← Case studies

AI-native PIMS: Run 100 sites from the office, without chasing chat

Industry -
Construction & real estate development
Client -
Madhav Limaye Group (MLC), Pune
Focus area -
Multi-site project delivery across a growing portfolio
Status -
Live in production · native mobile apps in build
Multi-site project delivery

Summary

A Pune construction group running a large portfolio of concurrent sites on Excel and WhatsApp — leadership reported spending upwards of 8 hours a day chasing status in chat instead of reading it. Vendors went un-appointed because a prerequisite step slipped — and no one could see why. Cost overruns surfaced only at month-end. Adding a site meant another spreadsheet.

Mnerva built an AI-native PIMS: a single role-based system, live in six weeks from first requirements, with a construction-specific vendor tracker, per-activity cost tracking and an audit trail on every change — then extended it with AI catalog search, safe plan merging, per-person accountability and a client-ready reporting layer.

The firm can now see the whole portfolio from one screen — and the system surfaces the next required action before a milestone slips.

01

Live in six weeks

Requirements to go-live

From the client's real Excel files to a running system, first week of September.

02

2,000+ tasks

Under management

Vendor-tracker tasks across the portfolio — with built-in checkpoints that surface blockers early.

03

Proof-backed

Audited by design

Every change carries evidence and an audit trail — who, what, when, and what it was before.

The starting point

The project record lived in a dozen Excel files on a shared drive — project details, vendor tracker, cost tracker, meeting register, drawing logs — reconciled by hand and chased over WhatsApp by about 160 staff across Planning, Estimation, Contracts, Purchase, Design and Execution. Status was stale by noon, blockers were invisible until they became overdue, and there was no record of who said a task was on track. Tasks were marked "done" without proof, and the model simply could not scale as the portfolio grew.

Onboarding a project now auto-populates its task list, drawing register and schedule from the firm's own master templates — the "enter it once" principle that replaced the spreadsheet dozen. Every role, head office to site, has scoped access to exactly what it owns.

Every role lands on what matters to them

On login, each person sees a dashboard tuned to their job — no navigation, no chasing.

Leadership

Portfolio KPIs — sites on-track, delayed, at-risk or overrun — plus cost position and pending approvals across every project.

Project managers

Overdue tasks, work due this week, and items needing a client or partner decision — for their sites only.

Contracts & purchase

Vendors to appoint and prerequisite steps pending — surfaced before a contractor goes overdue.

Planning & design

Deliverables due, drawings pending, and the backend rows they own kept current for everyone downstream.

The blocker surfaces before the deadline does.

The flagship: vendor tracker

A construction cycle, tracked as it really runs

Every task is tracked by construction phase through its real lifecycle — from prerequisites ready to vendor appointed, due and completed. The firm's own process steps are built in as checkpoints: when work stalls, the system flags the exact pending step, to the right people — instead of a generic "overdue".

Vendor appointment is modelled as the cycle it really is — each step a tracked state with its own owner, not a free-text cell in a spreadsheet.

One backend, many views

Per-activity cost tracking

Estimated vs. work-order vs. work-done for each activity, rolling up to project and portfolio — over-budget items flagged, not discovered at month-end.

Plans import directly

Existing project plans import as they are; actuals track against the baseline so slippage shows up as it happens.

Drawings drive execution

A drawing register whose status feeds each task's readiness — work can't proceed on drawings that aren't ready.

Daily labour, three taps

Fast mobile headcount entry from site, rolling up to a portfolio view of labour deployment.

Meeting notes to closure

Every action point tracked to closed, with open points auto-carried forward and searchable across projects.

Who is holding up what

Open items grouped by responsible party — one screen answers where every blocker sits.

Proof-backed and audited by design

Consequential changes demand evidence before they save: marking a task complete needs a note, a site photo or certificate, and a confirmation; a delay needs a reason; a key process step needs its document. Routine edits stay friction-free.

Every create, update and delete writes an append-only audit entry — who, what, when, and what it was before — and uploaded documents are sealed and immutable. Cost figures are visible only to the roles that should see them; site roles never see the numbers. Continuity and accountability become controls, not conventions.

The client review, turned into product in days

Built at review speed

A recorded client review produced eighteen action items. Twelve of the thirteen product changes were built, tested and delivered within two days.

Plan updates merge safely

Importing a schedule into a live project produces a reviewed diff — clean updates apply, conflicts show file-versus-system side by side, and site-reported progress can never be erased by an import.

Critical path, visible

The critical path and its end-date impact are computed and visible — and activities that are incomplete or never updated stand out instead of hiding.

Accountability made personal

Where a team shares one login, each person picks their name and enters their own passcode — every change is recorded against the person, not the mailbox.

AI catalog library

Product catalogs are uploaded as PDFs and become attribute-searchable with no manual tagging — "products with green coloured lights" just works.

Client-ready minutes

Letterheaded meeting minutes with a self-building agenda, open points carried forward, and a highlighted "pending from the client's side" block with overdue counts.

A weekly rhythm

A Saturday-evening pending-task report per person, project-wise, grouped by age — plus a statutory documents register with validity dates and expiry badges.

The current phase

On site, on the phone

Two native mobile apps are in build, consuming the same backend — no parallel system to secure and audit.

For site teams: minutes captured in under two minutes standing on site, the daily labour count in under a minute with offline queueing, and one-tap task status updates with camera evidence.

For the client's side: a strictly read-only view of their own project — progress, schedule summary, deliverable status, the minutes and the points pending on their side. Access is provisioned only by MLC administrators.

What we deliberately did not do

We did not build a parallel mobile backend — one source of truth to secure and audit. We did not adopt side channels — meetings persist to the system, not to a shared sheet. We did not let AI extraction fail silently — a catalog without a usable result says so visibly and is retried. And we did not overwrite site-reported progress from any import, ever.

How it's different

vs. spreadsheets & chat

One source of truth with role-based visibility and proactive triggers — instead of stale sheets, lost photos and status buried in message threads.

vs. general-purpose PM suites

Construction-specific by design — the firm's own process steps act as gates, drawings gate execution, and vendor appointment is a real tracked cycle, not a generic task with three status columns.

vs. schedule tools alone

A plan is only the start: cost, vendors, drawings, labour, meetings and evidence all live against the same tasks, with an audit trail for disputes.

Why this matters

Growth in construction usually means more coordination overhead, not less. By turning the firm's operating reality into one governed, proof-backed model, the PIMS lets leadership scale the portfolio without scaling the chasing — and lets each role act on exactly what's in front of them.

Intelligence surfaces the next action; people make the call.

Scaling a multi-site
operation on spreadsheets?

See where a governed, AI-native backend could remove the chasing — starting with a Business View of how your organisation works today.