\n<\/p>\n
As conversational AI begins to take over the world, chatbots are being given a new lease of life. However, there are clear hazards on the road ahead.\u00a0Parcel delivery giant DPD recently had to disable part of its online support chatbot after it swore at a customer<\/a>.<\/p>\n A user managed to convince the chatbot to be heavily critical of DPD, and convince it to criticize the firm in the form of a haiku.<\/p>\n The incident is even more of a warning sign for regulated industries such as healthcare and insurance, where leaving the heavy lifting to a haphazardly designed AI chatbot could leave the organization open to legal action.\u00a0 OpenDialog<\/a>, a UK startup from some serial entrepreneur founders, thinks it has an answer by combining the new world of LLMs with a customisable platform for regulated sectors. <\/p>\n It\u2019s now raised $8 million in a Series A round led by Alboin VC. Dowgate Capital and several notable angel investors also participated, meaning the startup has now raised $13 million in total.\u00a0<\/p>\n The demand for conversational AI is skyrocketing, and is set to explode<\/a> to a mind-boggling $38 billion globally by 2029.\u00a0<\/p>\n However, regulated sectors are still grappling with Natural Language Understanding (NLU) and Large Language Models (LLMs). OpenDialog says its focus on healthcare and insurance will make it a good fit for those industries.\u00a0 OpenDialog claims its no-code platform could allow insurance giants and the likes of the UK\u2019s NHS to mix and match the best AI models into one automation platform.\u00a0<\/p>\n It also claims one insurance client automated 9 out of 10 tasks, from selling policies to handling claims. What Walby is hoping of course is that by not being built on one foundational model, and having a more flexible platform, OpenDiolog will be able to mix and match various LLMs so that clients can get the flexibility they want.\u00a0 He counters, saying that disaggregating from a single language model means it can \u201cexploit them for certain parts of the conversation for which they are appropriate.\u201d<\/p>\n Ed Lascelles, Partner at AlbionVC, commented: \u201cConversational AI has evolved massively since early chatbots\u201d and noted that the OpenDialog has \u201cbuilt and scaled enterprise businesses before.\u201d<\/p>\n Indeed, this is the second startup for the founding team, who scaled and sold Thoughtonomy, a software automation solution, to Blue Prism in 2019<\/a>. <\/p>\n<\/p><\/div>\n
While an LLM might be able to sound like a human and understand context, regulated industries need high guard rails on an AI-driven approach. \u00a0<\/p>\n
Co-founder and CEO Terry Walby, told me over a call that the platform has a \u201cfluid architecture which means you can manage the conversation in a way that\u2019s much more like humans have conversations, which is not predictable. It can expect the unexpected, and I can react to it, respond to it, and not lose the flow of the conversation.\u201d<\/p>\n
That may be the case but it could leave OpenDiolog open to the charge that it is a mere \u2018middle-ware\u2019 player rather than having a robust platform of its own.\u00a0<\/p>\n