Content Is Key to Conversational AI Success
Conversational artificial intelligence and automation isn’t just for the contact center anymore. There’s growing use of the technology by marketing and sales to deliver real-time interactions and outreach at scale, according to a new report from Forrester Research.
However, the research firm cautions that quantity isn’t always accompanied by quality. Many organizations are relying on technology rather than strategy to drive their conversational approaches, Forrester says.
The quality of the conversation can’t be overlooked, according to Jessie Johnson, a Forrester principal analyst, who says AI is becoming increasingly important in conversational analytics. “When we think automated conversations, predominantly, there’s some level of conversational AI in play. We have automated conversations that take place through decision trees and automated rule sets,” Johnson adds, noting that humans can still control automated conversations. “We don’t need to be completely reliant on the robot overlords to run these conversations.”
Still, many of these technologies are getting, “very, very good at being able to interpret a response or an intent from that buyer or customer,” Johnson says.
The research also found major differences in business-to-business conversational interactions compared to business-to-consumer ones. This includes the number of interactions needed before completion of the sale. With B2B, the average is 27. B2C cycles are harder to quantify because so much depends on the type of purchases, the consumers themselves and their individual behaviors and attitudes, peer influence, and previous experiences with the brand or product, Johnson says.
“When [a B2B prospect] is on your website, they’re not looking to buy right now. They are likely doing some research that is part of a considered purchase, then they will take that back to a group,” Johnson says.
B2C purchases, however, are usually made by an individual or a household, typically not by a group.
Sellers and marketers need to know in which basket each potential buyer falls. “We want to understand, is the person that we’re interacting with that buyer customer? Are they the influencer of that decision process? Are they the end user of a product? Are they all about [total cost of ownership], looking at the total cost to see if they can get something cheaper? The role they play within the buying group gives us a really important signal,” Johnson says.
But even with those differences, there are some important similarities as well, according to Johnson. “We’re all human at the end of the day. Sometimes with B2B, we try to get a little more formal with the language, then we try to drive the sale, which isn’t always what the customer is participating in the conversation for,” she cautions.
B2B sellers should prepare for the following three basic types of conversations:
- Enable and influence. These conversations are at the very beginning of the buying process. They focus on understanding, predicting, and responding to audience needs at the individual and group level across various scenarios.
- Transact and sell. These conversations support the earlier stages of the customer life cycle. The report recommends considering routing rules that respond to buying group signals, high purchase intent, scoring thresholds, and other factors.
- Support and serve. These conversations proactively provide information to support the customer purchase and to increase use of the product or service after the purchase.
So when designing these conversations, how they flow is becoming increasingly important for conversational AI, according to Johnson. The right content needs to be integrated into the B2B content engine to do this, and so organizations need to do the following:
- Create modular conversation content. By connecting appropriate phrases, companies can build good automated responses for chatbots. These phrases are modular content that, along with related metadata, will lead to more value-producing conversations.
- Define conversational taxonomies. Metadata and taxonomy are essential for delivering the right content and learning from responses to it.
- Experiment with a primary-derivative approach to the conversational content. Organizations can experiment with different inputs for content to see what produces the best results.
But while modular content is so key to success, Forrester’s 2022 State of B2B Content Survey found that only 8 percent of respondents use modular content for chatbot conversations. That should be much higher, according to Johnson.