By Barb Mosher Zinck

August 21, 2023

Reaching customers where they spend their time has become increasingly important for B2C brands, and social messaging apps like Facebook, Instagram DM, and WhatsApp are the perfect place to offer personalized product information and discounts and provide ongoing customer support.

I talked with Max Koziolek, CEO of Spectrm.io, a conversational marketing platform that enables brands to use social messaging to engage with customers about the growth of this channel and the impact generative AI has had on its growing popularity.

According to Koziolek, Spectrm began as a solution for media companies, sending out articles to people, but the media market was small, and they realized there was a greater need for marketing. So, in 2017, the company pivoted to support B2C customers.

Koziolek argues that messaging and chatbots in social feeds are a great way to engage with customers in a privacy-conscious way, but it does require getting people to opt-in to receive messages. He explains that offering some incentive is typical, whether a discount on a purchase or something else. Many brands use ads that drive people to Facebook Messaging or WhatsApp, or they share links on their social channels, website, content, and emails.

There are 931 million people using Facebook Messenger and 2.78 billion people worldwide using WhatsApp. Many are using these apps to talk to brands about everything from questions about products and services to customer service support. They are also receiving content and offers from brands.

In Spectrm’s State of Social Conversations 2023, 78% of respondents messaged a brand through Facebook Messenger, Instagram DM, or WhatsApp, and 86% said that experience positively impacted their perception of the company. In addition, 82% made a purchase after messaging with a brand. In another study Spectrm did on B2C retention, 35% said these messaging apps are a top tool for retention.

What generative AI brings to social messaging

The company initially received funding from Google to build conversational AI. This was when natural language processing (NLP) was the primary model. They started with intent messaging, looking at exact match, keyword matching, context (using an algorithm called word2vec), and GANs (generative adversarial networks). And now Spectrm applies large language models (GPT 2 is currently implemented in their platform).

Generative AI has enabled brands to move away from orchestrating journeys through manual processes (think the decision path process you had to do with early chatbots). By combining LLMs and company data, you can more quickly understand what the user wants from the brand (and the bot) and generate a more natural conversation. Koziolek also said that by using LLMs and company data together, Spectrm has put in place a system to avoid the hallucinations that LLMs typically have, especially when the AI generates something that hasn't been seen.

Koziolek says that as people ask questions, the AI builds a corpus of data around what is going to happen, generating potential avenues and identifying the different ways people might ask for something. This approach can cover up to 95% of the requests. For the rest, Spectrm has a brand security mechanism where they will check the response to be sure it's not completely made up.

Concerns around privacy and engaging with a bot

There is a lot of excitement around using conversational AI now because more is possible than before. As for brand concerns, they exist, and Spectrm works to ensure the legal checks are done and that they aren't sending data back to Open AI.

I asked Koziolek if their customers tell people they are talking to a bot, and he says they do say it is a bot. Most people don't care if they are talking to a bot, though, according to Spectrm's study, 56% don't care if they message with a bot or a human, so long as they get what they need. Of those who said they'd rather speak to a human, 63% said they'd message with a bot, as long as they were able to receive instant and useful responses.

People tend to assume they are messaging with a bot, Koziolek says, even though many conversations are hardcoded via a decision-tree-like experience:

A lot of stuff is hardcoded, right? So a lot of stuff is a decision tree, “if this, then that,” which is the right thing because the user has an intent and wants to get something from the brand. They're not chatting for the sake of chatting and talking about the weather. I think in many ways, it's completely also still very important to have these crafted user journeys in order to guide the user because my take is that UI beats AI.

What does that mean, "UI beats AI?" Koziolek says everyone says the conversational interface will take over everything and that people just want to type and chat. He doesn't see it that way. He said that people want to get their answers fast. They want to see the options, and they are very visual.

Whatever is the fastest way to get their answers, that's what they want. That's why he thinks the UI will beat the AI every time. Ultimately, it's the customer's choice of how they want to engage, so brands need to consider providing both options, and both experiences need to be efficient and effective.

My take

From a consumer perspective, I have no issues dealing with a chatbot if I can get the answers to my questions quickly. And that's the thing most consumers want - answers quickly. Getting those answers via social messaging apps appears to be the way most younger people (Gen Y and Gen Z) want to engage, according to Spectrm's study. But a significant number of Gen Xs were also responding to the study.

If a brand's customer base includes Gen X, Y, and Z, not having a social messaging strategy could be a significant loss. They may not need full-blown commerce built-in to start, but they need some engagement. Platforms like Spectrm enable brands to implement everything from lead generation to selling to retention and loyalty programs. It's a long way from where they started sharing media articles, and it's much bigger in-demand market.