U.S. last-mile costs hold at 12% as network control outperforms AI spend

Last-mile delivery costs in the United States rose by a median of 12% year on year in 2026, matching the prior year's figure almost exactly, according to a survey of 84 shippers, carriers and third-party logistics providers conducted by delivery management software vendor FarEye. The result, drawn from 76 cleaned responses, is being read by the company as evidence that cost inflation has become a structural feature of the market rather than a temporary disruption working through the system.

Nearly half of respondents (46%) said costs are growing faster than revenue, while 42% said costs and revenue are growing at roughly the same pace. Only 12% reported costs growing more slowly than revenue, suggesting that volume growth is no longer a reliable offset for the industry.

Technology adoption is rising; trust in automation is not

AI adoption among the survey cohort advanced sharply over the past twelve months. The share of operators at implementation stage or beyond rose from 46% in 2025 to 66% in 2026, with extensive operational adoption growing 3.6 times over the same period. Despite this, mean trust in AI for real-time operational decisions scored just 1.98 out of 4, a figure that remained flat across low, mid and high maturity tiers. Seven of the eleven respondents reporting extensive AI adoption rated their trust at 1 or below, the lowest reading in the sample.

The data suggests operators are deploying AI heavily in advisory roles, ETA prediction (39%), demand forecasting (36%) and customer support (31%), while remaining reluctant to hand live routing decisions to automated systems (21%). The blockers cited are not awareness or budget: lack of internal expertise (28%), distrust of automated decisions (21%) and poor data quality (18%) are the dominant barriers for those yet to implement.

Control, not technology, predicts reliability

The most striking finding concerns the relationship between perceived network control and on-time performance. Among the 41 respondents who answered both questions, the correlation between control and on-time delivery was 0.474 (p=0.003). The correlation between technology maturity and control was 0.073, statistically indistinguishable from zero. Investment level showed a slight negative relationship with control.

Operators rating their outsourced-network control at 4 or 5 out of 5 achieved mean on-time performance of 85.9%, against 74.5% for those rating control at 1 to 3. A more granular breakdown shows operators at the highest control tier delivering inside the promised window 95% of the time, versus 65.5% for those at the lowest tier, at identical median investment levels. High-control operators also reported a WISMO (where-is-my-order) contact rate of 6.2%, against 20.8% for low-control peers, and median year-on-year cost increases of 8.3% versus 14.5%.

The FarEye analysis attributes the control advantage to contractual and governance arrangements rather than software configuration. Control correlated with customer-facing capability (0.274) but not with routing capability (0.201), leading the authors to describe the pattern as control being "bought contractually, not installed."

Market context

The findings land at a moment when last-mile software vendors are competing hard on AI-led routing and dynamic optimisation narratives. FarEye itself sells delivery orchestration software and has a commercial interest in how operators interpret the relationship between technology investment and outcomes. Readers should weigh the survey's framing accordingly, while noting that the statistical relationships reported, correlation coefficients with p-values, are more granular than is typical in vendor-sponsored research.

The broader picture the survey describes is consistent with patterns visible across the logistics technology market: enterprises are scaling AI deployment faster than governance frameworks can absorb, creating a gap between capability and operational trust. As agentic AI systems begin to enter supply chain tooling from vendors including Google Cloud, Microsoft and a range of specialist logistics-AI startups, the question of override, audit trail and explainability is moving from a compliance concern to a procurement criterion.

The survey also found that predictability has decisively replaced speed as the primary consumer promise operators wish to deliver, with 55.7% naming predictable delivery time or successful first-attempt delivery as the metric that matters most, against 11.4% selecting fastest possible delivery. Operators prioritising speed reported the lowest on-time rates in the sample (76%) and double the median cost inflation (24% year on year).