.png)
What happened in Pt.1? - TL;DR
In my first piece on 3PL warehousing, I laid out the foundational reality of an industry that handles $13.1-16.4 trillion worth of freight annually yet remains largely invisible.
- Scale - 3,000-3,500 providers managing 60-75% of all US freight through warehousing facilities, ranging from giants like XPO Logistics to single-facility operators serving niche markets
- Economics - brutally thin 3-7% operating margins with labor consuming 40-50% of revenue, creating a demanding environment where technology must prove value quickly or face no market pull
- Switching paradox - theoretically easy to change providers, practically sticky due to contracts and migration complexity, with average contract lengths of 1-3 years and hardcore termination clauses
- Trust business - communication quality often trumps operational perfection; reliability compounds into defensible relationships where each perfect order builds lasting client partnerships.
Now, where do I see the real opportunities? Here's where my taste comes in... (and we’re not talking about amatriciana or pizza al taglio!)
Always on the hunt
As expected, AI is increasingly taking space on the central stage also in 3PL warehousing. As a testament, while in 2023 only 16% of 3PLs were exploring AI, this figure surged to 25% in 2024 - making absolute sense if we consider that ~75% of shippers would be somewhat likely to switch 3PL based on their AI capabilities. However, it emerged that 3PLs are still catching-up, as only 19% of them are actively implementing AI tools, with this number possibly reaching 41% in the next years (thx to 2025 3PL Industry Report for the stats).
As you keep reading, my stance on how AI is interesting (or not) crystallises.
What is not my taste?
My taste in this market might run counter to several innovation areas that currently seem to spark quite the interest.
The core WMS/IMS replacement play faces nearly insurmountable barriers to adoption. These systems form the central nervous system of warehouse operations, with switching costs - both financial and operational - that few operators willingly undertake. The entrenched nature of these platforms means they're more likely to be extended than replaced.
Hardware-intensive solutions like autonomous guided vehicles (AGV) or humanoid robots demand hefty upfront capital while delivering murky ROI timelines that make CFOs squirm. The AGV arena is absolutely packed - legacy heavyweights like KION Group and Toyota aren't just playing defence, they're dominating with established supply chains and decades-deep customer relationships. My number-crunching shows a newcomer would need to pull off 70%+ throughput gains (way beyond the standard 40% bump) just to get warehouses to consider switching. And let's be real - these incumbents have war chests that make startup funding rounds look like lunch money, plus they've already got maintenance contracts locked up tighter than a drum.
Similarly, robotic humanoids require prohibitive hardware and software development to reach the dexterity levels needed for (all) warehouse operations, making them impractical for most applications despite their sci-fi appeal. While current warehouse tasks do match existing robotic capabilities, the fundamental issue isn't technical feasibility but economic logic. Simple tasks that robots can handle today (palletising, basic picking) don't justify humanoid form factors when specialised automation already exists at a fraction of the cost. Meanwhile, complex manipulation tasks that would actually benefit from humanoid dexterity (returns processing, exception handling) remain prohibitively expensive to solve reliably. You're essentially paying humanoid premiums for robot-level performance, or betting on dexterity breakthroughs that have consistently disappointed for decades. Although these solutions target the 50-60% direct labor cost reduction sweet spot, I still haven't seen one approaching this budget mindful of the competition or hardware limitations—they're trying to replace humans in applications where purpose-built automation has already won on economics.
Beyond hardware, the software landscape presents its own challenges. Communication platforms (usually targeting carriers and brokers rather than 3PLs) and workflow tools face diminishing returns as the market saturates. AI communications with agentic workflows represent a red ocean with similar offerings proliferating faster than differentiation can be maintained. I won’t rant on this point as I believe my thesis is more relevant for freight forwarders, brokers and carriers.
Documentation automation startups face similar headwinds - despite increasing complexity from tariff-driven supply chain changes, this space is becoming commoditised as larger players integrate similar capabilities into their existing offerings. Besides, their extremely policy-driven nature (not my taste) means their disruptive potential is short-lived. Criticality seems to appear first, it then gets smoothened away as operators naturally adjust to policy shits and any ordinary point-solution generates the same outcome.
What is my taste?
Information flow unification represents another compelling opportunity that addresses operational inefficiencies head-on. Data availability and transparency remain massive hurdles in this space, with information often siloed between shippers, carriers, and 3PLs. Solutions that create novel information layers connecting these stakeholders can reduce friction in handoffs and improve coordination throughout the supply chain. I'm particularly interested in approaches that retrieve data from original sources rather than relying on third-party inputs—building the information infra layer instead of adding another AI layer on top of second-hand data creates sustainable defensibility.
Here's where my taste for hardware joins the party: when companies build their own custom systems rather than layering software on top of existing hardware infrastructure. I'm looking for what I call the "bathtub, not the water" approach - in an age where AI is flooding everything and leaving water damage everywhere, I want to invest in the bathtub that holds the water, not the replaceable water itself (Directed by: Patric Hellermann, Interpreted by: me).
Hardware represents atoms that simply cannot be displaced by AI, making it the ultimate defensive moat. When a company controls the full stack - their own sensors, their own data collection, their own processing - they create a system of record that becomes nearly impossible to replicate. It's not just about being 10x better; it's about building something so fundamentally integrated that competitors can't even attempt to copy individual components. The magic happens when the hardware enables entirely new workflows and data collection methods that weren't previously possible, creating a compounding advantage where the physical infrastructure becomes the foundation for breakthrough software capabilities (this is not only applicable to 3PL warehousing or logistics at large btw).
In the following example I link back up a point in the previous section (“What is not my taste?”) with the largest 3PL cost item (direct labor cost) and the bathtub metaphor. It’s hardware heavy (with vital software component), low-dexterity, and addresses key bottlenecks such as unpredictable staffing, inconsistent hiring cycles, and lack of existing skilled-workers. Indeed, the humanoid robotics landscape isn't entirely without merit when we shift focus from dexterity-heavy applications to purpose-built, task-specific deployments. While complex manipulation remains economically unfeasible, low-dexterity humanoid platforms designed for simple, repetitive tasks are beginning to show (some) early sings of commercial viability. Agility has effectively become the only player with a somewhat meaningful warehouse floor presence precisely because they didn't chase the dexterity holy grail - instead focusing on basic mobility and simple object handling that makes economic sense for operators.
However, the current deployment reality provides some contexts around current market dynamic: the only operators I've seen actually using these robots are Amazon and GXO (in testing phases with Agility, Apptronik, and Reflex) - massive logistics operations with billions to spend on automation experiments. It got me thinking: there simply aren't many clients with similar pockets and needs down the long tail of smaller 3PLs - against what current are these players swimming? Death? Consolidation? - it’s a story for another time.
Still, these low-dexterity humanoid manufacturers could be interesting to follow since we're not looking at a 10-year horizon for larger-scale deployment (at least according to Adrian Stoch and the recent VC appetite). In my opinion, the competitive battlefield will center on software intelligence (language commands and inferability), floor adaptability, commercial viability, and optimal robotic form factors - is something that looks like you actually the best fit for all warehousing tasks? With no clear category creator having emerged yet and the competition focusing on controllable technological differentiators rather than moonshot dexterity breakthroughs, there seems to be space for (hedged) betting.
Shifting our focus from manufacturers to operators, I want to keep my taste in check and mention how vertical integration remains a model I consider of tremendous value also in warehousing. How would it look like? A company that reimagines the entirety of warehousing operations through technology integration and takes on the role of operator - a tech-enabled warehouse 3PL. Companies on this spectrum are attacking labor costs that grow linearly with traffic volume, high human turnover rates, labor shortages, and frequent mistakes at the floor level. Hereby, I’m not referring to robots manufacturers anymore, but fully autonomous warehouse managers using off-the-shelf AI-robotics, providing custom software to manage and integrate all systems and securing 24/7 operational uptime (with minimised human intervention). Finally, such companies would have a very limited uncertainty window to execute fast and gain customers trust - tough competition form the likes of GXO, Amazon, and Symbiotic.
On a different note, space optimisation systems with accurate demand forecasting deserve special attention as warehouse space becomes both more expensive and strategic. Tools that help project-based businesses scale space up/down on demand connect directly to the P&L concerns identified earlier, especially for hard industries with variable inventory requirements.
- Space optimisation directly hammers the lease/utilities line item that's quietly eating into margins. When warehouse real estate costs keep climbing and you're stuck with 60-70% utilisation rates due to clunky layouts, bloated inventory, or seasonal swings, you're essentially lighting money on fire at $8-12/sqrft annually. These systems squeeze more revenue out of every sqrft while cutting the fat from wasted space expenses - helping you avoid oversized lease commitments upfront and maximise the value of your existing fixed costs rather than bleeding money on underutilised square footage
Otherwise, I'm still on the lookout for companies directly addressing maintenance (nice chunk of cost base). Frankly, I've witnessed some eye-catching stuff flying around other logistics fields (e.g. trucking), yet none aimed at warehousing - will have to spend some more time on this slope (or maybe warehouse managers simply don't experience enough of a pain in these regards if compared to manufacturers/carriers for instance).
Finally, unlike internet businesses that succeeded through demand aggregation, physical logistics thrives on supply integration - resembling the vertically integrated models of the second industrial revolution rather than modern tech platforms. This approach recognises that physical problems require physical solutions even in our digital world; and what is more physical than real goods being stored, moved multiple times, and carried around a gridlocked monument of boxes, forklifts, and steel.
Conclusion
I've taken you all through a view of the 3PL warehousing landscape—from core activities and market demographics to the economics driving decisions and where I see the most compelling opportunities. The exciting opportunities might not be in flashy AGV replacements or dexterity-chasing humanoids that burn cash faster than they deliver value. Instead, I see some of them in systematically attacking the P&L pain points that actually keep warehouse operators awake: labor bottlenecks that compound with every hiring cycle, information silos that create expensive coordination failures, and space optimisation challenges that directly hammer the bottom line.
What draws me to this industry is its fundamental resistance to purely digital solutions and its demand for the "bathtub, not the water" approach (like many other frontier industries). The most compelling opportunities combine physical infrastructure with breakthrough software capabilities - whether it's low-dexterity humanoids that focus on economic viability over sci-fi appeal, or information systems that retrieve data from original sources rather than layering AI on second-hand inputs.
Success in this space demands founders who understand that moving atoms requires more than managing bits. I'm drawn to those who can bridge operational expertise with technological vision, and who recognise that in a gridlocked monument of boxes, forklifts, and steel, the most sustainable advantages come from controlling the physical infrastructure that enables everything else.