Delhivery says self-hosted AI cuts support, POD costs over 50%

Delhivery Ltd. says fine-tuned, self-hosted open-weights artificial intelligence models have reduced customer-support and proof-of-delivery (POD) validation costs by more than 50%. The company said output-quality metrics remained comparable with frontier models, enabling these workflows to scale during the Q2FY27 peak period, when it handled 325 million Express shipments.

How did Delhivery reduce customer-support and POD costs?

Delhivery reduced customer-support and POD-validation costs by more than 50% through fine-tuned, self-hosted open-weights models deployed in production. POD is the confirmation that a shipment has been delivered, and Delhivery said the models maintained output-quality metrics comparable with frontier models while supporting scale during the Q2FY27 peak period.

The company said its initial AI work focused on giving agents greater autonomy in software development and testing, using its proprietary knowledge and processes. Over the last 12 months, Delhivery has deployed several AI agents and agentic workflows into production, extending the programme from development work into operating processes.

The deployment occurred alongside higher operating volumes in Q2FY27. Delhivery carried 325 million Express shipments, up 32.1% year on year, and 545,000 metric tonnes of part truckload (PTL) freight, up 14.2% year on year. The company linked the lower-cost self-hosted models to its ability to scale high-volume support and validation work without a stated decline in quality metrics.

Where is Delhivery using AI agents in operations?

Delhivery is using AI agents across the shipment journey, including order manifestation, POD auditing, returns-quality inspection, delivery-associate and shipper coordination, escalation handling, and claims resolution. Its agents also issue alerts on festivities, local disruptions and weather events for estimated-time-of-arrival planning and resource allocation, extending their role beyond the Q2FY27 support and POD workflows.

The company has launched agents that analyse video feeds to monitor facilities, guide floor and dock operations, and provide problem-solving inputs to operating teams. Delhivery did not disclose a rupee saving for those tools, but said they form part of a wider operational-automation programme aimed at staffing volatility, productivity-adjusted labour costs and climate-related working conditions.

That programme also includes physical automation rather than software agents alone. Delhivery said it is deploying bin and pallet automated storage and retrieval systems and shuttle systems across fulfilment centres, and expects to deploy 100 self-developed autonomous pallet transportation vehicles by Q4FY27. The company estimates that autonomous mobile robot pilots could reduce in-facility movement costs by one-third, though those pilots remain under way with Indian and international partners.

How is Delhivery applying AI in freight procurement and sales?

Delhivery is using voice agents to automate fleet procurement and tracking, and is externalising that capability to clients through its OS1 stack. In Supply Chain Services, the company said agentic transportation-management systems are supporting transportation-mix optimisation and freight procurement, which it identified as mechanisms for further profitability growth rather than separately quantified savings.

In full truckload (FTL), Delhivery is moving client servicing, quotations and vendor calling to agentic workflows. FTL freight margins exceeded 5% in Q2FY27, which the company attributed primarily to higher-quality leads and a shift from seasonal fixed contracts towards spot business; it also said the agentic flows are generating operational efficiencies.

Delhivery's field sales teams use AI-based beat planning integrated with Delhivery Maps and visit analytics captured in its customer relationship management system. AI-powered pricing simulators produce customer-specific rate cards from enriched profile data, with the stated aims of faster quote turnaround and adherence to benchmark margins. These tools operated as Delhivery reported 32.1% year-on-year Express-volume growth in Q2FY27.

The sales tools are being introduced while Delhivery expands Delhivery Local. The intra-city platform was live in eight cities at the time of the Q2FY27 letter, with annualised revenue run-rate rising from about Rs 100 crore at the end of Q1FY27 to about Rs 200 crore. Management re-baselined its Q4FY27 annualised revenue run-rate target to about Rs 280 crore from its earlier target of about Rs 250 crore.

What does Delhivery's AI programme indicate for financial progress?

Delhivery said AI-led billing automation and faster claims and dispute resolution contributed to a reduction in net working capital to five days in Q2FY27. That compares with 11 days in March 2026 and 17 days in September 2025. The company also cited tighter credit controls, improved client selection and lower receivable days, and said seasonal or business-mix factors could reverse some of the improvement.

Profitability also improved sequentially in Q2FY27, although Delhivery did not attribute that movement solely to AI. Adjusted earnings before interest, tax, depreciation and amortisation (EBITDA) rose to Rs 123 crore from Rs 76 crore in Q1FY27, while the adjusted EBITDA margin increased to 4.0% from 2.6%. Reported EBITDA rose to Rs 205 crore from Rs 142 crore, and consolidated profit after tax increased to Rs 41 crore from Rs 32 crore.

Delhivery identified fuel-linked price increases, minimum-wage-related price revisions, higher network utilisation, and changes in customer and shipment mix as additional drivers of financial performance. Management expects Express service EBITDA margins to end FY27 at 16% to 18% and PTL margins to exceed 15% by FY27 exit, while saying pricing revisions and utilisation should support expansion through Q3FY27.

The company has also established a dedicated frontier AI team to run focused experiments with new tools and scale those that prove effective. Delhivery said frequent open-weight model launches, lower proprietary-model prices, infrastructure-cost optimisation and lower-cost decision models are reducing the cost of comparable intelligence, which could make more high-accuracy workflow automation viable.

Conclusion

Delhivery's disclosure shows AI being used across a broader set of logistics processes than customer support, spanning shipment validation, claims, facility monitoring, freight procurement, pricing and sales planning. The company reported these deployments during Q2FY27, when Express volume grew 32.1% year on year and adjusted EBITDA reached Rs 123 crore, but it identified several commercial and network factors behind the financial improvement.

The next milestones are management's planned deployment of 100 autonomous pallet transportation vehicles by Q4FY27, results from autonomous mobile robot and loading-and-transfer pilots, and the frontier AI team's experiments. Management's disclosed FY27 targets also include 16% to 18% Express service EBITDA margins, PTL margins above 15%, and about Rs 280 crore of Delhivery Local annualised revenue run-rate by Q4FY27.