AI for Humans

To understand humans’ complex relationship with artificial intelligence (AI), start with the stories we tell about it.

Movies and TV, from Stanley Kubrick’s 2001: A Space Odyssey in 1968, to 1999’s The Matrix, and recent HBO series Westworld, present AI as a computerized version of the Icarus story. Inevitably, people create an AI so perfect that it’s practically human, only to realize too late that it’s too smart for our own good.

In the five decades since Kubrick’s masterpiece, AI has become deeply embedded in our everyday lives. If you used predictive text today or asked Amazon’s Alexa a question, you used AI.

AI has also become a buzzword in the buzzword-loving tech world. Everything is “powered by AI” to the point that the phrase has become meaningless.

Except it’s not meaningless. AI is a powerful tool and developers are constantly making it smarter. But like any tool, it’s most effective when it’s designed and used to meet a specific need.

For example, Conversational AI uses natural language understanding (NLU) to more accurately comprehend customer questions and to communicate helpful answers. It can’t tell you that there’s a traffic jam on your usual route home, or if you need an umbrella. That’s not its job. But it can help with converting website visitors into customers, scaling your sales and marketing teams, and a whole lot more.

Despite the ubiquity of AI today, some of the fears generated by those fictional versions remain. People worry that AI is too powerful; that it can think for itself, and use that insight to take over humanity — or at least steal our jobs. On the other end of the spectrum are the skeptics. They think it can’t possibly do everything that’s been claimed.

The truth is that AI is both more insightful than those artistic depictions envisioned and far less threatening.

This ebook will look closely at Conversational AI in particular, in an effort to explain how it works, how it’s helpful, and why it’s time to stop worrying and embrace the robot.

Kyle Bastien

👋 Who they are: VP of Sales Readiness at Drift

👀 What you should know about them: It became clear that Kyle knew more than a thing or two about AI when he was running sales enablement at Drift and quickly became the go-to person to lead demos of our first AI product. As AI has gotten smarter about people, Kyle too has become an expert on AI.

🔥 Their bot take:  “I think the measure of a good AI is if it’s useful. Does it remove friction from a process? It’s not trying to be human. It’s simply trying to make decisions and automate workflows.”

Aurelia Solomon

👋 Who they are: Senior Director of Product Marketing at Drift

👀 What you should know about them: Aurelia understands all the ins, outs, and benefits of Conversational AI. More often than not, you can find her geeking out about AI with Drift’s product team and customers. Her enthusiasm for the power (and potential) of AI is pretty infectious.

🔥 Their bot take: “The best thing about Conversational AI is that it provides a frictionless experience for customers — they ask questions in their own way and get answers really quickly. And the answers are much more relevant because topics are 10x more accurate than keywords.”

Rob Stevenson

👋 Who they are: Director of Performance Marketing at Zenefits

👀 What you should know about them: Drift customer Zenefits sells human resources management software. Rob originally saw Drift’s Conversational AI as a way to reduce waiting times for customers chatting on the website. As the AI developed, so did the relationship. His once skeptical SDRs now see Bottie Botterson (the name of their bot) as a strategically valuable member of the team.

🔥 Their bot take: “AI for us is about accelerating a customer’s journey to whatever piece of content, information, or service they’re looking for. I could staff it with people — but what happens at three in the morning, when I’ve got somebody coming to the website looking for their 1989 W-2? I have a tool now that will serve their needs.”

Monique Lemieux

👋 Who they are: Head of Marketing Operations at Drift

👀 What you should know about them: Monique understands first-hand the challenges facing marketers today, especially in B2B. As part of Drift, she also recognizes AI as an exciting tool to help her and others achieve their most ambitious marketing goals.

🔥 Their bot take: “I think of Conversational AI as another tactic you can use to get smarter about the conversations you’re having and get more personalized about the recommendations you make, in a way that doesn’t require your team to do a ton of back-end work.”

Although you probably encounter the term “artificial intelligence” regularly, you may have forgotten what it actually means. In this section, we’ll explain what people are talking about when they refer to AI — including examples you use for fun and work — and dig into Conversational AI specifically.

What Is AI?

Don’t worry; acronyms are weird like that.

Early AI adopter IBM writes: “Artificial intelligence leverages computers and machines to mimic the problem-solving and decision-making capabilities of the human mind.”

Dictionary-writer Merriam-Webster defines artificial intelligence as:

“1: A branch of computer science dealing with the simulation of intelligent behavior in computers.

2: The capability of a machine to imitate intelligent human behavior.”

As you can tell from these definitions, AI has a broad meaning. “AI is an umbrella term that explains a bunch of very distinct technologies,” explains Kyle Bastien, Drift’s VP of Sales Readiness. “But generally, it’s training a machine to recognize patterns, and then do things automatically in response.”

Examples of AI that you likely interact with in daily life include:

• Facial recognition software, e.g. the kind that unlocks your phone
• Short video summaries and recommendations on streaming services
• Spam filters on email inboxes
• Personalized social media feeds
• Personalized ad recommendations

AI Is Already on the Payroll

Many companies are also using tools based on AI internally as business leaders are starting to appreciate AI’s efficiency and speed. For example, customer relationship management (CRM) tools, customer data platforms (CDPs), and accounting software can all benefit from AI that observes and analyzes vast quantities of data and responds appropriately, often across different programs.

AI is also becoming more and more popular with sales and marketing teams. In Drift and Marketing AI Institute’s 2022 State of Marketing and Sales AI Report:

• 51% of respondents predicted that AI will be critical or very important to their marketing success over the next 12 months.
• 41% of respondents predicted that AI will automate half (or more) of their tasks in the next five years.
• The three most common uses of AI in marketing today are large-scale personalization (41%), extracting insights from data (40%), and accelerating revenue (40%).

Clearly, many companies have set their sights on AI and they anticipate it will have a massive impact on businesses for many years to come. The bot take here? If you aren’t planning to put AI on your payroll, you’re going to be missing out.

What Is Conversational AI?

As Kyle explained, AI is a broad term that covers a wide range of technologies.

For example, another popular buzzword, machine learning (ML), is a branch of AI. Entire books can and have been written on the subject — but the one-line version of ML is that bots are told to complete a task, they try over and over, are corrected by another bot or by a human, and use that information to gradually get better at the task.

Conversational AI uses an advanced version of a branch of AI called natural language understanding (NLU). In its most basic form, NLU can recognize keywords around a topic. For example, if someone types in a query that includes the word “price,” “priced,” or “cost,” the AI understands that they want to know about pricing, but not the exact nature of their question.

Common examples of NLU include:

• Automated phone trees
• Predictive text

Conversational AI takes NLU to the next level. It can recognize not just individual words, but the deeper context of queries. Common examples of Conversational AI include:

• Virtual assistants, e.g. Amazon’s Alexa and Apple’s Siri
• Google Search

Google Search is a helpful example of the sophisticated NLU method used in Conversational AI. In 2018, Google switched from keyword searches to a technique called Bidirectional Encoder Representations from Transformers (BERT). Instead of looking at each word in a search query individually, this AI looks for relationships between words in order to build context.

🤖 Fun fact: BERT is estimated to have helped Google better understand 10% of all searches in English.

One way it does this is by identifying words by category, e.g. proper nouns and conjunctions, as this can help determine a word’s role in a query. “It’s a more nuanced way of understanding many different words strung together in a sentence, which is how people communicate when they chat on a website,” Kyle says.

For example, if one user types the question, “There’s a problem with my pricing,” and another types, “Tell me about your pricing,” a basic type of NLU will recognize that they need information about pricing. Depending on its programming, the AI may proceed to send them an identical response. In contrast, Conversational AI that uses a system like BERT can understand the difference between these queries and respond to each person appropriately.

However, that doesn’t mean it’s thinking for itself. The bot is trained on relevant topics by a person. They’re the one who determines how it responds in different circumstances. What makes Conversational AI intelligent is not an ability to formulate answers to questions, but its ability to take unstructured data — like customer phone calls, search queries, or chats — and extract a nuanced meaning from that language. This enables it to then choose the most appropriate response from the presets given to it by human programmers. “That makes it seem like you’re talking to a human, but the AI is just understanding what people mean when they type characters on their keypad,” Kyle says.

Hopefully, this has laid out three foundational aspects of Conversational AI:

1. It’s a more sophisticated version of technology you’ve been using for decades.
2. It cannot operate effectively without human intervention and oversight.
3. It provides more advanced customer conversations than chatbots that rely on keywords or multiple choice.

Now you know the nuts and bolts of Conversational AI, you probably already have ideas for how it could apply to sales and marketing. In this section, we’ll dig into some of those use cases.

Engaging with New Customers

Most customers who come to your website have very similar questions. They want to find out what you’re offering and if it meets their needs. You want to know the same thing — and how serious they are about buying.

Having these very similar conversations over and over is not necessarily the best use of your sales development representatives’ (SDRs’) time. But doing something very similar over and over is one of the things AI excels at.

🤖 In particular, Conversational AI understands many different ways of saying the same thing, so it’s more effective at answering customer questions than a standard chatbot.

This helps you identify intent. Your potential buyer tells the chatbot in their own words exactly what they want as opposed to, say, clicking a pre-set button. You immediately know what products and services they’re looking for, and what part of the pipeline they’re in.

Based on this, you can qualify the lead. If it turns out the potential buyer isn’t a fit, you haven’t wasted time that the SDR could have been using on a better lead. If they are a good fit, you can tailor their content recommendations, communication options, and user journey to maximize the chances of conversion.

That user journey might not result in an immediate sale, in which case you can use Conversational AI to nurture leads until they are ready to buy. The bot can recommend relevant content depending on where your customer is in the pipeline, including videos and blog posts.

Letting the bots handle customers who are still on the fence gives SDRs more time to spend on hotter leads. And if the AI gets to a point in the chat where the lead is ready for the next step, it can alert an SDR or set up a future meeting.

Improved Customer Experience

In addition to helping you get insights into your customers, Conversational AI makes it much easier for your potential and existing customers to get the information they need about you. This is key when it comes to creating a positive experience.

“Buyers today do research online before we even talk to a sales rep,” says Aurelia Solomon, Senior Director of Product Marketing at Drift. “I get really annoyed when I’m just trying to understand if a product is relevant to me and the website keeps trying to connect me to a salesperson. That’s where AI is so powerful: It’s there to answer your questions.”

The advanced NLU can answer much more specific questions than a standard chatbot. Customers don’t have to settle for vague, general FAQs. They don’t have to rely on a chatbot that only answers pre-programmed questions using multiple choice — or completely misunderstands their query. In addition to having a more efficient and less frustrating experience, these customers identify your brand as an industry expert — in some cases, before they’ve even spoken to a salesperson.

The bots can also help you identify which topics your customers ask about most frequently and which you need more information on. During the course of their hundreds and thousands of conversations, the bots recognize topics that keep coming up and flag them to their human overlords. You, said human overlord, can then make sure you have enough information on that topic and train the bot to respond to questions about it. In this way, Conversational AI even makes it easier for companies in niche fields to answer highly specialized customer questions.

Speaking of specialized content, when you integrate your Conversational AI with your CRM, you can create a VIP experience for high-priority accounts. It could be something as simple as addressing them by name and talking to them as their appointed sales rep. If you know that they recently watched a webinar, or downloaded a white paper, you could ask if they enjoyed it. “You can create these really tailored experiences for select customers or prospects who you want to have a whole different experience,” Aurelia says.

Bots can also answer customer service questions. As with lead-qualifying, they’re able to take over the basic questions, which frees up your human agents to take on the thornier issues only a person can handle. Funneling customers through a conversational chatbot also means your sales reps don’t accidentally get stuck in a customer service chat that keeps them from selling. And all these customer service chats provide more opportunities to identify and flag common topics.

The New Voice of Your Brand

The benefits of Conversational AI extend beyond sales opportunities. Bots rely on humans to give them a voice, so you can make them an extension of your brand. “The nice part is you can add your own flavor to it,” Aurelia says. “For a lot of people, it’s their first interaction with the brand, so having that human touch is super important.”

Embrace the fun factor. Humans cannot resist playing around with bots — as you know if you’ve ever asked Alexa to tell you a joke or asked a chatbot its favorite color. “For example, if you ask the bot, ‘What’s the cost of a banana?’ it replies, ‘I’m sorry, we don’t sell bananas. Is there something else I can help you with?’” Aurelia says.

As with any type of branding, your bot is an opportunity to stand out. “B2B is so boring a lot of the time, so we aggressively try to make it fun,” says Rob Stevenson, Director of Performance Marketing at Zenefits, a Drift customer whose software automates payrolls, HR, and benefits.

Rob’s team even experimented with a bot that popped up and told visitors to go away. “Your immediate response is, ‘What is happening with this bot?’ — and you click on it,” he explains. “Then the bot’s like, ‘LOL Just kidding, tell me why you’re here,’ and it’s a happy unicorn GIF. Our engagement was up something like 400%.”

Collect and Analyze Sales and Marketing Data

On the backend, Conversational AI bots are busily collecting data on your customers, both individually and generally, to help you understand them better.

“The data and reporting piece of AI helps marketers understand and optimize their campaigns: what to say to who, what things are trending within the industry, how to tailor your follow-up,” says Monique Lemieux, Drift’s Head of Marketing Operations.

Having access to all this data can also help SDRs create a more personalized and efficient experience for individual customers. The reps can check on the bots’ conversations in real time or before a meeting. They can see what the customer asked about, what they’re interested in, any information they gave about their budget, and, of course, their name. This saves time for SDRs and customers, and creates a customized experience out of the gate.

This data can also help you design your website navigation to be more user-friendly for a general audience. For example, if the bot lets you know that multiple customers are asking questions about pricing tiers every day, you might consider moving that information to a more prominent position on your website, or labeling it more clearly.

Alternatively, if the bot tells you that a topic keeps coming up that you haven’t covered thoroughly, it’s a good indicator that the people coming to your website would click on a blog post, FAQ, or white paper about that subject. “They’re telling me in their own words what they care about — and what they want,” Aurelia says.

You’ve learned what Conversational AI can do, so let’s look closer at the potential business benefits of this technology. These can be divided into two broad themes: speeding up sales and scaling up operations.

Accelerate Your Sales Cycle

Chat can be a strong indication of intentionality. “Someone who chats with us is three times more likely to become a customer than a prospect who doesn’t chat with us,” Rob says. “Someone who chats on two separate occasions is 10 times more likely to become a customer. And if it’s three-plus chats, they are 30 times more likely.”

Before Conversational AI, your options for customer chats were limited to:

1. Chatbots that could talk to millions of people at once but only answer the most basic questions, often from a pre-set list.
2. Humans, who could answer more complicated queries, but only one or two at a time.

If you use basic chatbots, many of your customers ultimately have to wait to speak to a person anyway. And if you only use human agents, you’re limited by how many you can afford (and their working hours), which means long wait times for website visitors.

“It would take us a minute on average to get to people on chat,” Rob says. “One minute is an eternity when you’re waiting for a response.”

A bot that uses Conversational AI can:

1. Recognize and respond to a wider variety of questions.
2. Speak to millions of people at the same time.

This helps you accelerate several early stages of the sales pipeline, including qualifying leads, defining their needs, and setting up the initial meeting.

Conversational AI bots can direct every buyer who connects with them to the exact information they’re looking for more effectively than regular bots and more efficiently than people. This removes friction from the browsing and buying experience, which reduces bounce rates. And because the bot records the chat, you learn details about the lead’s needs, wants, and budget, in their own words.

Through this, both parties find out whether they’re a mutual good fit much faster than when talking to a basic chatbot or overwhelmed human. The Conversational AI bot can also recognize when a lead is hot enough that it needs to redirect them to an agent. And it can schedule a meeting between the lead and SDR in the future, cutting down on email back-and-forths.

“Best case scenario, we wouldn’t even need that first discovery phone call,” Rob says. “[The AI] would be able to chat, get their name, their phone number, their email address, their company size, the number of employees, and set a meeting, all without having to pick up the phone. Now I’ve saved time, I’ve saved money, I’ve saved resources.” With Conversational AI, your sales team can skip the interrogation and get straight to the good part (building a relationship!) so that your buyers can get the most value out of every conversation.

In short, Conversational AI helps you harness the power of chat to identify and nurture intent and convert those leads to customers.

Scale Up Your Sales and Marketing Efforts as a Small Team

Conversational AI bots may eventually replace traditional bots, but they won’t replace human agents. Instead, they’ll work alongside them, empowering small teams to scale their sales and marketing strategies as well as customer service.

For example, as a marketer, Monique points to the data insights bots can extract across thousands of conversations at the same time. “I think of Conversational AI as another tactic you can use to get smarter about the conversations you’re having, more personalized about the recommendations you make, in a way that doesn’t require you to do a ton of backend work on your team,” she says.

Having watched his sales team work alongside the Conversational AI bot, Rob says that people quickly get used to treating it as a tool and can see its benefits. “At first they’re like, ‘Oh, my gosh, the bot’s talking, I’ve got to jump in,’” he says. “But after a few weeks, they trust that the bot is going to get the prospect to the stage where it can fire it over to the sales team. You see them start to embrace what the bot can do for them.”

It helps that the bots are handling early-stage conversations, which take time but don’t necessarily lead to a sale. This allows the people on the team to focus their attention on the more in-depth work, putting together marketing strategies that come into play further down the funnel, or talking to leads who are closer to buying.

As the bot filters out less promising buyers, your team can be sure that they’re having more productive conversations, without needing to add more agents. “Our Conversational AI team sets demos at an 18% rate vs. 5% in any other channel,” Rob says. “They set as many meetings as the rest of the team put together in some months.”

🤖 Another thing that makes AI such a great teammate is that it doesn’t mind doing the late or early shift, and it doesn’t take holidays. It can meet leads whenever needed, and pass the information to a rep during their working hours.

“If you’re a C-suite executive on the hunt for a new solution for your team, you’re not online chatting in the middle of the day: You’re probably doing that at 10 p.m., after you’ve hung out with your kids, when you’re catching up on work,” Aurelia says. “And as a business, you definitely need to make sure you’re engaging people at that level.” Having a bot you can trust to help high-level prospects around the clock means you can reach more customers without hiring a night staff or outsourcing agents in a different timezone.

Congratulations: You are now enlightened on the capabilities and benefits of Conversational AI. We don’t have a sticker or a medal to give you, sorry. But hopefully this whirlwind education has shown you why Conversational AI has a place in your organization.

However, since you are clearly the inquisitive, thoughtful type, you probably have some lingering questions or concerns about this technology. Especially if you saw I, Robot, or heard what that Google engineer said about the company’s AI.

To address these, we’ve put together the top five most common myths people in sales and marketing — and more generally — believe about AI. No judgment: We’d want to know about the chess-playing robot too.

💬 AI Myth 1: “My business is too unique/too complicated for a chatbot to understand.”

We get it. Everyone thinks their business is far too unique and complex for a mere robot to come to grips with. However, AI doesn’t need to understand your business to answer questions about it because its responses are programmed by people.

“Theoretically, if you can train a human on it, you can train a machine on it,” Kyle says. The exception, he adds, would be if every single chat is completely different, since AI thrives on repetition. But the vast majority of businesses are not so complicated that an AI bot could not learn to answer the most common questions.

Consider also that AI can identify and flag niche topics your customers want to know about that you may not have considered. And AI’s consistency makes it appealing to highly regulated industries like law and medicine, which are required to give clients certain information that a human agent might forget.

💬 AI Myth 2: “The technical expertise you need to run an AI chatbot makes it too expensive.”

45% of marketers still consider themselves AI beginners…which is understandable. Without technical training, it’s natural to find AI daunting.

Fortunately, conversational AI bots have been designed to be user-friendly. You can train existing and new employees how to use it, no advanced degree required, and you don’t need to hire developers or engineers to maintain your bots. And it’s certainly cheaper to pay for one million chatbots than one million SDRs.

“Approachability is really important,” Monique says. “You don’t need extra resources: You can do it with the team you have because the product makes it really easy for you. We’ve done a lot of the work to pull out those insights.”

💬 AI Myth 3: “AI isn’t technically advanced enough to help customers.”

AI is often talked about in futuristic terms but it’s been helping consumers for decades.

Natural language processing (NLP), a type of language-analyzing AI which preceded NLU, was first developed after World War II. It’s navigated a few bumps in the road since then, but NLP has been mainstream since at least 2011, when Apple released AI-powered assistant Siri. NLU, meanwhile, is used in Google Search, translation apps, predictive text, phone trees, autocorrect, and autofill, to name just a few.

“There are tons of examples,” Kyle says. “Every time you use an automated phone system, that’s the same technology; it’s just applied to voice instead of text. You’ve been using those things for years.” The fact AI has blended into so many parts of our daily lives proves just how powerful and effective it can be.

💬 AI Myth 4: “AI can think for itself.”

It may depend on what you mean by “think.” AI can learn through machine learning. Deep learning takes that even further, using a complicated system of neural networks to collect feedback and develop. Conversational AI uses machine learning to get better at understanding user questions and finding relevant answers. But it doesn’t create new answers, or even understand what it’s saying.

“You have to tell it what to do, and what to say to customers,” Monique says. “It isn’t going to take over your website or change the conversations. It’s not going to override anything that you tell it to do.”

If you define “thinking” as making predictions and acting on them, “That kind of stuff does exist, but you and I don’t interact with that caliber of AI,” Kyle says. He gives the example of the AI built by developers DeepMind, which beat the world’s best Go player, Lee Sedol, in 2016. In 2018, DeepMind developed an AI that taught itself chess. After two hours of deep learning, it was able to beat human players. Two hours after that, it could beat the world’s best chess-playing computer.

An example of high-powered AI that combines natural language processing with deep learning is Google’s LaMDA. This came to public attention in June 2022, when engineer Blake Lemoine, who worked on the technology, claimed that the AI was aware of its own existence and experienced emotions.

While LaMDA’s conversational skills are very compelling, unfortunately for those who would like a robot friend to chat to, it is not sentient. It is just a very advanced version of the same technology we’ve been talking about. Which makes it very good at talking like a human, but doesn’t mean it understands what it’s saying.

Ultimately, this is all a bit like comparing your car to a rocket. These very advanced types of AI take a lot of processing power, and are currently too expensive and too complicated to be accessible to general consumers. And unless you want to lose a lot of chess, or get philosophical with a computer, they’re not necessarily even all that useful.

💬 AI Myth 5: “AI will replace humans on sales and marketing teams.”

Excluding world domination in the style of The Matrix, the general public’s number one concern regarding AI may well be that it will take over their jobs. But in sales and marketing at least, much of the work relies on developing meaningful human connections that AI can’t replicate.

“AI is never going to replace our SDR team,” Rob says. “It’s going to make them more productive and provide a superior customer or prospect experience. At some point, given the nature of sales, we need a real person. Drift’s Conversational AI frees up their time to focus on selling, and to make sure they’re offering the best possible experience.”

Companies are right to be excited about AI. It can be game-changing when used in a way that plays to its strengths. And it’s these strengths that make Conversational AI an excellent tool for sales and marketing reps without making it their replacement.

Conversational AI chatbots allow companies to dramatically increase how many leads and customers they can serve at one time and accelerate early-stage sales conversations. This empowers the sales reps to focus on hotter leads. The bots also collect data, which helps marketing identify key trends and topics as well as gaps in their offerings.

We’ve long relied on science fiction to show us what AI will look like. The reality might be less thrilling — but it’s also a lot more hopeful.

If you’re looking to learn more about Conversational AI, and how to get started tactically, check out The Drift Conversational AI Starter Guide.