Chatbots are learning a new trick: Instead of simply interacting with people, they are starting to speak to each other.

Within the past nine months, bot-related messaging volume has grown from zero to an estimated 10% of the total message volume on Symphony’s open-source messaging platform, David Gurle, founder and CEO of Symphony, told Markets Media.

“We now have more than 600 chatbots in production and several hundreds of them in development from our customers,” he said. “Those chatbots are a proxy for a range of business application. Moreover, in some organizations, chatbots represent 55% of the total message volume that we have seen within some clients.”

Gurle initially estimated that it would take the Symphony users base until 2020 or 2021 before they would start to build applications using the platform’s architecture.

“I thought first people would focus on the collaboration use case, straighten it out, and then start building applications on top of the platform,” he said. “I guess I underestimated the need for simplicity and automation, as well as the cost pressure that exists in the industry.”

Although most chatbot traffic with human users represents the vast majority of chatbot-related messaging traffic, chatbots are beginning to communicate between each other.

It is too early to estimate just how much traffic chatbot-to-chatbot conversations, four banks demonstrated their chatbot-to-chatbot connections during a recent Symphony event earlier this month.

Deutsche Bank and BNY Mellon have connected their back office chatbots, Debbie and Selina respectively, to communicate trade resolutions, trade allocations, and trade-break denials, according to Gurle.

He cited a second instance where AllianceBernstein and the Royal Bank of Canada have linked their front offices via Abbie and Arthur, respectively.

The growing client interest in chatbot-to-chatbot communications has sparked a conversation within a Symphony working group on developing a standard lexicon to simplify the process.

Gurle expects that a resulting standard would be closer to an extensible markup language than to other financial messaging protocols, such as FIX or ISO 20022, due to the need size of the needed vocabulary.

“We are kicking this off with the buy- and sell sides,” he said. “They already have standardized a common set of hashtags that define the vocabulary that people can exchange with each other.”