Coupa acquires AI-powered supplier discovery platform Scoutbee

Spend management vendor Coupa adds an important component to its supplier management capabilities today with the acquisition of Scoutbee, an AI-powered supplier discovery platform. This is Coupa's second acquisition this year, following on from its purchase of category management vendor Cirtuo, which it unveiled at its annual Inspire conference in May. Leagh Turner, CEO of Coupa, told us then that the company was ready to do more acquisitions to put the components in place towards its goal of building an agentic network for collaborative commerce.

If you've ever tried to shop for a product in a specific color, size or model on Google or Amazon and found yourself wading through hundreds of irrelevant search results — and who hasn't? — then you've got a good idea of the problem Scoutbee was founded to solve in the B2B space, where buyers typically have detailed specifications for their purchases. We spoke to Gregor Stühler, the company's co-founder and CEO, on Friday to find out more about what his company brings to Coupa. He sums up:

We are essentially a matchmaking engine. If you have a demand as a buyer or a sourcing event that you need to fill, [you] just fill in the sourcing event [and] we know enough to recommend you the right suppliers. I think that is way beyond what a common search does.

Traditional search engines rely on keywords and categories to find results, but this approach falls short for B2B buyers, whether they're using Google, Amazon, or the specialized marketplace search functions in the likes of Coupa and Ariba. This is because buyers are often looking for a highly specific 'long tail' search result that won't stand out in the midst of the generic results produced by these traditional approaches. Stühler gives an example:

There was a customer that was looking for 35-liter, top-loaded, verticalized sterilizer, an autoclave. So very specific. If you type this into Google, you will find all those autoclave providers. But if it's actually a 35-liter and top-loading, it's very specific, and you will not be able to form a query to get that done well. That's where we shine, essentially. That's where we go and break down the website and all the web content to say, 'Hey, this is exactly the part that you are looking for.'

The other issue with general-purpose search engines is that they're optimized to serve results based on an advertising model, which, for example, when consumers are shopping for something, prioritizes suppliers from their local area. A B2B commerce network has a different set of priorities. He explains:

If I'm a B2B buyer and I'm looking for coffee beans, I want to actually get results from the hubs, from Colombia, for example. Search engines are just not designed for that.

Supplier landscapes

Scoutbee's origins date back to the early 2010s, when Stühler was project manager for an electronic drive manufacturer and developed a system for discovering supplier information to be better prepared for supply chain disruptions. Invited to speak at a meeting of the German automotive industry to discuss the sourcing of rare earth materials, he realized how much pent-up demand there was for this type of database:

I realized how much need is there for simple supplier information, to map supplier landscapes worldwide. That was essentially the trigger moment.

One of the biggest difficulties in building such a system to serve this kind of B2B market is the specialist language that it needs to understand. He explains:

Mapping supply capabilities is very complex. It's not like a shoe, where you say you have a brand, a size and a function and a color. Suppliers have very different dimensions.

Scoutbee solved this by building — in 2016 — something that was effectively a precursor of today's Large Language Models (LLMs). The purpose was to find results based on the meaning of the search terms, rather than relying purely on predetermined keywords. He goes on:

It was called a hyperdimensional technics cube. We used that, essentially, to put the supplier website in a vector space. That was essentially our attempt to put an abstraction layer on the supplier website, so that we do not have to compare keyword by keyword.

That was the biggest issue actually with search engines. If I would search, for example, for VA steel CNC drilling and milling, it would not show stainless steel, but it's actually stainless steel. What we have essentially created was a text cloud or text cube, where those relationships between the words were already mapped. That enabled us to go way beyond standard search engines, and was much easier for us to identify the right supplier.

Scoutbee's supplier sign-up process is highly automated with AI support, with a basic profile already created by ingesting supplier data from websites and other sources of publicly available information. He explains:

As soon as the supplier says, 'Hey, I want to maintain my profile,' they log in, and the whole process from logging in to then completing the profile in a very well done manner, that's minutes, five to 10 minutes, and then you have essentially a very good profile awareness.

While the AI will take care of elements such as drafting a company description that the supplier can then edit and approve, the supplier is encouraged to add information that typically isn't publicly available. He goes on:

There are always certain data components that are not public on the World Wide Web, and intentionally so. Sometimes suppliers do not want to disclose certain products to competition. Some buyer or customer relationships cannot be disclosed. That's where we try to motivate the supplier, of course, to provide more information and to also make them rank higher. Essentially, the more our AI can trust the supplier, the higher they will also rank in the search results. That's where the supplier is very much motivated to provide additional data.

Network scale

Scoutbee has already been successful in creating a network of suppliers, but teaming up with Coupa gives access to a huge community of buyers. He explains:

What Coupa already has is essentially the missing piece for us — and we had the missing piece for Coupa. So we had essentially a matchmaking AI connected to a network of a million suppliers, with a very active community there. And on the other hand, Coupa had this extremely wide set of customers and hundreds of thousands of buyers that are daily logging in.

Putting this together, it's essentially automatically forming the biggest supplier-buyer network in the world. And that is essentially where we realized, okay, if we stay standalone, we might achieve our ambition, but it's definitely five, six, seven years out. But if you join Coupa, we can accelerate that and potentially be there in just a couple of years and have a massive impact.

He is also supportive of Coupa's goal to populate that network with autonomous agents that will take care of much of the groundwork that currently soaks up so much time and resources. He goes on:

How do buyers and suppliers do business today? It's a super-high friction market. If you're a supplier, you blast out emails, you pay $200,000 to attend a conference, a qualified opportunity often costs you often $5,000-10,000 just from sales and marketing spend. And all of that spend is essentially just creating noise on the buyer end.

What the buyer gets is this blast of marketing, and they cannot really actually filter that out. This filtering, however, is very unemotional. It's very deterministic to say, 'Hey, either you have this product, the credentials and the price, or you don't.' So the pre-filtering, in my mind, can be actually done, in the future, very much autonomously on agents. Filtering down from the 10,000 opportunities of suppliers down to the top ten, I think that can be done autonomously in the future, and that can be done by the agents.

Then later, however, when it comes around to, how do those two parties click? Do the innovation roadmaps map? Do they see eye-to-eye? I mean, there's also always a personal, emotional element in the final decision. That's where, I think, the human comes in and pulls the trigger. I don't think that this will be automated in the future. Maybe for like, tail spend, but not the bigger decisions for big projects.

My take

As Stühler points out, there are enormous opportunities to simplify the buying and selling processes for businesses and strip out inefficiencies. This acquisition slots another key building block into place along Coupa's roadmap towards that goal of an autonomous commerce network.