Most of the AI conversation inside UK consumer products and logistics businesses has been about software: agents that forecast demand, agents that chase deductions, agents that draft the board pack. That conversation matters, and it is one we have covered in depth, from why demand forecasting AI fails in month four to how to integrate AI with your WMS, TMS and carrier systems.
But there is a second wave forming, and it does not live in a dashboard. It lives on the warehouse floor.
Gartner's 2026 supply chain technology trends report names two forces reshaping the sector: agentic AI, which most operators are now at least piloting, and physical AI, which is newer and far less understood (Gartner, June 2026). Physical AI is the label for AI models that perceive and act in the real world: computer vision on the pick line, autonomous mobile robots moving stock, and machine-learning-driven quality inspection on the production line, rather than AI that only reads and writes data.
For a UK mid-market CPG or logistics business, that distinction is not academic. It is the difference between a project your data team can scope and a project that touches capital expenditure, health and safety, and the physical layout of your site.
What physical AI actually means for CPG and logistics
Strip away the marketing and physical AI in this sector tends to show up in three forms.
The first is computer vision for quality and safety: cameras and models trained to spot damaged packaging, mislabelled cases, or a forklift operating in an unsafe zone, in real time rather than in a weekly audit. Industry coverage through 2026 has repeatedly flagged this as the fastest-payback entry point, because it slots onto existing lines without a full automation rebuild (Automation World).
The second is autonomous mobile robots and forklifts, which move goods around a distribution centre without a fixed rail or a human driver. This is the most capital-intensive form and the one most commonly reported with payback windows inside twelve months when deployed against a well-defined, repetitive task such as pallet moves or replenishment (Neural Wired, February 2026).
The third, and the one most relevant to a business already running agentic AI in planning or procurement, is the link between the two. Gartner's own framing treats agentic AI and physical AI as converging rather than separate: the software agent decides what needs to happen, and the physical system executes it, closing a loop that used to require a person in the middle (Supply Chain Digital).
If your business has already read our piece on agentic AI in procurement, the governance-first lesson there applies here with even higher stakes. An agent that makes a bad purchasing call is a cost problem. A physical system that makes a bad call on the floor is a safety problem.
Why 2026 to 2027 is the window that matters
Three things are converging at once.
Hardware and model costs have come down enough that mid-market payback periods are now being reported in the low-teens of months rather than multi-year horizons, which changes the investment case for businesses that could never have justified a full-scale automation programme (Market Scale, mid-2026).
Adoption is moving from a small set of large logistics operators to mainstream mid-market deployment, which analysts are now describing as a 2026 inflection point rather than a future trend (AI CERTs).
And the regulatory and trade-compliance backdrop is shifting alongside it. If your business already has one eye on EU AI Act exposure, as we covered in does the EU AI Act apply to your UK business, physical AI systems that make safety-relevant decisions are more likely to sit in a higher-risk category than a forecasting model, which means governance needs to be designed in from the start, not retrofitted the way we have seen happen with software agents.
Where the real payback is, and where it is not yet
Not every part of this is ready for a mid-market budget, and it is worth being direct about that rather than selling the hype.
Computer vision for defect detection and safety monitoring is the closest thing to a safe first move. It typically layers onto cameras and infrastructure you may already have, the model can be scoped to one line or one gate, and the failure mode if it underperforms is a missed alert rather than a physical incident.
Autonomous mobile robots for internal stock movement are proven technology with real UK deployments, but they need a site that can absorb the change: clear floor layout, defined routes, and a team that has already been through some form of AI change management. This is where the lessons from what AI change management actually requires transfer directly, even though that piece was written about software rollouts.
Fully autonomous, agent-directed physical operations, where a planning agent dynamically redirects robots without human sign-off, are still early for anyone outside the largest logistics networks. Most mid-market operators piloting this in 2026 are keeping a human approval step in the loop, which mirrors the staged rollout approach we recommend for agentic AI generally.
A first step that will not put you ahead of your governance
Before any hardware conversation, the same discipline applies that we use for every AI investment decision: understand where the business case is strongest before committing capital. That is the purpose of our AI FlightCheck diagnostic, which assesses data, process, and organisational readiness, including for the physical and operational questions that a software-only audit will miss, such as site layout, health and safety ownership, and existing automation debt.
For a business that has already run FlightCheck and is looking at a scoped physical AI pilot, the FlightPath Sprint is the natural next step, structured the same way we approach any high-stakes, low-margin-for-error deployment: narrow scope, clear ownership, and a defined measurement window before anything scales.
If you want a plain read on where your operation stands before physical AI becomes a board agenda item rather than a choice, take the AI Readiness Scorecard or book a strategy call.
Sources: Gartner, Top Supply Chain Technology Trends for 2026; Supply Chain Digital, Gartner: Agentic and Physical AI Top 2026 Supply Chain Trends; Automation World, Physical AI: New Possibilities for Mid-Market Manufacturers; Neural Wired, Physical AI 2026; Market Scale, Physical AI Converges on the Warehouse Floor; AI CERTs, Physical AI Mainstream Adoption Poised for 2026 Surge.