Most pipeline reviews begin with a slide and a memory. Ours begins with a query against a data mesh of infrastructure projects, tenders and the organisations behind them. That difference decides who is in the room when a scope is still being shaped, and who arrives at bid opening with four weeks to price a plant.
Why infrastructure answers have to rest on records
A project is a record long before it becomes an invitation to bid. It appears as a line in a capital improvement plan, a permit application, a funding award, a planning agenda item. Contractors who read those records at identification or preparation stage hold a lead measured in months, not weeks. Contractors who wait for the tender notice are pricing someone else's assumptions.
Within Taiyo Canvas coverage, the US project records we index distribute as follows by status. These are record counts, not project values, and they include permit-level and plan-level records, so read them as a distribution of where activity sits rather than a headcount of distinct schemes.
| Status | Records |
|---|---|
| Prospective | 10,105,531 |
| Under Construction | 9,774,488 |
| Closed | 7,533,737 |
| Failure | 85,969 |
| Unknown | 25,021 |
| Bidding | 20 |
| Active | 9 |
The proportion sitting at Prospective is the commercial point. It is the stage where a utility has a need and a budget line but no delivery method, no shortlist and often no named contractor. It is also the stage where a failure count of 85,969 records reminds a bid team that not every funded scheme reaches procurement. Qualification matters as much as discovery.
What ConstructChat does
ConstructChat is an AI assistant built on the Taiyo data mesh. It works the way a bid team works, in five moves.
It searches and filters project and tender records by sector, country, stage and record type, so a query narrows to a defensible set rather than a keyword soup. It breaks a market down by status, sponsor, lender and contractor, which answers the question a pursuit lead actually asks: who owns this pipeline, and who is already in it. It opens individual records, so a hit can be read field by field before anyone commits pursuit budget.
It looks up organisations, their associated projects and their contacts, described by role. It reads risk data and risk sources when cost or schedule exposure needs a number behind it, and it searches the web and news to place a record in current context, from funding announcements to award outcomes.
A worked example: five real projects from nineteen thousand
We asked a plain question: which wastewater treatment plant projects are prospective in California?
The filtered universe for prospective Water Supply and Storage projects in the state returned a count of 19,050 records. Adding the keyword "wastewater treatment" surfaced a page of 25 hits. Of those, five were genuine plant projects: Pala Wastewater Treatment Plant Improvements, Tule River Wastewater Treatment Plant Expansion, Southern California Wastewater Treatment Project, Solvang Wastewater Treatment Plant Upgrade and South Bay International Wastewater Treatment Plant Rehabilitation. The remainder were building permits and tangential news, which is the working reality of any infrastructure search and the reason a record must be opened, not counted.
We opened the Pala record. It returned a named owner: the Department of Health and Human Services, through the Indian Health Service. It returned status as prospective, a location in California, and a Clean Water funding category under the Bipartisan Infrastructure Law. It returned no delivery method, no key dates and no contacts.
For a pursuit lead, that absence is the finding. Federal funding is identified, the owner is named, and the procurement route is not yet fixed. That is a qualification conversation to start now, not a bid to price later. A second record from the same funding source shows the same pattern, which tells a contractor this is a programme to pursue rather than a one-off.
The organisation layer completes the picture. The Jacobs entity record returned 11 associated entity record categories, spanning government, consultants, banks, infrastructure investors and contractors, plus named business-development contact records held by role. The Great Lakes Water Authority record returned 8 associated records and 3 contact record groups. That is the difference between knowing a name and knowing the structure around it.
What the evidence says about timing
Two external data points frame the same conclusion. In the US water sector, our tender filters returned 847 records at bidding status, the late-stage pool that every contractor can see. And US construction input pricing continues to climb: the BLS producer price index for new industrial building construction stood at 200.935 in June 2026, against 193.846 a year earlier, a 3.66% rise computed from those two returns. Cost escalation rewards the contractor who priced the scope early and penalised the one who arrived at the deadline.
Competition at the award line is equally tight. On the San Francisco Southeast Water Pollution Control Plant programme, Jacobs and PCL won a best-value award by 4.4 points, 1,006.98 to 1,002.58, despite the losing team taking maximum price points. Technical differentiation decided it, and technical differentiation is built months before a tender drops, on records.
The next phase of infrastructure procurement will be shaped by owners who publish intent earlier and by contractors who can read it. Utilities are already signalling scope through plans, permits and funding awards long before an advertisement appears. The contractors who win will be the ones querying that record layer as a routine, qualifying out the failures, and spending pursuit budget only where the owner, the funding and the route are visible.
Follow us on LinkedIn for more updates.