Fleet management software for a Surat truck fleet: a transport ERP with AI and automation

Why is a contract truck fleet hard to run on generic software?

A fleet that runs government and PSU contracts earns money slowly and spends it fast. Diesel is bought at many pumps, trucks are financed through EMIs, every vehicle carries a dozen documents that expire on different dates, and invoices to government customers are paid on 45 to 60 day cycles. Accounting software sees invoices, a GPS tracker sees location, and a spreadsheet sees whatever someone typed in. None of them ties a litre of diesel to a trip, a trip to a contract and a contract to the profit of one truck.

• Diesel is the biggest controllable cost. It leaks through unregistered pumps, cash fills and mileage drift that nobody notices until the month is over.
• Documents can stop a truck. A lapsed permit, insurance or fitness certificate is a dispatch risk, so expiry dates have to be watched across the whole fleet.
• EMIs hide in the bank statement. A loan debit that is not mapped to the right truck quietly distorts that truck's profit.
• Government invoices age. Receivables outside the 45 to 60 day cycle need chasing, and each invoice needs the right e-invoice and e-way bill paperwork.

Who is this fleet management software built for?

This is transport ERP software for the kind of business that owns its trucks, finances them, buys diesel every day and bills customers who pay late. It was designed around a Surat fleet running government and PSU work, but the same shape fits most Indian transporters with somewhere between a handful and a few dozen vehicles, in Surat, across Gujarat or anywhere in India.

• Contract transporters running freight for government departments, PSUs or large corporates, where invoices are paid on long cycles and every bill needs e-invoice and e-way bill paperwork.
• Truck owners with financed vehicles, where EMIs for different trucks are debited from different banks on different days and nobody is quite sure which debit belongs to which truck.
• Fleet owners who run the business from a phone, relying on calls and WhatsApp to know which truck is where, who filled diesel and which permit expires next month.
• Mixed fleets of tippers, tractor-trailers and multi-axle trucks, where mileage, tyre wear and service intervals differ per vehicle and a single average hides the problems.
• Logistics companies in Surat and Gujarat that have outgrown Excel and want one truck fleet management software rather than a GPS app, an accounting package and a diary that never agree.

Yukti AI has built the same kind of one-dataset system for factories too: see the Laxmi Fastener ERP and the box factory manufacturing ERP case studies.

It is less suited to a business that only needs live GPS fleet tracking, or a two-truck operator who is happy with a notebook. The value comes from tying diesel, documents, service, loans and billing to each truck, and that matters most once the owner can no longer hold the whole fleet in his head.

Why a custom fleet ERP instead of off-the-shelf fleet software?

Packaged fleet tools are often strongest on GPS tracking, and general accounting tools on invoices. A contract transporter needs both, plus the Indian specifics in between: GST bills for diesel, e-way bills, FASTag tolls, EMI days, permits and government payment cycles. A custom build lets the data model match how the fleet actually runs, which is the principle behind every AI-powered business management system Yukti AI builds. A clickable prototype first means the owner reviews every screen before any integration is paid for.

Fleet owner dashboard, live fleet board and GPS tracking

The control screens are where the owner starts the day. They answer which trucks are running, which need attention and what the money looks like, without opening a single report.

Owner dashboard

A single view of the fleet's position, with an AI briefing at the top, money figures on a rolling 30-day window (labelled as such, because a calendar month is a poor headline on the 3rd) and the items that need a decision. Why it matters: the owner sees what changed and where to look first.

Live Fleet board and map

A board of every truck with its status (Available, Running, Idle, Workshop, Breakdown, Sold or Inactive), plus a map drawn from real coordinates with no map API key required. It is vehicle tracking software for fleet owners in prototype form, fed by simulated positions.

GPS alerts, route replay and driver behaviour

GPS alerts, a route replay and driver behaviour views sit under the GPS screen. One truck in the sample data has a faulty GPS device and falls back to driver-entered odometer readings, which shows how the system degrades when a device goes offline. An automation rule watches for offline devices every 30 minutes.

At trip start the system takes the truck, the contract and route, the odometer photo and reading, and the fuel gauge photo. At trip end it takes the end odometer, the gauge and the proof of delivery (POD). Kilometres, mileage, cost per km and trip profit are then calculated, not entered. A rule can hold dispatch if a required document has lapsed, and another raises an alert if a trip is running late. Delivered trips can feed a drafted invoice.

Vehicle maintenance software: job cards, tyres and batteries

Workflow D runs from a reported issue to a closed job card, with the cost landing on the right truck.

• Job cards. Issue, job card, estimate, approval, work, parts, bill, payment and close, each step recorded.
• Workshop board. Which trucks are in the workshop and where each job stands. In the sample data one truck is in the workshop on a gearbox job and past its service interval, and another is due today.
• Preventive schedule. Service intervals by kilometre and by days, such as an engine oil change every 10,000 km or 120 days, checked on every odometer update, so a service is proposed as a job card before it is overdue.
• Tyres. A wheel-layout view by axle position, tread tracking, tyre life, tyre cost and tyre cost per km, plus a replacement-due list below the tread limit.
• Batteries. Tracked alongside tyres as part of each truck's running cost.

Reports cover maintenance cost, service history, cost per km, breakdowns, workshop spend by vendor and parts consumption. The AI layer adds predictive maintenance notes based on the same records.

Expenses, approvals, EMI tracking for transporters and bank statement mapping

Expenses and bills with smart checks

Vouchers by category, truck and approval state. Smart checks flag patterns such as cash fills at an unregistered pump, roadside repairs and GST credit leakage, always as items to review.

Approval centre

One queue for fuel holds, bills, driver advances and other requests, with an AI recommendation in the drawer and a history of who did what. The sample data includes a driver advance request waiting for a decision.

EMI and loans, and bank statement mapping (workflow F)

Each truck's loan, EMI schedule, principal, interest and outstanding are tracked. A bank statement is imported, and narration, amount and EMI day are matched to propose a truck. Each suggestion carries a confidence score and a plain-language reason. The owner approves, an EMI voucher posts to that truck's ledger and the bank row is reconciled. Uncertain rows are left unmatched rather than guessed, because a wrong mapping silently distorts a truck's P&L. In the sample data a UPI debit with no matching pattern is deliberately left unmatched.

39 fleet reports and the truck profit and loss

Every report reads the same selectors as the screens and opens with a written summary. There are 39 in the report registry, in six groups: Fleet (5), Fuel (6), Maintenance (6), Tyres (4), Finance (10) and Compliance (8).

Truck P&L (workflow G)

Revenue less diesel, toll, maintenance, tyres, EMI, driver cost and compliance gives net profit, margin and cost per km for each truck. Because every cost row carries a truck, this needs no manual assembly.

How fleet management software can increase a transporter's revenue and profit

A transport business rarely loses money in one big event. It loses it in small, invisible ways: a truck that stands for four days because nobody offered it to a contract, a diesel bill that was never questioned, an invoice that slipped past ninety days, a permit that lapsed and kept a loaded truck at a checkpost. The prototype is built to make each of those visible. Below is how each lever works in the system. We are describing mechanisms, not promising a number, because this build has run only on sample data.

1. Know the profit of every truck

Because every cost row carries a truck, the Truck P&L subtracts diesel, toll, maintenance, tyres, EMI, driver cost and compliance from that truck's revenue and shows net profit, margin and cost per km. Trip P&L and contract P&L come from the same rows. That is often the first time an owner sees that one truck carries the fleet while another quietly loses money, and it is the basis for deciding which truck to sell, which route to reprice and which contract to renew.

2. Put idle trucks to work

The Live Fleet board shows every truck that is Available or Idle, and the Idle report lists trucks with no trip assigned along with the revenue being forgone. A paused automation rule, idle truck to contract matching, is built to propose a contract for an idle truck when it is switched on. A truck that earns on more days in a month spreads its EMI and insurance over more trips.

3. Collect government payments faster

Government and PSU customers commonly pay on 45 to 60 day cycles, which is exactly why invoices must go out quickly and be chased on time. The invoice draft rule proposes a draft as soon as a trip is delivered, the receivables ageing report shows every invoice past its cycle, and the overdue follow-up rule proposes a reminder on the due date and every 7 days after. The cash forecast in AI Insights then shows what the next 30 days look like.

4. Stop fuel leakage before month end

The fuel anomaly detection compares each truck's actual km/L with its own expected figure, looks at fill patterns, and reconciles litres expected from distance against litres billed. Smart checks flag cash fills at unregistered pumps. People often search for this as diesel theft detection software, but the system is deliberately careful: a flag reads Review Required or Watch, explains what was observed, and a person decides. It never labels anyone a thief and never auto-rejects an entry. What it does change is timing: a mileage drift gets looked at in the week it happens.

5. Avoid document-lapse fines and downtime

A lapsed insurance, permit, fitness certificate or PUC can mean a fine, a stopped truck or a lost load. The reminder ladder starts 60 days before expiry, the renewal list can go to an RTO agent on WhatsApp (mock today), and a rule proposes holding dispatch on a lapsed document before the truck leaves the yard.

6. Keep GST input credit from leaking

Diesel, parts and repairs bought without a proper GST bill, or bills that never reach the accountant, are input credit the business paid for and cannot claim. The diesel workflow asks for the GST bill on every fill, and the GST credit-leakage panel in smart checks surfaces spending that is missing it.

7. Fewer breakdowns, longer tyre life

Preventive service by kilometre and days, a tyre tread watch and tyre cost per km all aim at the same thing: spending on maintenance when it is planned and cheaper, rather than on a roadside repair when it is urgent and expensive.

AI for fleet management: it explains the numbers, it never invents them

The AI layer produces the daily owner briefing, the written summary at the head of every report, a per-truck insight panel, predictive maintenance, a 30-day cash forecast, approval recommendations, a renewal plan and driver coaching notes. It appears on the dashboard, on the AI Insights hub, on every report, on each Truck 360 profile, in the approval drawer and in the fuel review queue.

The constraint that makes it usable: the AI never originates a number. Every figure in a generated sentence comes from the same calculations the tables use; the narration only decides what is worth saying. Each block shows its confidence, names the records it came from and has a thumbs up or down.

Fleet AI assistant

An assistant answers questions by mapping them to the same selectors the reports use. If it cannot resolve a question, it says so and points to the screen instead of producing a plausible figure.

How the fleet ERP makes a transport business AI-enabled, exactly as built

A lot of software calls itself AI for fleet management. It is worth being precise about what this build does, because the precision is the point. The system computes every figure first, in one analytics layer. The AI layer then reads those computed facts and decides what is worth telling the owner, in plain language. It never creates a number of its own.

Deterministic narration today, an LLM as a swap-in adapter

In the prototype, every sentence is assembled by deterministic code from the computed facts. There is no language model generating text today. A hosted LLM is planned as a swap-in AiProvider adapter: it would receive the same computed facts as context and improve the wording, but the figures would stay exactly the same because they never come from the model. Each generated block shows its confidence, names the records it came from and has a thumbs up or down, so the owner can always check where a statement came from.

That design is what makes the AI safe to rely on in a business where a wrong number means a wrong payment. If you want to read more about the approach in general, see AI ERP and what an AI agent is.

Integration interfaces

Each external service sits behind a TypeScript interface with a mock adapter bound in one place: GPS, WhatsApp, accounting, e-invoice (IRP), e-way bill (NIC), bank statements, OCR, FASTag, notifications and AI. Screens, workflows and reports stay untouched when an adapter is replaced. Evidence is immutable: replacing a photo or document creates a new version and the audit log records who replaced what, so a driver's original reading stays recoverable.

A day for the owner, the driver and the accountant

These are illustrations of how the prototype is meant to be used, walked through on its sample data. They are not a record of the client's actual day.

The owner

He opens the dashboard with his morning tea. The AI briefing says which trucks are running, which are idle, which documents expire this week and what is waiting for approval. He opens the approval centre, reads the recommendation on a driver's advance request and approves it. A truck shows Review Required in the fuel queue, so he opens its Truck 360° and sees the mileage trend before calling the fleet manager. Before leaving the office he checks the receivables ageing and the 30-day cash forecast.

The driver

At the yard he opens the driver app on his phone, picks his truck and contract, photographs the odometer and the fuel gauge and starts the trip. At the pump he photographs the meter before and after, enters litres and rate and snaps the GST bill. If he forgets a photo, the evidence reminder prompts him instead of the office calling. At delivery he captures the POD and the end readings, and the trip closes itself.

The accountant

Diesel vouchers are already there, read from the bill and attached to the right truck, waiting only for a check. She imports the bank statement and reviews the EMI matches the system proposes, approving the confident ones and leaving the uncertain ones unmatched. Delivered trips have produced invoice drafts, and the overdue follow-up rule has a list of reminders ready. Vouchers flow to the accounting package on the hourly sync, so nothing is typed twice.