AI and software for project and PMO teams, across the schedule, the cost, the approvals and the documents that decide both.
Every line on the right comes from a record somebody already files. The report on the left is compiled by hand, by the person being reported on.
In brief
EigenSpark builds AI systems and the software around them for project and PMO teams. We read the records a project already produces: daily progress reports, drawings and revisions, RFIs, measurement sheets, variation notices, invoices and minutes. We reconcile them against the schedule and the cost model, so a slip shows up as soon as the evidence does. We keep the variation, risk and action registers current from those same records. Then we hand it to your team to run.
Your systems
Nothing here asks you to move the schedule out of the tool that holds it.
Use cases
The record first, then the schedule and the cost, then the documents, then what carries forward.
Daily progress reports, measurement sheets, photographs and logbooks read into the project record on the day they are made, with anything doubtful queued.
Document Intelligence Engine →Activities flagged as at risk on predecessor completion, approval age and resource actuals, with the specific reason named for each one.
AI & Data Strategy →Every drawing, RFI and submission tracked by age and by who holds it, mapped to the activities that cannot start until it comes back.
Agentic AI →Instructions, site records and correspondence read into a live register, with the entitlement and the evidence for each variation held together.
Contract Risk Analyser →Invoices checked against measured quantities, rates and the contract, with only the differences passed to the quantity surveyor.
Document Intelligence Engine →A forecast built from committed spend, unpriced variations and actual productivity, with the assumptions visible so the commercial team can argue with it.
AI & Data Strategy →The register updated from what the records show, so a risk that has not moved in five months is raised for a decision.
Agentic AI →Meeting recordings and minutes turned into actions with an owner and a date, then chased against the record until they close.
Generative AI →Drawings and specifications matched by revision across the register, the site copy and the contractor’s set, with anyone working to a superseded sheet flagged.
Document Intelligence Engine →Planned against actual deployment by trade and by shift, so a shortfall shows up in the week it starts, well before the monthly review.
Data Engineering & Platforms →Delay, rework, quality observations and invoice disputes scored per subcontractor across projects, so the next award is made on the record.
Data Engineering & Platforms →Every past project’s drawings, claims and correspondence made searchable, so a design change or a delay argument starts from what happened last time.
Generative AI →How an engagement runs
Reconcile the record first, automate one register, then hand it over.
Two to three weeks connecting the schedule, the cost model, the document register and the progress records into one view per project, with every activity tied to the evidence behind it. Everything else on this page depends on this.
One register taken end to end: variations and claims, the approval and RFI log, or progress intake from site. Small enough to prove on one project, and measured against your own last one.
Activity mappings, cost codes, approval routes, risk thresholds and report formats live in a console your PMO runs. A new project, contract form or contractor does not need us.
Training
Ninety-minute hands-on sessions, run on your own drawings, progress records and contracts.
Project managers and engineers
Reading, summarising and drafting against your own daily reports, correspondence and drawings, including the parts that decide a claim.
Walk away withA drafted notice and a weekly report, both built from your own records in the session.
Planners and quantity surveyors
Querying schedule and cost data directly, reconciling measurement against the contract, and automating the reports rebuilt every week.
Walk away withOne weekly report automated, and a measurement check that runs against the contract.
Project directors and the PMO
How to tell a prediction from a reported figure, what evidence a claim needs, and where a model belongs in a review that carries commercial consequences.
Walk away withA standard for what a status pack must show, and where AI output is allowed in one.
Project teams usually take AI & Machine Learning and Data & Analytics, with Enterprise Systems for the planning tools. The full catalogue is in Training & Enablement.
The engineering half
Four layers of ordinary engineering. The first one decides whether the schedule and the cost can be compared at all.
A slip is invisible until the schedule activity, the cost code and the drawing refer to the same piece of work. It is the least interesting layer and it decides the outcome.
Primavera or Microsoft Project stays the schedule of record and your ERP stays the ledger. Most of the effort is everything that arrives from site and from consultants.
Built for the person who signs the report, with the record behind each figure one click away, so a flag can be checked before the review starts.
Defined once and used everywhere, so the site report, the PMO dashboard and the board pack cannot disagree. Your team runs and owns it.
We also run the training that goes with it: SQL and data engineering, cloud, enterprise systems and cybersecurity, alongside the AI tracks. A PMO that can query its own project data stops depending on the person who built the spreadsheet.
What we offer project teams
Most teams start with progress intake or the variation register, and use three of the four.
Where a project programme starts.
Where a single workflow starts.
AI tracks, and the foundations under them.
Where the specifics change.
The specifics change by sector. See how this lands in Infrastructure & Public Works or Real Estate & REITs.
Security and controls
The project manager signs the report, every flag opens onto its record, and nothing changes the schedule on its own.
Flags, forecasts and register entries go to the person accountable for the project, with the record visible and the override kept.
A schedule risk, a cost variance or a claim entry links back to the progress record, the drawing or the invoice it came from, on the date it arrived.
No baseline is revised, no certificate is issued and no payment is released by a system. Your own approval chain does that.
Activity mappings, cost codes, approval routes and risk thresholds are managed by your PMO in a console. A new contract form does not need a release.
FAQs
No. The schedule stays where it is, and so does your ERP. The work is reading the records that should be driving the schedule, reconciling them against it, and giving the PMO one place to see where the two disagree. Updates go back through the tool’s own interfaces.
Yes, and that is the usual starting point. The template is fine as a format; the problem is that it is the only input. We add the records underneath it, which are the daily reports, the measurement sheets, the approval log and the photographs, and reconcile the template against them. Where they disagree, that is the finding.
It can tell you which activities are at risk today and name the reason: an unfinished predecessor, an approval that has been out for six weeks, productivity below the rate the plan assumes. That is more useful than a probability, and it is checkable. We will not promise a date, because a project record that is two weeks stale cannot support one.
It matters little for the schedule and cost work. For variations and claims it matters a great deal, because the entitlement and the notice requirements come from the contract. We hold those as configuration per contract form, and your commercial team owns them.
Both, for different reasons. Site gets the intake work, which is time back on paperwork the team already fills in twice. The PMO gets a report it did not have to compile and can defend in a review. The site benefit is what makes the PMO benefit last.
No, and the parts that are not are usually what makes the AI work. Mapping the work breakdown structure to cost codes, cleaning the vendor master, connecting the document register and building the rule engine are ordinary engineering, and they are most of the effort. The same is true of the training: alongside the AI tracks we run SQL, data engineering, cloud, enterprise systems and cybersecurity.
In your own cloud tenancy. Project cost, claims and correspondence are commercially sensitive and often contested, so they stay inside your boundary. Where a hosted model is used for text work such as drafting minutes, what is sent to it is agreed and recorded before anything is built.
One, for the intake and the register work, because the effort it saves is on that project. The cross-project work, which is subcontractor performance and search across finished projects, needs a portfolio, which in practice means five or more completed projects with their records intact.
Related
The products this ships as, the services under them, and the industries where the specifics change.
Progress records, drawings and invoices read into the project record.
→ ProductVariations, entitlement and obligations per contract.
→ FunctionThe award decision the performance record should inform.
→ IndustryWhere the claim and the contract form take over.
→ IndustryWhere the same work meets approvals and handover.
→ PillarThe engineering side, for systems a PMO runs.
→Send the reports, the current schedule and the document register for the same project. We call you within 48 hours and go through where the report and the records disagree, which activities are at risk today, and what the first automation would be.
Talk to us