This A.I. startup is saving Walmart and other big companies millions by automating negotiations

April 27, 2021, 12:00 PM UTC
Updated April 27, 2021, 3:55 PM UTC

Never split the difference.

That’s one of the negotiating dictums promulgated by Chris Voss, a former FBI negotiator who has become one of the world’s best known experts on bargaining tactics. (Splitting the difference, for one thing, increases the risk of both parties being dissatisfied.)

Voss always says his negotiating strategies are rooted in psychology, meaning they’ll work as long as the other side in a bargaining session is human.

But it was studying under Voss and others on Harvard Business School’s famed negotiation master’s course that inspired Martin Rand, a former commercial director at Monsanto’s digital agriculture division, Climate Corporation, to try to take the human out of at least one side of the equation.

Rand went on to cofound Pactum, a company that uses artificial intelligence and a chatbot interface to automate contract negotiations for some of the world’s largest corporations, including Walmart and electrical equipment distributor Wesco International. Shipping giant Maersk Group is also trialing the software.

Today the company, which is based in Mountain View, Calif., announced it has raised another $11 million to fund its expansion, bringing to $15 million the total amount of venture capital that the two-year-old startup has raised to date. The new funding round was led by Atomico, the London-based venture capital firm run by Niklas Zennström, the Swedish billionaire cofounder of Skype. Rand had been a product manager at Skype earlier in his career. Atomico partner Ben Blume is joining Pactum’s board as part of the deal, Pactum said. Berlin-based venture capital firm Project A, which had previously invested in Pactum, also participated in the funding round.

In addition to Zennström’s Atomico, Pactum has attracted funding from other key members of the “Skype mafia,” veterans of the voice-over-Internet software pioneer, who have since become prominent tech investors. They are Europe’s closest equivalent to America’s “PayPal mafia,” whose members include Elon Musk, Peter Thiel, and Reid Hoffman.

Among the Skype mafia investing in the new funding round for Pactum are Jaan Tallinn, the former Skype engineer who has become a leading investor in many A.I. and video gaming companies; Ott Kaukver, the chief technology officer of; Taavet Hinrikus, the cofounder and chairman of fintech company TransferWise; and entrepreneur and investor Sten Tamkivi.

Skype’s original engineering team was based in Estonia, and Rand is a native of the country as are his Pactum cofounders, Kaspar Korjus, who developed Estonia’s e-Residency program, and his brother Kristjan Korjus, who is Pactum’s chief technology officer. Pactum has engineering and R&D offices in the country.

Pactum is designed to take on contract negotiations with the “long tail” of suppliers and vendors that most large corporations have. These are contracts that are not high-value enough to warrant much, if any, attention from a big company’s negotiation teams but which collectively add up to significant money. “The average Fortune 500 company has $240 million locked in inefficient deals in that long tail that people can’t renegotiate,” Rand tells Fortune.

He says that by using the company’s software, a large customer like Walmart has been able to gain 2.8% to 6.8% in profitability from each supplier deal Pactum negotiates, with one company seeing additional $1.5 million come in monthly on these contracts, Rand says. He says Pactum has just 28 employees but has seen its revenues grow more than 10 times over in the past year. He says the company plans to double its team within the next six months.

Pactum charges its customers $25,000 for a proof of concept. For full deployments, it licenses its software in one of two ways: either $120,000 annually for a three-year contract plus 25% of the first-year gains seen from using the system, with the annual license fee deducted from that cut, or $120,000 for an annual license and then a fee of $600 to $6,500 per successful negotiation.

When a corporation uses Pactum’s software, it works with human negotiation experts from Pactum to answer a series of questions that will set the bargaining targets for Pactum’s software: What terms does the company want for price, delivery, payment dates, and quality guarantees?What other terms and conditions does it ideally want to achieve? The system also needs to know what tradeoffs the company is willing to make among these variables as well as any red lines that the company cannot cross. Having established what the large corporation values and what scope the software will have to make concessions, Pactum’s software then uses a chatbot interface to conduct negotiations with the suppliers. It asks a series of questions that prompt the other side to reveal its preferences. It then seeks to find the solution that will maximize the overall value for the large corporation, while trying to give the vendor a satisfactory deal too.

In the back-and-forth, the chatbot even incorporates some of the psychology-based tactics Voss advocates, such as mirroring back to a person what they’ve just said so that they feel listened to. “The system learns how people negotiate,” Rand says. “It tries different strategies and tactics and gets better with each subsequent negotiation.”

The only downside of the system is that it isn’t a human, so there may be times when its natural language processing system won’t understand a particular nuance. And some contractors may be offended that the large corporation has left them to negotiate with a chatbot. For one thing, it means a vendor can’t hope to cut a better deal by taking the purchasing manager out for a fancy dinner or inviting them for a round of golf.

But given that many of these vendors or suppliers were too small to have garnered much meaningful attention from a human executive on the other side of the table in the first place, Rand says that in Pactum’s experience so far, some vendors say they actually feel more “heard” and had a more meaningful negotiation with the bot than they ever had with the company previously.

Jacob Gorm Larsen, the head of digital procurement at Maersk and author of the book A Practical Guide to E-Auctions for Procurement, says his company is currently planning to test Pactum’s software to see if conducting multiple simultaneous negotiations with suppliers could augment auctions as a way of contracting for things such as inland freight transportation or small capital projects. He says he thinks Pactum will be particularly useful in areas that are not particularly competitive, such as “spot” negotiations—one-off contracts for projects or services that would never be worth the time and effort of a human negotiating team.

One of the main advantages of Pactum’s approach to negotiations, Larsen says, is that by asking questions through the chatbot that force a customer to choose between alternatives, the system can build up a much better picture of both sides’ “value function,” the graph of tradeoffs they care about. Auctions can also be configured to reveal some of this information, and Maersk already uses sophisticated e-auctions to do this when it is putting out contracts for bid.

But, Larsen says, a simple “best price” auction won’t uncover much about other preferences such as delivery times and payment terms. As a result, you may get the lowest cost but not the best overall value, he says, and some parties might drop out of an auction conducted solely on price that could have actually provided better service on other contract terms. One of the goals of the pilot Maersk hopes to conduct with Pactum is to see when it might be more useful to use negotiations rather than auctions.

Clarification April 27, 2021: This article has been updated to clarify Maersk’s relationship with Pactum and how it plans to pilot the software.

Our mission to make business better is fueled by readers like you. To enjoy unlimited access to our journalism, subscribe today.

Read More

Artificial IntelligenceCryptocurrencyMetaverseCybersecurityTech Forward