All solutions
Book a demo

Proposed demo for Rodic · simulated data

Infrastructure / EPC · Use case 3

Predict the idle crew three weeks out. Fix it while it is still cheap.

Proposed Procurement, Execution and Project Manager agents correlate POs, vendor history, equipment logs, attendance and weather to predict bottlenecks per site, then rank actions for named approvers.

Book a demo Open the live demo

Rodic Consultants (proposed tenant) · proposed, not deployed

Constraint detail with root-cause chain and evidence (simulated data) Proposed demo for Rodic · simulated data
  1. RiskSite A week 2 turns red
  2. CauseLate steel + crane service clash
  3. ActionSplit-source 40 t, move service
  4. ApprovedSlip 5 days → 1 day

The problem

Bottlenecks are visible in the data weeks before they hit the site.

A supplier that has been late on its last four POs, a crane due for service in the same week as a deck pour, and a short crew: each is known to someone. Combined, they become an idle crew and a slipped pour.

  1. 01

    Signals in separate systems

    PO status, vendor scorecards, maintenance logs, attendance and weather rarely meet in one view.

  2. 02

    Idle time is expensive

    Crews and equipment waiting on a missing input cost money every hour and push the critical path.

  3. 03

    Mitigations need owners

    Re-sourcing, rescheduling a service or re-sequencing work each needs a different person to approve.

How it works

Predict, explain, rank, then route each action to its owner.

Proposed Procurement, Execution and Project Manager agents, governed by ThoughtShieldAI. Supplier data access is scoped per agent by policy.

  1. Constraint radar

    21-day lookahead per site across material, equipment, labour, permits and weather, with confidence.

  2. Root-cause chain

    Evidence cards from PO status, vendor scorecards, maintenance logs, telemetry, attendance and weather.

  3. Ranked actions

    Each action with modelled effect on idle hours, ₹ and days, its added cost, trade-off and a cited why.

  4. Named approvers

    Procurement actions go to the Head of Procurement; site actions to the Site Manager.

    Human step

  5. Utilization ledger

    Idle hours avoided, net ₹ and critical-path days, before and after approvals.

The human approval

Meera Iyer and Vikram Rathore · Head of Procurement and Site Manager (fictional)

Each approves only the actions in their remit. A fourth option, moving a crawler crane from another site, is shown as not recommended.

The scenario

A deck-slab pour that would have slipped five days.

Constraint CN-A-412 on the fictional NH-48 Pkg 3, Site A. Suppliers and people are fictional.

  1. 01

    The heatmap turns red

    Sites by weeks, scored for risk. Site A, week 2 turns red; each cell opens its drivers.

    Bottleneck heatmap of sites by weeks (simulated data) Proposed demo for Rodic · simulated data

  2. 02

    On the radar: a 5-day slip

    The MJB-07 deck-slab pour is predicted to slip 5 days at 0.82 confidence: 640 idle crew-hours and ₹18.7 lakh.

    Constraint radar with a 21-day lookahead per site (simulated data) Proposed demo for Rodic · simulated data

  3. 03

    Two causes in the same week

    PO 4500-22318 for 64 t of Fe 550D TMT is 6 days late, from a supplier late on its last four POs. The only tower crane is due its 500-hour service 12–14 Nov.

    Constraint detail with root-cause chain and evidence (simulated data) Proposed demo for Rodic · simulated data

  4. 04

    Three actions, three signatures

    Split-source 40 t from a second supplier (Meera Iyer, Head of Procurement), move the crane service by four days and re-sequence parapet spans 1–3 (Vikram Rathore, Site Manager).

    Head of Procurement approving a split-source action (simulated data) Proposed demo for Rodic · simulated data

  5. 05

    The ledger counts what was avoided

    Slip cut from 5 days to 1, 560 idle crew-hours avoided, ₹14.3 lakh net after a ₹2.1 lakh sourcing premium.

    Utilization ledger before and after approvals (simulated data) Proposed demo for Rodic · simulated data

Governance and lineage

From source data to outcome, every link kept.

The decision chain behind this scenario, as ThoughtShieldAI records it. Each step can be opened in the live demo.

  1. Source data

    PO 4500-22318 status, vendor scorecard, TC-01 maintenance log, attendance, weather

  2. AI recommendation

    Prediction CN-A-412: 5-day slip, 640 idle crew-hours, ₹18.7 lakh, confidence 0.82

  3. Agent collaboration

    Procurement, Execution and PM agents rank R-101 to R-104 with trade-offs

  4. Human approval

    Meera Iyer approves R-101; Vikram Rathore approves R-102 and R-103

  5. Execution outcome

    Slip 5 → 1 day; 560 crew-hours avoided; ₹14.3 lakh net

Atlas

Sources behind each prediction

Every evidence card names its system, owner and timestamp.

Nexus

Supplier-data permissions

Policy SUP-02 scopes which agent can read supplier pricing; tool permissions are shown on each action.

Pulse

Model cost

Cost of the prediction and ranking runs, by agent and model.

Security

Append-only audit

Prediction, ranking, approvals and outcome recorded in order.

Live demo

Play it yourself.

The full Supply & Site Constraint Predictor app runs right here, offline, with simulated data. Press Play demo for the guided scenario (about 74 seconds), or click around freely. Reset is in the top bar.

/demos/rodic-supply/radar Proposed demo for Rodic · simulated data Open in new tab (opens in a new tab)
Open the live demo

Designed for screens 1280 px and wider. On a phone it opens full screen.

Outcomes

What the scenario changed.

5 → 1
days of slip on the deck-slab pour
560
idle crew-hours avoided
₹14.3 lakh
net saving after the sourcing premium
0.82
confidence of the original prediction

SimulatedSimulated outcomes from a proposed demo scenario. Rodic has not deployed these agents.

Questions buyers ask first

How far ahead does it look?

The demo shows a 21-day lookahead per site, with confidence on each predicted constraint.

Can an agent place a purchase order?

No. Agents recommend. Each action shows the tool permission it would need and routes to a named approver.

What happens to actions that are not recommended?

They stay visible with the reason, so the approver can see what was considered and rejected.

Has Rodic deployed this?

No. This is a proposed demo with simulated data and fictional sites, suppliers and people.

See it on your own workflow.

A 30-minute walkthrough of the live demo, then a conversation about the agents, data and approvals in your environment.

Book a demo Open the live demo

Opens an email to hello@anvevoice.app · Proposed demo for Rodic · simulated data

More from ThoughtShieldAI