the company; no public supplier registration is implied.
Evidence: procurement coordination identifies supply constraints affecting planned work.
And The Layer Underneath It
What the marketplace records is what Resource.AI is designed to read. A dedicated specialist agent links supply commitments and delivery status to planned work, and procurement records are ingested as authorised evidence. Resource.AI does not issue orders, process payments or publish procurement dashboards. It supplies the evidence the marketplace’s decisions rest on.
Boundary: Resource.AI does not issue orders, process payments or publish procurement dashboards.
Resource.AI Is The Knowledge Base And The Resource Layer
Resource.AI is a Self-Improving Knowledge System: a project intelligence core that connects documents, site conditions, resources and decisions into one operational picture, turns that picture into prioritised and evidence-backed action, and then learns from what actually happened. Its most valuable component is not the ability to read documents or chat with users. It is the ability to build, maintain and refine a connected understanding of the project and of the way projects like it behave. Each cycle begins with the procurement record itself and ends with what actually happened, so the next decision starts from a better record than the last one. Nothing in the cycle takes the judgement away from the people running the transaction.
Project Knowledge
Contracts and subcontractor agreements, architectural and engineering documentation, specifications and scope, schedules and work packages, RFIs and design decisions, inspection and test requirements, procurement commitments, quality, safety and compliance procedures, and meeting minutes and approved instructions. The objective is to make the right knowledge available at the moment a decision is needed, with traceability to the underlying evidence.
Live Operational Knowledge
Project documents describe intentions. Site information reveals reality. Resource.AI can incorporate authorised operational data such as site progress reports, mobile inspection records, QR-code and RFID tracking, photographs and approved video analytics, attendance records, material and equipment status, defects and corrective actions, and work order completion. The aim is to connect the planned project to the conditions actually encountered.
Execution Intelligence
This is where the layers combine. Instead of four administrative items, the project manager sees one prioritised execution risk with supporting evidence, responsible parties and a recommended response.
It Acts, Within Limits
Where authorised integrations exist, it can initiate approved workflow actions. Within configured permissions it prepares the escalation, creates the task, notifies the responsible person or updates an integrated workflow. Anything needing professional judgement or formal approval goes to the right person first. Integration itself is described as happening where technically feasible.
The Seven Stages That Run Behind Every Purchase
The cycle is not an archive step. The final stage feeds directly back into how the next issue is understood, ranked and recommended. Most construction software is worth the same late in a project as on the first day; Resource.AI is designed so that it is worth more.
What Each Team Can Actually See
The value of an evidence layer is what it puts in front of a decision-maker at the moment a decision is needed, with traceability to the underlying record. Resource.AI assembles that context around each operational issue, separates verified facts from assumptions, conflicting records and predictions, and says so when the evidence is insufficient.
The Buyer Sees The Evidence Behind A Choice
A recommendation identifies the documents, observations, dates and assumptions behind it, and the system separates verified facts from assumptions, conflicting records and predictions.
- Requirement and qualification evidence held with the catalog entry
- Conflicting records and gaps shown rather than hidden
- Users can review, approve, reject or override any recommendation
The Supplier Sees The Collaboration, And Its Commitments
The marketplace’s stated design provides a platform for companies to collaborate with their suppliers, partners and customers. Behind it, a dedicated specialist agent links supply commitments and delivery status to planned work, so a supplier’s commitment is read against the work it has to support.
- Scope and commitment attached to the planned work
- Delivery status linked to planned work
- Recurring subcontractor and procurement patterns held in a pattern library with their evidence
Finance Sees The Commercial And The Administrative Picture
Analytics and reporting give data-driven insight into procurement performance and identify areas for improvement, and the design automates information gathering and routine workflows. Administrative hours per reporting period is a proposed measure for tracking that, agreed against a baseline.
- Reporting drawn from the records the work happens in
- Administrative effort tracked as a proposed measure
- Benefits quantified with project-specific evidence wherever possible
Operations Sees Consequence, Not Just Status
Project-wide consequence analysis maps how one delayed decision cascades through procurement, installation, inspection and completion, so issues are ranked by their potential effect rather than by how long they have been open.
- Consequences traced across procurement, installation, inspection and completion
- Dependencies and readiness gaps surfaced ahead of delivery
- Preliminary warning distinguished from a reliable forecast where data is incomplete
One Engine And One Learning Framework, Reused Across Projects
The same execution engine and learning core can serve multiple stakeholders across the construction and development lifecycle. The strategic advantage is reuse: one execution engine and one learning framework supporting many applications, rather than a separate AI solution for every problem. The value is not a set of separate assistants. It is coordination: a procurement risk is visible to schedule planning, a design issue informs subcontractor coordination, and a quality failure shapes inspection planning and future risk assessments.
- Across a portfolio A subcontractor pattern that preceded delay on one package is flagged earlier on the next, with the supporting records attached.
- Across time A recovery sequence that worked on a previous project is offered as an option, with its assumptions and the differences that matter.
- Across people When a new project manager joins, they inherit the verified history of decisions and outcomes instead of starting from a folder of unsorted files.
- Across organisations A consistent framework for identifying execution risk across projects, respecting project boundaries, permissions and confidentiality. Learning is scoped by default to the project and then the customer organisation, and cross-project or cross-organisation learning happens only under explicit permissions and agreed data policies.
The Decision Stays With The Person Accountable For It
Every recommendation shows its evidence and its confidence, and every learning is validated, permissioned and auditable. The system separates verified facts from assumptions, conflicting records and predictions, and says so when the evidence is insufficient. Consequential decisions, including contractual notices, technical approvals, safety-critical matters and financial commitments, stay with accountable people. That is a design principle, not a feature we switch on later.
Consequential Decisions Stay With People
Users can review, approve, reject or override any recommendation, and can see the documents, observations, dates and assumptions behind it before deciding how to act. This holds across the marketplace’s ordering, invoicing and payment features as much as anywhere else: the system prepares, people approve.
Learning Is Permissioned And Auditable
Only verified outcomes and confirmed human feedback feed the learning loop, while unverified claims and conflicting records are held for review. Every learned pattern links back to the evidence that produced it, and any learning can be reviewed, corrected or retired.
Access Controls, Audit Trails And Data Segregation
Access controls, audit trails, data segregation and retention policies protect commercially sensitive information. Customers keep control of their project data, and any cross-project learning requires explicit permission.
Changes Are Validated Before They Count
Improvements to risk indicators, rules and playbooks are validated against real outcomes before they change recommendations. The result is a system that improves through disciplined feedback rather than unexamined drift, and a customer who can see exactly why a recommendation changed.
Outcomes Become The Next Recommendation
Learning is not an archive step. Every resolved issue, rejected recommendation, failed inspection and recovered delay becomes verified knowledge that sharpens the next one, and every learned pattern links back to the evidence that produced it, so it can be reviewed, corrected or retired.
Proposed Measures, Agreed Against A Baseline
Resource.AI should therefore be introduced through clearly defined execution problems, measured against an agreed baseline. Because it is designed to learn, it should be held to a learning scorecard as well as a delivery scorecard. The measures below are proposed for the first deployment, to be agreed before a pilot period is set. They describe how progress would be tracked, not outcomes already demonstrated.
| Proposed measure | How it is agreed |
|---|---|
| Material-related delays and readiness failures | Counted per reporting period, compared against the baseline agreed at the start |
| Missed commitments and corrective action closure | Tracked per reporting period against an agreed baseline |
| Time to decision and action closure | Tracked per reporting period against an agreed baseline |
| Recommendation acceptance rate | Reported on the learning scorecard from the first deployment |
Start With A Small Number Of High-Value Workflows, Then Expand On Evidence
Every engagement starts with a defined execution problem and an agreed baseline. The proposition is one execution engine and one learning core, entered through a small number of clearly defined execution problems, measured against a baseline, and expanded on evidence rather than on a plan. Capital would be applied to engineering and product work on Resource.AI, the integrations that let it read the records a company already keeps, the marketplace and catalog build, and the delivery team that runs the first measured deployments. Bain’s Global Private Equity Report 2026 names the shift in private capital plainly: “We call this ’12 is the new 5′.” The firms that win “turn that differentiation into a system, not a slogan, and one that is backed up by data.”
No funding amount, valuation, round size or financial outcome is claimed anywhere on this page. Every engagement starts with a defined execution problem and an agreed baseline, and it is measured against both delivery and learning scorecards. Investment, valuation, legal and funding decisions remain subject to professional assessment. [email protected]