Field operations intelligence

Software was built for desks. India runs on doorsteps.

Groundrealiti builds field intelligence for the people software forgot. We see the ground reality every tool keeps missing — then build the intelligence to fix it. One sector at a time.

Sector roadmap

One sector at a time, deliberately

We go deep before we go wide. The same field floor sits at a different doorstep in every sector that follows. Pick a sector to see how far down we go.

We started where the day actually starts

The field floor in four moves: the journey, where it breaks, the fix, and what it costs.

The field journey

Walk the lifecycle the way the field walks it

Each stage leaks. The annotation under every node is what is lost, rejected, or defrauded there — before the branch can see it.
01

Find the group, collect documents.

LeakTargets rushed; documents outsourced to ringleaders; bad IDs enter.

Sourcing
02

Compulsory Group Training.

LeakTreated as a checkbox; the group never really learns the terms.

CGT
03

Group Recognition Test.

LeakCoached, rehearsed answers; one member understands, the rest recite.

GRT
04

Loan is released to the group.

LeakAbsent guarantor; proxy member; pipeline case shared on day one.

Disbursement
05

Center meetings; demand list collected.

LeakData arrives after the window closes; EMI pocketing hides in 'partial'.

Collection / EMI
06

Overdue accounts worked by the team.

LeakChronic cases caught at 60+ DPD that should have surfaced at 0–30.

OD & Recovery
07

Non-performing; legal / Lok Adalat.

LeakRecovery cost has multiplied; the loss is now largely locked in.

NPA
Problem → Fix

Where it breaks, and how we fix it

Every problem the field actually has, paired with the feature that solves it. Flip a card to see the fix.

Sourcing · KYC

The form bounces a week later

A wrong Aadhaar digit, a name spelled two ways across IDs, or a borrower who already holds too many loans — found days later, after the trip is wasted and someone has to go back.

The fix

Rejection predictor

We check it before the agent submits and say exactly what would bounce. The agent fixes it at the doorstep. No wasted second trip.

CGT / GRT · Disbursement

A borrower who isn't real

Ghost clients who show up for training but don't exist, doctored photos, forged Aadhaar, and groups quietly stuffed with relatives — none of it visible to a back-office check.

The fix

Anti-fraud at the door

A live face-check against the document, edited-photo detection, and a flag when one group shares too many surnames or addresses. On the phone, offline, the moment it happens.

CGT / GRT

A weak group is six months of pain

Rehearsed GRT answers, one woman who actually understands the EMI, members who live too far apart — passed through because the meeting was treated as a checkbox.

The fix

Group-readiness check

A quick readiness test before approval — a short comprehension quiz, distance and kinship checks — that says this group will hold, or fix this first. Better to lose a week than six months.

Underwriting

Income, guessed

Agents estimate a family's income on instinct. The regulator has openly said this guessing is what drives bad loans.

The fix

Smarter income check

The same structured questions every time — work, stability, existing EMIs — producing a consistent income figure and a clear can-afford or overstretched call you can show an auditor.

OD & Recovery

The playbook lives in one head

Chasing overdue money is the hardest part of the job, most agents are new, and the veteran's negotiation instinct walks out the day they quit.

The fix

Recovery copilot

The veteran's playbook at the right moment: cap the promise at three days, bring in the family, take a little daily from the shop instead of waiting for a lump sum. A day-one hire collects like a ten-year hand.

Collection / EMI

A day behind, and unprovable

Collection data lands after the window has closed, EMI pocketing hides inside the noise of partial payments, and no one can prove who was where.

The fix

Verified presence and activity trail

GPS punch in and out confirms the officer was at the center; every visit, form, and capture is geo-tagged and timestamped into an auditable trail, with no manual reporting.

A two-second error becomes a multi-week loss

One name mismatch at sourcing doesn't stay small. Follow it.

The 2-second error “Sunita Devi” on the form. “Sunita D.” on the ID.
  1. +0 sec

    The mismatch is written

    An agent who can't fully read the English form records the name slightly off.

  2. +2 days

    Rejected before sanction

    The loan bounces at validation — never the borrower's fault, but it reads like it.

  3. same day

    The borrower is humiliated

    She showed up, trusted the group, and is told she failed a test she never saw.

  4. same day

    The group's effort is wasted

    CGT, GRT, the center meeting — the whole group's time collapses on one field.

  5. +1 day

    A re-visit is scheduled

    The agent must travel back out to the village to redo what should have worked.

  6. +24–48h

    Disbursal is delayed

    The cycle slips a day or two while the corrected case moves back through the queue.

  7. weeks

    The agent is demoralized

    Enough of these and a 60–90 day ramp never finishes. The agent quits, and the knowledge leaves.

One unreadable name mismatch at sourcing, fanning out into a multi-week, multi-party loss.
What it costs

Where a mid-tier MFI quietly bleeds

  • Rejection re-work

    Re-visits and re-keying for loans that bounced on a sourcing error.

    ₹ {{ rejection_rework_annual }}
  • Ramp time lost

    Agents paid through a 60–90 day ramp that attrition cuts short.

    ₹ {{ ramp_time_annual }}
  • Attrition re-hire

    Re-hiring and re-training against 75–80% annual churn.

    ₹ {{ attrition_rehire_annual }}
  • NPA from bad field data

    Losses traceable to integrity and capture failures on the ground.

    ₹ {{ npa_field_data_annual }}
Annual bleed ₹ {{ total_annual_bleed }}
Where a mid-tier MFI quietly bleeds every year. Figures in {{ }} await the owner's ROI model — shown as placeholders, never invented.

Depth in the problem, restraint in the cure. The fix is one product, built to drop into the app your agents already use.

The product

The Field Ops Intelligence SDK

One field-intelligence layer that drops into the app your agents already use — built on four things we refuse to compromise.

Voice, in their language

The officer speaks the form aloud — Hindi, English, Tamil, Telugu and more — and gets back a clean, validated field. No typing, no English, nothing lost in translation.

On-device AI, works offline

A full assistant runs on the phone the agent already carries. It answers from your own rulebook at the doorstep — with no signal bars and no server to wait on.

Data never leaves the device

Capture and checks happen on the phone. What's collected goes only to you, and never to us. Personal data does not cross our line.

Location intelligence

Every visit, punch, and capture carries a verified place and time — presence you can trust and a field trail you can show an auditor.

The moat, as principle

In

A voice in Hindi, at the doorstep

device
Nothing crosses this line
Out

A clean, validated field

  • On-device
  • Zero-PII
  • Offline-first
  • Vernacular-native

Stated as principle, never as blueprint.

Built to drop into the existing Android field app your agents already use — not to replace it.

Request a pilot
About

Why field operations, why sector-by-sector, why now

Field operations is where the real economy meets software and loses. The work is rural, vernacular, and offline; the tools are urban, English, and online. We think the only way to close that gap is to go one sector deep at a time — learn its floor cold, earn the right to its data, then move.

On-device Indic AI is finally viable

What needed a server two years ago now runs at the doorstep on the phone the agent already carries.

DPDP makes cloud field-apps risky

Pushing raw field data to the cloud is now a liability. Keeping it on the device is both safer and the only way that works offline.

Scrutiny of field data quality is rising

Regulators are looking harder at how field data is captured and verified. Quality at the source is becoming non-negotiable.

Start a conversation

Request a pilot

Tell us what your field floor struggles with most. We read the ground reality first, then talk about what to build.

We're DPDP-minded and we don't sell data. No phone-spam capture — just one conversation.