How Long Can Humans Be In the Loop?
Welcome to the 58th edition of The LogTech Letter, a weekly look at the impact technology is having on the world of global and domestic logistics. Last week, I looked at the bilingual mandate of LogTech. This week, I’m assessing the whether the concept “Human in the loop” is a real thing or a buzzword in logistics.
As a reminder, this is the place to turn on Fridays for quick reflection on a dynamic, software category, or specific company that’s on my mind. You’ll also find a collection of links to stories, videos and podcasts from me, my colleagues at the Journal of Commerce, and other analysis I find interesting.
For those that don’t know me, I’m Eric Johnson, senior technology editor at the Journal of Commerce and JOC.com. I can be reached at eric.johnson@ihsmarkit.com or on Twitter at @LogTechEric.
A long time ago, in a newsletter far, far away, I wrote about how the logistics industry needed to get to grips with a reality: that automation was in fact going to fundamentally reshape, and potentially shrink, the workforce of people moving freight.
I think back often to this story from June, particularly this passage from Jonah McIntire: “Any company in this industry has to wrestle with this discussion. We actually have to make it hard to take control, so it’s sort of the opposite of the ‘call-to-action’ button on the website. We don’t want them to click anything. The hard part is that people will listen to automation if it pretends to listen to them, so that’s comfort food. But we’re not assisting decision making; this is the decision making.”
Which brings me to a concept I’ve heard quite a few companies reference in recent times: human-in-the-loop. Essentially, human-in-the-loop is a branch of machine learning where human interaction is incorporated into the machine learning model alongside data (I won’t go any further into the definition as to further expose my lack of expertise).
Human-in-the-loop is one of those ideas that makes the idea of automation more palatable. It gives people a role in a future where programs do the heavy lifting. It makes people feel in control of decision-making. It might give them the final say. It might make them feel as though the foreboding automated future will involve real human instincts instead of cold, Gattica-like robotic decision-making.
In a logistics context, human-in-the-loop brings a whole host of things into play. People’s distrust of programmatic decision-making, the fear of job loss, the belief that moving goods is first and foremost a set of physical moves in a dynamic physical world that can’t be entirely conducted automatically. It’s a nice economic bridge as well, keeping people employed but theoretically allowing humans to be more productive, and thus allowing the company to use software instead of adding headcount.
To be clear, I don’t think human-in-the-loop is a buzzword, at all. It’s the opposite, in fact, because it is happening as I type this (including to me while I type this). I am contributing in real-time to a cloud-based editing tool’s understanding of how humans write, and how they make mistakes. Every time the doc autosaves, it’s also saving data fragments of how I type, what I type, what words I know, and what I erase. The mountain of knowledge grows. The question for me to ponder is, when will the program be evolved enough to take me out of the loop? I could be humble and say that time will come before I retire. Or I could be prideful and say, engaging with readers still requires a degree of human creativity, of unpredictable wordplay, that can never be captured by a program no matter how much data it captures from me and all the billions of humans in the loop contributing to its understanding.
This is what the logistics industry needs to work out in the decade ahead. What processes, or even individual tasks, require a human, what can move to human-in-the-loop, and what can remove the human from the loop? My guess is that a lot more than we think can move to the last of those three options, but it requires humans with various levels of responsibility to basically let go. That’s the hard part, not least because companies still need humans to do most of the work in the logistics today and machine learning models benefit tremendously from having humans contribute. There’s a symbiotic relationship that feels transitory and ominous, and yet necessary because this all feels inevitable.
The most tangible example is a co-bot, a physical warehouse robot that assists a human employee in their daily tasks. At some level, the human knows it is helping to train the robot to replace him or her. At some level, the human knows that will happen anyway, and so why not take the work while it’s still there. These uncomfortable transitions are likely to become a more standard situation in the years to come. That’s because capital is flowing to platforms built on human-in-the-loop models, where the desired outcome is an automation of tasks that are expensive for humans to handle, and for which humans aren’t well-equipped to handle anyway.
In a sense, the scenario that McIntire described, where the band-aid is ripped off, might be more painful in the short term but more representative of what logistics will look like in the future.
So much is riding on this from an investment perspective. Growth investors know that the more a system does and the less a human does, the more valuable their investment. Companies that can make faster, more accurate decisions and reduce the human employment costs of making those decisions, have the potential to scale quickly and make them lots of money.
As we ponder the future of the logistics industry, humans will not go away, but nor will a lot of current jobs look like they do today. For the things that can be automated today, those jobs are already on the way out. The place to look for substantive changes in the future workforce is in the processes where human-in-the-loop is gaining traction, where the model is learning from the data the human inputs, and from the way the human inputs the data.
Here’s a roundup of pieces on JOC.com the past week from my colleagues and myself (note: there is a paywall):
Another week, another wild array of acquisitions and funding rounds in LogTech. In an especially notable week, I’ll lead first with Transfix’s decision to go public via a SPAC. The reason this takes priority over some other pretty consequential happenings is Transfix is the first company in the 2010s crop to take the public plunge, something investors said that startup cohort needs. The question is, how representative will Transfix be in that regard, as a domestic freight broker among many others (both “digital” and traditional). Does its success in a public market say anything about how a digital forwarder might do, much less a data provider or SaaS company?
Speaking of companies rumored to be going public, project44 made news again this week with the biggest LogTech acquisition I can think of, acquiring last-mile visibility provider Convey for $255 million. It adds to the picture project44 is creating of acquiring adjacent competitors to fill out its modal puzzle. The company has designs on more acquisitions, and a war chest to make them happen, so stay tuned.
In any normal week, month, or even quarter, Emerge’s $130 million series B would be the biggest news. This week it’s only the third bullet here! Read my colleague Bill Cassidy’s excellent breakdown of what Emerge is trying to build here. Spoiler alert: mini-bids come up a lot.
Altana AI has a pretty ambitious aim: to create a map of global supply networks and subnetworks that allows shippers to understand their vulnerabilities and to allow customs agencies to weigh shipment risks. On top of that, Altana believes it can help 3PLs incorporate risk into operational and pricing decisions that will induce shippers to reduce their vulnerabilities. As CEO Evan Smith put it to me, “We’re imagining a FICO score for world trade.” The company this week announced a $15 million funding round (again, in normal times, that’d be a pretty big deal this week).
Interesting move from CH Robinson to incorporate DAT’s RateView data into a pricing benchmark tool for its customers. Bill Cassidy describes the product and its aim here. DAT told me such an avenue is open to any freight broker that uses RateView. As an aside, interesting to note which media outlets didn’t cover this, and whether it’s because they have a competitive product.
And finally, a funding round this week for Portchain, a company I’ve covered a lot the past year as it gains global traction for its berth planning and optimization software. In a world overcome by crippling port congestion, this approach won’t solve all the ills of hampered supply chains, but it could help ports maximize existing resources and infrastructure.
And here are some recent discussions, reports, and analysis I found interesting:
If Brian Glick is sounding off on industry buzzwords, it’s probably worth your time to read.
Good listen here about our declining interest in trade.
On a session I moderated this past week, someone dropped into the chat this essay from mid-’20 about how to grow US manufacturing. Not saying I agree with this all, but worth a read, especially in relation to the podcast directly above.
Xeneta veteran Thomas Sorbo is starting a new venture, focused on automating ocean invoice audits. Seems a hot topic given the craziness in the market, especially around switches from contract to spot, ancillary fees, loading premiums, etc.
Some upcoming events I’ll be involved in:
Next episode of LogTech Live is coming up at 10 am ET Oct. 1. My guest is good friend Angela Czajkowski, head of supply chain at forwarder and customs broker Shapiro. Can’t wait for this, and there will the normal news of the week, buzzwords explained and dadjokes aplenty. Use this page to navigate the numerous ways to catch the show.
I’ll be moderating a panel at JOC’s Inland Distribution webcast at 2 pm Oct 14 with Ryan Schreiber of CarrierDirect and Bruce Chan of Stifel. The topic is What Does the Next Phase of Digital Brokerage Look Like? Given the events of this week, this should be a lively discussion! Register here for all three days of the free webcast series Oct. 12-14.
Disclaimer: This newsletter is in no way affiliated with The Journal of Commerce or IHS Markit, and any opinions are mine only.