The generative AI boom kick-started by ChatGPT in late 2022 has led to a number of companies and brands introducing conversational assistants.
Powered by generative AI, these assistants introduce a conversational layer between the user and a platform or service. They offer new features that can be accessed via natural language, or a streamlined way of accomplishing certain tasks.
Consultant Neil Perkin, speaking during Econsultancy’s latest Digital Shift presentation, remarked that the integration of conversational assistants by companies like Meta across their owned platforms shows “how much they see AI as central to the way that we will interact with their services – but also each other – in the future”.
How are these assistants influencing the way that brands interact with consumers, and what opportunities – and drawbacks – does that present? Here are some key examples highlighted by Perkin during Shift that shed light on where this trend is headed.
Generative AI has become a significant area of focus for Meta, leading to the release of a proprietary assistant, Meta AI, which has since been embedded into Facebook, Instagram, WhatsApp, and Meta’s Ray-Ban Smart Glasses (from generation 2 onwards).
“Users of Meta’s platforms will be able to message Meta AI … to ask for advice, answers to questions, and even recommendations,” said Perkin, while also highlighting the AI’s image recognition and analysis capabilities that allow users to ask questions about things they see around them.
The assistant, which launched in the UK last month, can even be added into a group chat as an additional participant in order to answer questions that arise during a conversation.
If Meta AI and assistants like it, such as Snapchat’s ‘My AI’, find favour with users, this could see social media supplanting search for some uses cases that would previously have involved a search engine, keeping them within the social network’s ecosystem instead.
Many of these inspiration and research use cases, such as recipes or travel planning, suggest opportunities for brands to be present in the ‘conversation’, although in the case of Meta AI, there is no obvious indicator as to how likely any given brand is to be recommended by the AI. (Snapchat has begun offering ad placements within My AI in which sponsored brands are surfaced depending on the context of the conversation).
For this, the “share of model” metric – highlighted in Shift’s opening chapter, which focused on applications of generative AI in marketing research – may come in useful.
The ability for creators and businesses to build their own chatbots using Meta’s API … further compounds the potential reach and also growth of this type of technology.
At Cannes Lions this year, Meta announced the rollout of a customer service solution using AI chat, beginning with select businesses such as medical apparel company White Coat Manila. The underlying generative AI has been trained “on the most popular questions businesses receive on Messenger”, according to Meta, and can respond to simple queries such as colour and style recommendations or detail on return policies.
Users can also initiate a purchase within the chat, although this causes the customer to be transferred to a human agent to complete their transaction.
Generative AI can be very useful in triaging basic customer service questions, as long as it responds accurately; and Meta is clearly hoping that gen AI’s more fluid conversational capabilities will ease users along the sales funnel in the same way that a helpful shop assistant can by responding to queries ahead of a potential purchase.
Meta’s AI Studio, launched in July, has also opened up conversational assistant creation to users – who can create a specialised AI ‘character’ to interact with – and influencers, who are invited to create a virtual extension of themselves to answer “common DM questions and story replies” from fans.
“The level of integration we’re seeing into Meta’s platforms extends the reach of AI agents dramatically, mak[ing] it far more easy for people to become used to interacting with an AI,” Neil Perkin commented. “The ability for creators and businesses to build their own chatbots using Meta’s API … further compounds the potential reach and also the growth of this type of technology.”
While much of the investment into and integration of generative AI is coming from tech behemoths like Meta and Google, brands are also finding ways to embed this technology directly into their product or service.
Perkin gave one example from the world of dating apps, several of whom have introduced ‘AI wingmen’: conversational assistants that can give users tips on chat-up lines, generate icebreakers, and help them navigate the world of online dating.
“The [wingman] concept is useful to consider when it comes to AI agents – to show how they can be integrated into customer journeys; to enhance personalisation; [and] to reduce user friction in onboarding or interaction phases,” said Perkin. On dating app Volar, for example, users create a profile by messaging with a chatbot instead of filling in a form.
With this said, app users have remarked that they would avoid partners on the app who appeared to be using AI – while online dating can’t be said to be entirely authentic, users no doubt prefer when it is more authentic rather than less – and a Wired article noted that Volar’s chat tool “got weird” when messaging with potential matches.
A chat-based onboarding experience is a genuinely novel idea that offers a solution to a notoriously tedious process – filling out forms and creating an account. In general, however, brands need to think carefully about how they integrate conversational assistants and what the experience will be like from the user side: will this help, or harm, what users want to achieve? Can the problem be addressed effectively without a conversational assistant?
The [wingman] concept is useful to consider when it comes to AI agents – to show how they can be integrated into customer journeys
I previously wrote in an Econsultancy briefing about AI ‘copilots’ that “Brands need to avoid falling prey to ‘copilot mania’ (no matter how optimistic the think-pieces are) and give careful consideration to what, if anything, a conversational UX would add to their product … even if this might mean passing up a chance to jump on the generative AI bandwagon”.
We’ve been down this road before in the 2010s with ‘chatbot mania’, and at the time the effect was generally to prompt a backlash from users sick of interacting with bots. As conversational assistants become more ubiquitous, their usage may become normalised – but users may also become less tolerant of them if their experience with assistants has primarily been unhelpful.
Ideally, conversational assistants will be a seamless extension of the user experience – unremarkable because they simply accomplish what is needed without any additional friction.