Bots are everywhere these days. They play poker against you and help you order a pizza. They assist in getting you hotel reservations and chat with you when you contact customer service to find out why your pizza had extra onions instead of extra cheese. And now, thanks to Cork, Ireland-based Keelvar, they can all but take over a company’s ocean and air freight procurement.
There’s no question air cargo really needs help right now, given the volatility in the market. Capacity is down, way down—nearly 40 percent from China in mid-February, 20 percent over the last year as a whole. Ocean capacity has also dropped. Meanwhile, demand has been rising, due to the pandemic. Optimizing sourcing at a time when rates, transit times and carriers are changing so rapidly is challenging for even the largest firms.
For Keelvar, there are few better times to unleash their bots.
“We’ve been helping shippers to find ways to bring the product to market faster,” says Keelvar CEO Alan Holland. “It’s automated—that’s what’s different about what we’re doing. A bot can go to work as soon as someone wants to move something, say, from Montevideo to New York. It’s always available. That’s the biggest competitive advantage.”
The bots that Keelvar and many other companies make these days are simply software that automates specialized tasks. Keelvar calls the artificial intelligence (AI) bots it makes Sourcing Automation, which it defined in a June 2018 white paper as “a new category of software that leverages intelligent systems to automate complex human reasoning that exceeds expert standards.”
Holland likens his bots to those that entered the world of online poker a few years ago. The earliest poker bots could play the game well, but couldn’t best professional players. But as the software evolved, the AI learned how to play better. Today, even the best poker champions in the world can’t beat the latest generation of poker bots. Holland’s goal, which he articulated in the white paper, is creating software that performs sourcing work for companies better than the experts in the field.
“The sobering fact is that AI is defeating the best human experts in most tasks where the boundaries and constraints on decision making are well-defined closed systems,” states the Keelvar white paper. “It would be a mistake to assume that AI won’t be competitive in the task of strategic sourcing and then ultimately overtake humans in this role. Once the boundaries of decision making are communicated, then the game-theoretic reasoning for optimizing the mechanism for sourcing goods and services becomes just another complex but the tractable calculation for Artificial Intelligence.”
In other words, Keelvar says their bots can automate all bidder communications in sourcing: opening, feedback generation, data cleansing, closing, termination-criteria monitoring, and activation. They basically run a company’s sourcing events, though logistics officials are always free to override the bot’s preferred course of action. Sourcing events can be complex and require the labor of many employees, some with years or decades in the procurement field, but Keelvar’s sourcing automation can basically handle it all.
“This process is tedious to execute manually and the more bidders there are, the more onerous the tasks above become and also the more likely that short-cuts are taken, and mistakes are made,” states Keelvar’s white paper. “Furthermore, the slow pace in manual events leads to curtailment of the rounds of bidding and inevitable lost savings opportunities due to the frictional effect of manual operations.”
What’s more, Keelvar’s ocean freight bots can even account for the pollution emissions of cargo vessels when conducting sourcing events (a feature that will eventually be available on the company’s air cargo bots). “Humans can’t get to that level of detail to do emission-sustainable options,” Holland said.
While the ocean freight bot has been around a year or so, the air freight bot only became available in January. Keelvar says the bot can automate 90 percent of a company’s tactical sourcing processes.
“It’s a natural evolution from our first bot, the ocean freight bot,” Holland said. “There’s a finite set of airport codes, but different logic around recommendations in air freight. Bid sheets are different, and cargo tends to be weight-based, rather than container-based, which is how ocean freight works.”
For Felix Plapperer, a venture capital investor and CEO of Paua Ventures, Keelvar’s sourcing bots will “dominate” the procurement market. Not merely because the bots are inherently more efficient than manual labor, but also because they learn the sourcing job better every time they operate.
“When a tender/auction is conducted by a bot, the number of actions is between eight and 20, depending on the complexity,” Plapperer wrote in a June 4, 2020, post on Medium on why he invested in Keelvar. “If each bot action costs less than $1, then the cost per event is roughly one or two orders of magnitude (yes, that is 10x to 100x times) cheaper than for an event operated by manual labor. Now, these cost savings only capture the value driven by process automation. As ‘mini-tenders’ are not run by sourcing experts, little to no optimization takes place (in fact, often personal relationships drive the outcome). Sourcing bots, in contrast, analytically optimize each and every event based on business priorities. Thus, they create additional value in reduced spend—every time they are at work.”
Of course, the bots aren’t replacements for a company’s procurement teams, but were designed to work alongside them. Major companies such as BMW, Novartis, Siemens and Coca Cola are already using Keelvar’s bots for their procurement.
Another of Keelvar’s recent customers, McKesson Corp., is an Irving, Texas-based healthcare and pharmaceutical company founded back in 1883. According to Keelvar’s marketing materials, McKesson was on track to save 6 percent of its global freight budget prior to the pandemic, and had already saved 6 percent the year prior, through the use of sourcing optimization products from Keelvar.
“Excel is nice, but it’s not where we need to be,” said Tad Strong, McKesson’s Vice President of Global Operations during a March 2021 webinar hosted by Procurement Leaders. “And over the next few years, we’ve some pretty big plans on making that shift into a more automated, more robust system.”
So, what does the future hold? Holland wouldn’t comment on whether his company is developing a bot to handle ground freight logistics (though that would be a logical step for the company), but he did say the next generation of artificial intelligence bots are just on the horizon.
“This is all level four automation, but level five automation systems in the future will have more autonomy,” said Holland. “We want the bot to autonomously decide on new carriers. There’s a lot of strategy on negotiating rates. You learn by experience which strategies are best. In level five, the bot learns new strategies.”
Holland said the first of examples of his level five AI bots in ocean freight should appear in the fourth quarter of this year.