Most autonomy demos an operator sees are built around a new vehicle. The vendor designed the chassis and the stack together, so the demo shows both. What the demo cannot show is the maths that happens afterwards, when the operations director takes the idea back to a capital plan built around assets with twelve to twenty year lives, most of which are somewhere in the middle of that span.
That point of collision is where a large share of industrial automation programmes stall. Not on technology, and not on skepticism about autonomy. On the fact that the proposed path requires pulling forward a fleet replacement decision that the business had no intention of making this decade.
What 'just buy autonomous' assumes
The default path has three assumptions buried in it, and they are worth surfacing separately because each is defensible alone and implausible together.
The first is that replacement can be accelerated. Heavy industrial equipment is not a laptop refresh. It is financed, depreciated, maintained under negotiated contracts, and often still perfectly serviceable. Retiring it early means writing down assets that work to buy capability that is unproven in your specific environment.
The second is that today's vendor choice is a safe fifteen-year bet. Buying a purpose-built autonomous vehicle means buying that manufacturer's autonomy stack for the life of the asset. In a market this young, with this much consolidation ahead of it, that is a long time to be locked to one supplier's roadmap, one supplier's map format, and one supplier's view of what a safety case looks like.
The third is that the yard can absorb a parallel fleet. In practice, a small autonomous sub-fleet operating alongside a large conventional one creates two of everything: two maintenance regimes, two training programmes, two sets of operating procedures, and a set of interaction rules at the boundary that nobody has written yet. The pilot succeeds and the scaling plan turns out to be an organizational problem, not a technical one.
Three reasons retrofit wins on the merits
Retrofit is usually presented as the pragmatic compromise, or what you do if you cannot afford the real thing. That framing is wrong. On three dimensions it is simply the better strategy.
Capital rhythm. Retrofit stops automation from competing with the replacement cycle and lets it ride on top of it. The autonomy investment is decoupled from the chassis investment, which means it can be scoped, funded, and justified on its own timeline rather than waiting for a fleet decision that is years out.
Optionality. If the autonomy layer is separable from the vehicle, the operator's commitment to any one autonomy supplier becomes a renewal decision rather than a fifteen-year structural one. That is a materially different risk position, and it is the position a buyer should want to be in during the early phase of any technology market.
Operational continuity. The vehicles are the ones the team already knows. The maintenance contracts hold. The operators are the same people, doing a modified version of a job they already understand. Exactly one variable changes rather than five, which matters enormously when something inevitably behaves unexpectedly and somebody must work out why.
Retrofit is not the cheap version of autonomy. It is the version that does not require the operator to bet the capital plan on a single vendor's roadmap.
One autonomy stack, many chassis
The technical claim underneath the thesis is narrower than it sounds. Perception, localisation, awareness, navigation and control are, in their substance, chassis-agnostic. Understanding where you are in a GNSS-denied yard, what is around you, and what path is permitted does not fundamentally change because the vehicle underneath you is a different make.
What is not chassis-agnostic is the interface to the vehicle, and this is where retrofit projects genuinely get hard. Drive-by-wire integration differs by model and often by build year. Safety interlocks and emergency-stop behaviour have to be engineered per platform and validated per platform. Braking and steering response curves need characterizing individually. Warranty implications and liability allocation have to be negotiated with the OEM or accepted by the operator. Each new chassis type carries a real integration and validation cost.
The honest version of the retrofit claim is therefore: the intelligence transfers, the integration does not. Which is still a very good deal, because the intelligence is the part that took years to build and the integration is the part that gets cheaper each time you do it on a given platform.
What a retrofit-first roadmap looks like
The sequence matters more than any single step. Map and instrument the environment first, because every subsequent decision depends on a machine-usable representation of the site and none of it depends on which vehicles you own. Then bring one or two vehicles under supervised autonomy in a bounded area, which tests the integration path and the operational reality at the same time. Then extend across the fleet at the rate the integration economics and the operators' confidence allow.
The new purpose-built autonomous vehicles still arrive. They arrive later, in smaller and better-informed quantities, into a site that already has a semantic map, a live twin, a governance layer, and a workforce that has spent two years learning to supervise machines. That is a much better place to receive them than a greenfield yard and a hopeful purchase order.
The question to put to the board
Not 'should we buy autonomous vehicles'. That question forces a capital decision before the operational learning has happened, and it invites a vendor to answer it for you.
The better question is: what is the smallest change that makes the fleet we already own automatable? It is answerable this financial year, it produces learning rather than a bet, and it leaves every subsequent option open.
Do you want to work with us?
If any of this rings true for your yard, we'd like to hear about it. We're working with terminal and warehouse operators who want to test the retrofit path on their own fleet; starting small, on their own timeline and without a capital bet on a new chassis. If you're weighing a fleet decision, wrestling with a vendor lock-in question, or just want to compare notes on what "automatable" looks like for your specific mix of vehicles, get in touch. We're especially keen to talk to teams already running mixed autonomous/manual operations, or ports and terminals facing a Scope 1 reporting deadline that's forcing the automation conversation sooner than planned.
Reach out to the Kybera team at hello@kybera.com and we'll take it from there, together.
