Kodiak Hits 93% Autonomy Readiness, On Track for Driverless Long-Haul in 2026

The Analysis

Kodiak AI's Autonomy Readiness Measure (ARM) reached 93% at the end of August, up from 84% in February, after more than 40,000 paid driverless operating hours. The company says a driverless long-haul commercial launch is still on track before year-end. For anyone moving freight in North America, that combination turns a technology story into a scheduling question: capacity that can be contracted, priced and, if it fails, exited.

The policy backdrop is moving at the same speed. USDOT published its National Strategy for Automated Vehicles covering FY2026 to FY2030, and California's DMV lifted its ban on driverless vehicles over 10,000 lbs, letting Aurora and Kodiak test on state highways. Gatik extended its PepsiCo middle-mile programme across North America. Read together, the signals point to one pattern: autonomy will arrive on fixed, repeatable lanes first, not across an entire network.

This advice covers what driverless freight can and cannot do today, which lanes are worth piloting, how to write a pilot contract that assigns liability, insurance and exception handling, how to set measurable acceptance criteria, and the explicit trigger conditions for moving autonomous capacity into your formal network plan.

For Exporters

• Map U.S. inbound lanes from port of entry to distribution centre and flag fixed, high-frequency, weather-stable routes as pilot candidates by 15 November 2026.
• Ask your U.S. 3PL or carrier in writing by 30 November 2026 whether they operate or resell autonomous capacity, and demand a quote splitting driver cost from autonomous cost.
• Keep a human-driven backup lane contracted at 100% of baseline volume at all times.
• Log the operator's published ARM each quarter and treat the 84% to 93% path as your review cadence.
• Settle liability, insurance sub-limits and cargo claims in the pilot contract before loading any freight.
• Test the pilot against a 10% inland saving on one 40-load-per-month lane, roughly USD 52,800 a year, before extending it.

As a Chinese exporter shipping into the United States, your landed cost is decided less by the ocean leg than by what happens after the container leaves the pier. Kodiak AI reported its Autonomy Readiness Measure at 93% at the end of August, up from 84% in February, with more than 40,000 paid driverless operating hours behind it and a driverless long-haul launch still scheduled before year-end. That tells you the U.S. inland leg, port-to-DC drayage and DC-to-DC middle-mile, is where autonomy will first push rates down. It does nothing for your factory loading dock in China, your ocean booking, or Asian-side customs. Treat it as a 2027 to 2028 lever on the U.S. distribution leg, not a 2026 fix for a late shipment.

Turn the maturity signal into a number you can monitor. Take Kodiak's 40,000-plus paid driverless hours at an assumed blended average of 55 mph across highway and city running; that equals roughly 2.2 million driverless miles already logged on public roads. That exposure base, rather than a demonstration video, is what makes a pilot conversation defensible today. Now take the readiness curve: the ARM moved 9 points, from 84% to 93%, between February and August, about 26 weeks, which works out at roughly 1.5 points per month or 0.35 points per week on a linear assumption. If that curve flattens for two consecutive quarters, assume the launch has slipped and keep autonomy out of any customer-facing transit promise.

Assume a lane moves 40 containerloads a month at USD 1,100 inland per box. That is USD 44,000 of inland spend a month; a 10% autonomous saving is USD 4,400 a month, or about USD 52,800 a year on that lane alone. A number that size justifies a pilot. It does not justify repricing a contract on a press release. So do three things this quarter. Map every U.S. inbound lane from port of entry to distribution centre and flag the fixed, high-frequency, weather-stable ones, because only those are worth discussing, by 15 November 2026. By 30 November 2026, ask your U.S. 3PL or carrier in writing whether they operate or resell autonomous capacity on those lanes, and require a quote that splits driver cost from autonomous cost. Keep a human-driven backup lane contracted at 100% of baseline volume at all times, and give your logistics manager ownership of the pilot file with a one-page monthly report covering cost per mile, on-time performance and exception count.

The fallback if autonomy stalls is straightforward: hold the human-driven lane, renegotiate the inland rate on volume, and shift cargo to rail intermodal where distance and DC layout allow. The trade-off is transit time, since intermodal typically adds one to two days on a 1,500-mile lane, and cut-off discipline is tighter. The traps are contractual and commercial. Do not promise a customer faster U.S. delivery on the strength of a 93% readiness score, because autonomy today covers fixed daytime highway running, not peak surges, not residential final-mile, not winter routes through the snow belt. Before a single pallet is loaded, pin down liability, insurance deductibles and cargo claims in the pilot contract. The most expensive clause is a vague force-majeure line that quietly pushes incident risk back to you.

For Cross-Border E-commerce

• Tag your three highest-volume, most stable middle-mile lanes as pilot candidates by 15 November 2026.
• Negotiate a per-mile autonomy pilot with the saving split defined up front, target 60% to you, starting no later than the first quarter of 2027.
• Require cargo insurance naming your company as additional insured with a stated sub-limit.
• Give one operations lead ownership and a monthly report to finance on cost per mile, on-time performance and claim rate.
• Benchmark the pilot on your three lanes at 60,000 boxes a year and USD 0.85 per box, a USD 51,000 annual spend, before extending it.
• Do not advertise faster delivery: keep the customer delivery promise unchanged until the pilot has earned a change.

Cross-border import ecommerce is decided by landed unit cost and the delivery promise. Gatik's decision to extend its PepsiCo middle-mile programme across North America shows where autonomy lands first: repetitive, scheduled runs between fulfilment nodes, which is exactly the DC-to-DC and DC-to-regional-hub leg that sits inside the cost of every SKU. Kodiak's ARM at 93%, up from 84% in February and backed by more than 40,000 paid driverless hours, confirms the technology is ready for paid operation, but only on fixed lanes in good conditions. It will not rescue a stockout and it will not flex for a flash sale, so a pilot must not be allowed to reshape your peak-season plan.

Anchor planning on the readiness trend rather than the headline. The ARM gained 9 points, from 84% to 93%, over the six months from February to August, roughly 1.5 points a month on a linear assumption, and uneven in practice. Set an internal gate and date your go or no-go review from it: if your own lane-level readiness tracks the same curve, a 90% gate written down today would be reached in about four months of monitored operation. Assume your three busiest middle-mile lanes carry 60,000 boxes a year at USD 0.85 per box for the DC-to-DC leg. That is USD 51,000 a year, and a 12% autonomous saving on those lanes is about USD 6,100 a year, small enough that it has to be proven rather than assumed.

Move this quarter. Tag your three highest-volume, most stable middle-mile lanes as pilot candidates by 15 November 2026. Ask the carrier for a per-mile autonomy pilot with the saving split defined up front, for example 60% to you and 40% to the carrier, starting no later than the first quarter of 2027. Require cargo insurance naming your company as additional insured with a stated sub-limit. Give one operations lead ownership of the pilot and a monthly report to finance covering cost per mile, on-time performance and claim rate, so the business case cannot drift quietly, and carry the pilot cost in a rolling twelve-month forecast instead of booking it as a one-off saving.

The classic trap is margin arithmetic that ignores what autonomy cannot do. It does not cover residential final-mile, so your customer-facing delivery promise stays exactly where it is and you do not advertise faster delivery. It does not surge, so peak volume stays on human capacity. Watch for exclusivity clauses that lock you out of switching back to drivers when an autonomous lane stops for weather or maintenance, and watch for savings measured against an inflated baseline; insist that the comparison uses your actual trailing-twelve-month cost rather than a list rate.

For Manufacturing Plants

• List fixed five-day-a-week plant-to-DC and supplier-to-plant lanes as pilot candidates by 15 November 2026.
• Obtain a pilot quote by 15 December 2026 that prices driver and autonomous cost separately with a 95% on-time performance floor over 90 days.
• Name your logistics supervisor as trial owner with a monthly report to plant management on lane cost and on-time performance.
• Keep current safety-stock targets until the pilot has held its on-time floor for a full quarter.
• Put a 30-day termination right in the pilot contract.
• Benchmark one daily shuttle at 260 round trips a year and USD 900 per trip, a USD 234,000 spend, against an 8% target saving of about USD 18,720 a year.

If you run a U.S. plant or feed a U.S. assembly line, autonomy touches your inbound parts flow and your finished-goods outbound, but only on fixed shuttle lanes, not on irregular expedites. Kodiak's ARM reaching 93%, up from 84% in February, after more than 40,000 paid driverless hours tells you the technology is dependable enough for scheduled runs. It is not a substitute for the flexible spot capacity you need when a line goes down at two in the morning. Plan autonomy as a cost stabiliser on known lanes and keep your expedite budget intact until a pilot proves itself on your own freight.

Use the readiness curve to size your monitoring effort. The 9-point rise from 84% to 93% between February and August spans about 26 weeks, or roughly 0.35 points a week on a linear assumption, and real progress is lumpier than that. For a plant shuttle that runs five days a week, review autonomous readiness at least monthly and log it against the lane cost. The remaining 7 points to a full 100% would close in about 20 weeks at that pace, which is a reasonable way to date a capacity-planning review in mid-2027. Assume one daily plant-to-DC shuttle makes 260 round trips a year at USD 900 each, a USD 234,000 annual spend; an 8% autonomous saving is roughly USD 18,720 a year, and it only counts once on-time performance holds on your own schedule.

Act on the lane list rather than the launch date. By 15 November 2026, list every fixed, five-day-a-week lane, plant to DC or supplier to plant, that runs the same route; those are the pilot candidates. By 15 December 2026, obtain a pilot quote that prices driver and autonomous cost separately and carries an on-time performance floor of 95% measured over 90 days. Name your logistics supervisor as trial owner with a monthly report to plant management covering lane cost and on-time performance, and tie any expansion decision to a written threshold rather than to a vendor's launch calendar.

The mistake is letting autonomy change your safety-stock maths too early. Do not cut buffer inventory because of a press release; keep current stock targets until a pilot has held its on-time floor for a full quarter. Two other traps recur. A carrier quotes a low autonomous rate but excludes detention, liftgate and night operations, so the real cost jumps. A pilot carries no exit clause, so when the operator pivots you are stranded with freight and no alternative. Put a 30-day termination right in the pilot contract, and keep the true cost of the lane visible, accessorials included, in every monthly report.

For Brand Owners

• Request each 3PL's autonomy roadmap and a per-mile rate on your top three fixed lanes by 30 November 2026.
• Require cargo insurance naming your brand as additional insured, plus an incident-reporting feed into your control tower.
• Name one supply-chain manager as pilot owner and review cost per mile, on-time performance and claim rate monthly with sales and finance.
• Write lane-by-lane approval and a quarterly review gate into the 3PL agreement.
• Benchmark 500 truckloads a year at 400 miles and USD 1.05 per mile, a USD 210,000 spend, against a 9% target saving of about USD 18,900 a year.
• Do not market driverless delivery to customers while autonomy is confined to fixed middle-mile lanes.

As a brand selling into the United States, your promise to the end customer, the delivery date, the condition of the goods and the price, rides on the U.S. distribution leg that autonomy is starting to reshape. Kodiak's ARM at 93%, more than 40,000 paid driverless hours and Gatik's North America-wide PepsiCo middle-mile expansion all point the same way: steadier cost and capacity on fixed DC-to-DC lanes. None of it touches your storefront service promise yet, and none of it shortens residential final-mile, so manage expectations internally before briefing marketing and keep the customer promise exactly where it is.

Anchor planning on the maturity number rather than the headline. Take Kodiak's 40,000-plus paid driverless hours at an assumed blended 55 mph and you get about 2.2 million driverless miles on real roads. That is the exposure base behind a 93% readiness score and the reason a pilot conversation is now reasonable. Assume your brand moves 500 truckloads a year on fixed lanes averaging 400 miles at USD 1.05 per mile. That is USD 420 a load and USD 210,000 a year; a 9% autonomous saving is about USD 18,900 a year. Enough proof to open a pilot, not enough to drop the human-driven backup, which should stay contracted at 100% of baseline volume so a lane outage never reaches the customer.

Concrete steps. By 30 November 2026, ask each 3PL for its autonomy roadmap and a per-mile rate on your top three fixed lanes, with the saving split defined. Require cargo insurance naming your brand as additional insured, and a transparent incident-reporting feed into your control tower so exceptions are visible while they are still small. Put one supply-chain manager in charge of the pilot and review acceptance metrics, cost per mile, on-time performance and claim rate, monthly with sales and finance, so the trade-off stays explicit and any shift in liability is caught early. Write lane-by-lane approval and a quarterly review gate into the 3PL agreement.

Resist the marketing temptation to claim driverless delivery, because autonomy is nowhere near the doorstep and a false claim is both a regulatory and a reputational risk. Three traps deserve attention. A 3PL quietly routes your freight onto autonomous lanes without your sign-off, moving liability to you; savings are computed against an inflated baseline; and channel priority collides, with autonomous capacity handed to a retail partner while your own stores wait. Fix all three in the contract, not in a corridor conversation.

For Procurement Teams

• Open rate talks with two autonomous-capable carriers before 31 December 2026 and secure a quote splitting driver from autonomous cost, targeting 50% to 60% of the saving to you.
• Build indexation: an ARM crossing 95% with the year-end launch holding triggers an automatic rate step-down.
• Add the 9-point ARM gain, 84% to 93%, as a quarterly readiness checkpoint in the carrier scorecard.
• Benchmark the book at 1,200 truckloads a year, 450 miles at USD 1.60 per mile, a USD 864,000 spend, against a 30% cut in driver cost, about USD 103,680 a year.
• Ban cost-pass-through and exclusive-dealing clauses; require a 30-day exit and defined insurance sub-limits.
• Require an on-time performance floor, a cost-per-mile target and a claim-rate cap as contract acceptance metrics.

Procurement holds the contract, and autonomous trucking changes the leverage you carry into carrier negotiations. Kodiak's ARM rising from 84% in February to 93% at the end of August, on more than 40,000 paid driverless hours, is a maturity signal you can cite at the table: operators need volume commitments to fund a launch, which opens a window to lock favourable rates before the capacity is sold. California's DMV lifting its ban on driverless vehicles over 10,000 lbs means heavy testing is now legal on state highways, which widens your future options and your negotiating hand.

Turn the regulatory number into a scope check. The 10,000-lb threshold California removed is one-eighth of a standard 80,000-lb Class 8 tractor-trailer on an assumed gross vehicle weight of 80,000 lbs, so the lifted ban covers the entire heavy long-haul segment rather than light trucks alone. Expect autonomous bids on your longest lanes first, which are also the lanes where driver cost carries the largest share of the rate. Assume you buy 1,200 truckloads a year, averaging 450 miles at USD 1.60 per mile. That is 540,000 miles and USD 864,000 of spend. If driver cost is 40% of that, USD 345,600, a 30% cut in driver cost on autonomous lanes is about USD 103,680 a year, and that is the number worth negotiating towards.

Before 31 December 2026, open rate talks with two carriers that operate or resell autonomous capacity and secure a quote that separates driver cost from autonomous cost with a saving split, targeting 50% to 60% to you. Build indexation into the agreement: if the operator's published ARM crosses 95% and the year-end launch holds, a rate step-down triggers automatically, which protects you if adoption stalls. Add the 9-point ARM gain from 84% to 93% over six months as a quarterly readiness checkpoint in your carrier scorecard, name the procurement manager accountable for the file, and require monthly cost reporting so the saving is tracked rather than claimed.

Close out the clauses that quietly transfer risk. Do not accept cost-pass-through terms that let the carrier reclaim the labour saving through fuel or accessorial formulas. Avoid exclusive-dealing clauses that block a return to human drivers. Every pilot contract needs defined liability, insurance sub-limits and a 30-day exit right. The costliest omission is a contract with no acceptance metrics at all: no on-time floor, no cost-per-mile target, no claim-rate cap, which leaves you unable to prove the value or to walk away.

↑ Back to the key points