Last month I attended Tradeshift’s Innovation Day for analysts in New York City. It was an eye-opener.
Tradeshift’s Fascinating (and Daring) Evolution
When it was founded six years ago, Tradeshift started off in the fairly narrow (and slightly crowded) niche of buyer-supplier networks for automating invoicing and payment — with some key differentiation (such as their OCR-based machine learning facilitating the ability to intake invoices in any format1). Tradeshift is one of the few with a true many-to-many Networked SaaS Platform2 architecture. They have been successful at growing their network, to 600,000 suppliers last year with expectations of reaching 1.2M suppliers by the end of this year. They launched in China in January and already have over 100,000 suppliers there. Getting the long tail of small suppliers is one of their focuses. There is value in getting that long tail on board; for example, small suppliers usually benefit the most from dynamic discounting.
Addressing Supplier’s Needs
Tradeshift has been more supplier-friendly than some other platforms. For example, they don’t charge suppliers for transacting on the platform, regardless of the transaction volumes. They told us they have an NPS score over 50 with suppliers. They have a G2 Crowd rating of 3.9 out of 5 (based on 158 reviews).3 Tradeshift is providing supplier functionality that provides value for the supplier, such as the Tradeshift Product Engine.
Tradeshift Buy
As mentioned, Tradeshift has been developing P2P functionality beyond invoicing and payment. Tradeshift Buy includes Tradeshift Product Engine, a central, network-wide, searchable product database, maintained by suppliers, accessible to buyers. It can be thought of as a sort of free PIM for suppliers, that allows them to manage all their product data in one place and make it available to the buyers on the network. This lessens the content management burden for both buyers and sellers, and makes it more likely the information will be up-to-date and complete.
We saw a quick demo from the supplier’s perspective, importing product data from a spreadsheet. The system flagged one entry that was missing its UNSPSC4 code. Without leaving the import screen, the user typed in ‘notebook’ and the system brought up the correct code to insert. Once all the products had been successfully imported from the spreadsheet, the supplier could select which ones they wanted to offer to all buyers and which they wanted distributed only to a subset of buyers.

The demo then switched to the buyer’s perspective. The buyer user could manage by item and/or by company and choose to accept all the items to go into their catalog or only a subset of those offered. Tradeshift Shop then provides a way for thebuyers’ employees to search for items they need. The demo looked similar to the familiar Amazon-like online shopping experience. Items were added to a shopping cart, and then at checkout a requisition was created for approval. When searching, items frequently bought by this user would rise to the top of the search results. Third party data can be used to aid searching, in addition to the data provided by the suppliers. Rules can be added to the requisition workflow. For example, over a certain threshold could require at least three different bids.
Tradeshift has added contingent labor procurement to Shop. Contingent labor procurement has its own set of unique and complex requirements for the searching, requesting/approving, tracking, and payment processes. Tradeshift introduced Buy a year ago. They said they now have five large customers, two of them as complete replacements for existing Ariba systems.
Evolution to a Coherent, Multi-App Platform
Tradeshift started off by building all of the functionality themselves. Then around 2013-2015, they engaged with a handful of strategic partners to integrate those companies’ applications. The early efforts at integrating third party applications required a fair amount of ‘handholding.’ Meanwhile, Tradeshift has been investing heavily in building out a multi-app platform that can help them more rapidly and deeply integrate other applications, in order to scale out their expanded vision and footprint more quickly. The platform that Tradeshift is building could be described as a sort of ‘Force.com for the buy side.’
We refer to this approach as a ‘Coherent, Multi-App Platform’ (see The Rise of the Agile Networked Platform). Building a coherent multi-app platform is more than just exposing a few APIs. It involves providing a very complete and granular set of micro-service APIs, all of the tools and resources that developers need (especially development support resources), UI components/tools and guidance (to allow multiple apps to appear as one to the user), integrated tools (such as single-sign-on, single administration panel, and data management), and an app store letting end user organizations easily find, purchase, install, and manage subscriptions for new components.
Tradeshift is building out their developer support infrastructure: hiring a dedicated app support team (based in NYC), building out the App developers’ self-service toolkit, training their partners, and launching an app store with reviews and ratings. They are aggressively recruiting apps for their store. They said they have about 50 apps now5 and plans to double it by the end of the year.
Expanding the Footprint to full Source-to-Pay and Beyond
Tradeshift has been using this platform approach to build out a suite of cooperating, integrated applications for a complete Source-to-Pay (S2P) system, including spend analytics, contract management, and supplier management (info, performance, risk). They are building some of the components themselves and are looking to partners for others. Because of the platform approach, it is possible for partners to build tightly integrated modules that have the same integrated look and feel, security, and availability through the Tradeshift app store. In some cases, Tradeshift has acquired these app development partners, to broaden out their own portfolio, such as their recent acquisition of Merchantry (PIM).
In many other cases, Tradeshift is partnering to expand the footprint, such as with Quyntess (direct materials P2P, logistics, SC collaboration), C2FO (dynamic discounting), EcoVadis (supplier CSR ratings), Determine (Strategic Sourcing), ProcurePort (eSourcing and Contract Management), MetaProcure (managed procurement services), Baiwang (Chinese VAT compliance), and TimeStarter (time tracking). As well, they are forming strategic relationships and partnerships with various banks and financial service providers, such as HSBC, Amex, Citibank, and CreditEase, some of whom have invested in Tradeshift. There are also integration apps for integrating Tradeshift with QuickBooks, NetSuite, and Xero, with more coming (Uber, Magento, ShoppingCart, and a few others). Tradeshift told us they had 60 app partners last year, 120 committed this year, and aiming for about 250 by the end of the year. If they can reach those numbers, they should have a rich pipeline of apps coming.
Tradeshift’s Audacious Idea
Now you might think that trying to create the largest network of suppliers and buyers in the world, along with a complete set of buy-side applications, and a robust multi-app platform and development infrastructure, as well as a set of supplier apps would be ambitious enough for Tradeshift. You’d be wrong. They have bigger plans. They are building a platform and a set of services that brings together natural language, machine learning, and a hybrid machine-human approach to providing an enormous range of services. A good way to get your arms around this is first understanding Tradeshift Go.
Tradeshift Go — Starting with T&E

Tradeshift Go is trying to solve employee purchasing pain for midsized companies, which they described as 100-1000 employees.6 The financial people in these firms try to push as much spend as they can through POs in order to get visibility and control. But that is expensive to process, supplier payments are delayed (sometimes for months), and anyway it is a losing battle, as increasingly credit or debit cards (personal or company cards) are being used for more and more spend. As we move to an as-a-service economy, substantial monthly subscription charges7 are being paid via credit card. Unfortunately, credit cards lack the visibility and control until after the fact.

Go is a chat-based virtual business assistant to help every employee do T&E (travel and expense) and everyday purchases. Go is based on technology from Tradeshift’s acquisition of Hyper Travel, hence the initial focus on travel. The user simply types in a natural language query, like “I’ve got a meeting next Tuesday in New York. Get me a flight, hotel, restaurant.”
Behind the scenes is a combination of machine learning/AI + human intelligence (a real person) at work. The machine learning understands the intent, knowing the user’s location, it figures out what they are asking for. It then prefetches the appropriate information (available flights, company-specific pricing, rules, and restrictions, hotels near the meeting, and so forth). The appropriate information and options are presented to a live agent in the back end, who can now quickly make an intelligent recommendation and/or describe the options to the requesting user.
We saw a demo on the iPhone. The Go chat box popped up. They said it could bring up a template, prefilled with the user’s home airport and a list of common destinations. The person giving the demo pointed out that while waiting for the reply, the user could easily type in a natural language request to add information to their profile, such as TSA number, emergency contacts, loyalty program numbers, payment methods, and so forth — all entered in the same conversational view and Go would put them in the profile. Go returned three different flight options, which included details like where and how long the layover is. They were able to purchase the round trip ticket in under a minute. For the hotel, it showed five options on a map. They picked the Hilton, showed the cancellation policy, and confirmed payment. They also showed us the backend, which they said gave the human agents ‘super powers’ and made them ‘orders of magnitude’ more efficient than traditional call center agents.
Witnessing a Real-World Use, in Real Time
Something happened while we were listening to all this that was, for me, more impressive than the demo. The woman sitting next to me (who happened to be the head of procurement for UPS) had to leave soon to catch a flight. While we were listening, she downloaded Go and asked it the best way to get to Newark Airport. She showed me Go’s response, which not only described the options and which one it recommended, but some very specific instructions that sounded like they must have come from a person with local knowledge of the subway and train system — something like “be careful, there is a similar sounding train — don’t get on that one, get on this one” and “the price will be $X, but there is a surcharge added to that.” Though pure AI assistants get better every day, that kind of context-specific information and response seem like something that only a human can provide (for now).
Virtual Credit Cards
One of the things that makes Go powerful is the virtual credit card, essentially a one-time-use credit card issued uniquely for each purchase made. Later uses of the same card will be rejected, thereby dramatically reducing theft and fraud. Tradeshift is working with Amex, HSBC to figure out how to roll it out globally and piggyback onto their networks.
Other Expenses Beyond Travel
The concept of the virtual one-time credit card is being incorporated into Tradeshift Buy to address off-contract or spot buys. This helps accommodate the reality that not everything is bought off of the catalog or negotiated contracts. It allows employees to buy from anywhere on the Internet, while giving full visibility and control to managers via approval workflows and dashboard. Approved purchases generate a single-use credit card. This is one of the keys to move towards the goal of capturing and managing 100% of spend.
We saw a demo of Go as a browser plugin (they said in the future it will integrate with other tools, like Slack). In the demo, the employeewas shopping in the Apple online store, saw an iPad they wanted to buy, and opened a Go chat window to put in a request for the $800 purchase. They then showed the perspective of the approving manager, who was able to see the request, check her budget, and approve it. The buyer then saw the approval and the one-time credit card information auto-filled for them. In a real-world scenario, the manager might not be quite so real-time in their response, but this kind of system should still help a lot in reducing maverick spend.

Go is available for free, which includes the virtual assistant concierge service for travel, approval workflow, and payment via virtual credit cards. In exchange, Tradeshift gets a portion of the card rewards given by the issuing banks. Tradeshift has several beta customers already using Go.
Tradeshift Skills
Tradeshift Go by itself is cool enough — a new paradigm in the T&E world. But they have a bigger, grander vision. Christian Lanng (right), Tradeshift’s CEO, talked about the blending of services (BPO8 and other types of service) on their platform. This is where Tradeshift Skills comes in. It takes the concept that Go provides for T&E and extends it to other areas to provide combination AI/human powered services through the platform. The idea is to fill the expertise gap within an enterprise. Tradeshift is building Tradeshift Skill Store, a one-stop access point for all the Skills available on their platform.
Tradeshift ispartnering with major BPO providers, including Wipro, Infosys, Xerox and others. John Gentry, President of Finance, Payment, and HR Services for Xerox said that the Tradeshift Skills platform will give them a whole new sales channel, and a way to move beyond classical BPO processing to much more of an automated, as-a-service model. He said the market is moving to technology-assisted approaches; we are already seeing robotic process automation helping with a lot of grunt work tasks and it won’t be long before technology takes on higher order tasks.
The dividing line between what machines do well vs. what people do well is continually moving. Those who provide human-based services need to adapt. This blended approach is a great way to stay in the game as machines get smarter and smarter … using the people for the things machines still can’t do, even as that line continually evolves. Tradeshift is planning on launching Skills soon, rolling it out to some of the existing customers.
Including Skills in Tradeshift Buy

With Skills, Tradeshift is potentially disrupting their own development within enterprise, so they are figuring out how Go and Skills actually work in the Enterprise. Skills could be included inside Buy within the Products Engine, Shop, and Collaboration. We saw a sneak peek demo (still in development) of someone using Shop to look for engineers who could help set up a network in a new Atlanta office. The user types in “need IT engineers in Atlanta office to set up Wifi and internet, target price of $8K.” Based on the nature of the request (in this case based on the UNSPSC code), the system realizes that the organization doesn’t have the internal procurement expertise for this request. So instead of routing the request internally, it routes it to a temporary contractor procurement agent, who has expertise in IT installation projects.
Similar to the way Go parses the request for the travel agent, Skills parses the request and makes it quick and easy for the procurement agent handling it. In the demo, the agent responded by asking a question about the timeframe for the project, since it wasn’t specified in the original request. The user says “right away.” The agent was able to suggest a few different IT managers to run the project, including their photos, bios, recommendation, and experience. The user selected one, and then a purchase request was created and sent right to the IT service provider. They responded with a quote, sent back right there within the same app and the user was able to approve it.
The chat stream was combined with live documents used in the exchange (SOW, agreements, etc.). So, here we went from a simply worded natural language request to the complete procurement of a complex service, without the buying organization needing to hire the person with the skills needed to complete that complex purchase. Pretty impressive vision, if you ask me — the on-demand delivery of a skilled workforce, on an as-needed basis. As Christian put it, “We can give smaller companies the power of a massive, skilled procurement department.”
Expanding the Skills Available
Today, the human agents behind Go are all Tradeshift employees. Tradeshift does that to ensure quality as they get off the ground with these services. But to meet this grander vision, at scale, within any reasonable timeframe, the skilled procurement folks behind the scene can’t possibly all be Tradeshift hires — hence the partnerships with the BPOs. And even beyond BPOs, the platform could accommodate skills from organizations like 3PLs9 and freight forwarders, who can provide assistance when someone needs to move, store, or import something.
Turning Cost Centers into Profit Centers
Some companies have some strong competencies in their internal central services organization (procurement, finance, or other back office functions) or strong skills in engineering, field service, supply chain, or any other area. The Skills platform provides them with a pre-built framework and marketplace to turn that internal cost center function into a profit center serving external customers. Becoming a BPO has historically been a big business model change and investment commitment. For example, IBM had a very strong procurement organization, which they turned into a BPO service for others. For IBM, this was not as difficult as it might be for other firms, since IBM was already a consulting company providing other services. They already had in place the business model, sales force, billing mechanisms, and so forth, so they could add procurement to their pallet of BPO services.
But think about a completely different company, say an automotive OEM that is really good at buying steel and other metals. They could aspire to use those skills to buy steel on behalf of their suppliers (perhaps even leveraging spend volumes), and perhaps for others as well. To do that kind of business themselves from scratch, they would need to acquire a platform, put together a sales and marketing organization for that service — all things that are way outside their core business of building cars. Most likely they won’t bother, even if they really are very good at buying steel. A platform like Tradeshift Skills takes a lot of that friction out of the process. The company can plug into the platform as a service provider. On the other side of the equation, that very same automotive manufacturer might decide to jettison some of the resources where it is weak or for tasks it does infrequently and instead use the same Skills platform to fulfill those needs.
Enabling the On-Demand Hyperspecialized Economy
This approach makes it easier for companies to take what they do best and turn it into profit centers, and get rid of what they are not so good at or do infrequently. You could argue that the Skills platform approach thereby encourages hyperspecialization and the dynamic fluidity of enterprise skill sets, as described in Outsourcing and the Age of Rapid Startups and Hyper-specialization, which talks about the move to an as-a-Service economy. Why keep resources in-house that you aren’t that good at, or that you do infrequently enough that makes it hard to justify keeping an expert on staff (which includes a lot of things for a smaller company), especially when it is so easy to find someone with that expertise, on-demand, just when you need it. Tradeshift, that started as an invoicing and payment firm, is now proposing to propel us into a whole new way of doing business, the 21st century as-a-Service, on-demand economy. It may be incredibly ambitious, but I for one hope they wildly succeed.
_________________________________
1 This includes fax, .PDF, word docs, etc. Their machine learning technology meant that conversions to a common digital format are to be done with fewer and fewer errors, approaching zero errors over time for commonly used external unstructured documents (such as QuickBooks-generated .PDF invoices). — Return to article text above
2 Networked SaaS Platforms are also sometimes referred to as ‘multi-enterprise grids.’ — Return to article text above
3 By comparison, other G2 Crowd ratings: Ariba: 3.6 with 29 reviews, Coupa Procurement 4.2 but with only 12 reviews, Concur 3.9 with 229 reviews. — Return to article text above
4 United Nations Standard Products and Services Code — Return to article text above
5 I saw 20 apps listed on Tradeshift’s site. There may be more apps available ‘behind the scenes.’ — Return to article text above
6 There are about 100K companies in the US with 100 to 1,000 employees according to the US Census Bureau. — Return to article text above
7 For example, Tradeshift said they pay their Amazon AWS bill via credit card. I’m sure that is big. Some very large purchases are made via credit card. The largest single credit card purchase to date was $170M. — Return to article text above
8 Business Process Outsourcing — Return to article text above
9 Third Party Logistics providers. — Return to article text above
To view other articles from this issue of the brief, click here.