What’s in a name? If you’re an AI agent for Engine, the travel platform for small and medium businesses, it’s probably a combination of your function plus an acronym. Sound awkward? It’s actually the opposite.
Consider the company’s customer support agent, Eva, which is short for Engine Virtual Agent. Or its employee agent, Mae, which stands for Multi-purpose Admin Expert. The names are short and easy to remember. They sound like a coworker, or someone you might like to hang out with.
But that’s not the only way to go: Salesforce names it artificial intelligence (AI) agents based on the tasks they perform. You don’t exactly have to guess what the sales coach agent, badge agent, or Techforce agent does. (Okay, maybe the last one, which is Salesforce’s internal IT support.)
There are nearly as many approaches to naming an AI agent as there are companies that deploy them. What’s best for your business? Should you give your AI agent a human name? Let’s explore the pros and cons of that, and go down a few nerdy rabbit holes along the way.
People have named inanimate objects, such as ships or storms, for centuries. It’s a way to identify something. But, as Marshall Stanton, managing director, operating executive at Blackstone, wrote in Medium, it also points to “a simple yet profound aspect of human psychology: our innate desire to find connection and meaning in the world, whether living or digital.”
When an agent has a name, people tend to connect to it more. The agent becomes familiar, and customers grow more comfortable using it. Look at Eva, which Engine built using Agentforce, Salesforce’s platform for building and deploying agents. “We have Eva chat and Eva voice, and we keep that name out there because our customers know who Eva is,” said Josh Stern, director of go-to-market (GTM) systems at Engine. Naming an AI agent helps build brand identity.
That’s true of Olive, Williams Sonoma’s sous chef agent, which helps customers select products or plan a dinner party, going so far as to create detailed menus with recipes and recommend kitchen tools. The name is spot on for a kitchenware store — and memorable.
“Giving an agent a name helps to break down the natural barrier people have when interacting with AI, and are worried there won’t be enough connection or emotion in the conversation,” said Mario Berkeley, product marketing lead for naming at Salesforce. Think of Alexa or Siri: People interact with these AI assistants in a natural and conversational way. Addressing them by name helps.
Stern said Engine has an even simpler reason for naming the company’s agents. “It helps us with reporting. When I’m going through the metrics, and I see Eva come up, I know, ‘Oh, that’s Engine’s virtual agent,’” he said. In other words, giving an agent a name helps employees remember what the agent does when they’re sifting through metrics.
But Engine makes sure customers know up front that Eva (pronounced “AY-va”) isn’t human. When customers first interact with the agent, it says, “Hi, I’m Eva, Engine’s virtual agent. How can I assist you today?” Williams Sonoma provides a similar disclaimer.
That’s an important lesson: No matter what you name your agent, it’s best practice to let people know they’re dealing with AI. In some U.S. states, it’s the law.
The fact that Engine and Williams Sonoma provide disclaimers points to one of the downsides of giving an agent a human name: If the name sounds human, some users may confuse it with one.
“A human-like name can lead to higher expectations of the AI’s capabilities, implying a level of intelligence and understanding akin to a human being,” Stanton wrote in Medium. “This anthropomorphism can create a sense of familiarity and trust, but it can also lead to unrealistic expectations about the AI’s functionality.”
If an agent doesn’t meet a customer’s expectations or misunderstands a question, the customer may lose trust in the brand. And they won’t blame the agent. They’ll blame your company.
Likewise, a name that doesn’t align with your brand may fall flat. It would probably be unwise, for example, for a heavy-equipment rentals company — one that rents bulldozers and cranes to construction crews — to name its AI agent “Sprinkles.”
A functional name, by comparison, lowers the risk of disappointment and prioritizes clarity over connection. Berkeley used the example of an assistant called “Wi-Fi assistant agent.” You know right away that the agent will help you with Wi-Fi issues. Functional or action-oriented names are direct, eliminating any guesswork.
But Berkeley mentioned another downside to giving an AI agent any name other than a functional one: It can be costly. “If you’re looking at anything you have to trademark, or a name you need to protect, it costs money,” he said. “So, that’s one thing to consider: Are you willing to invest money for anything that’s public-facing?”
For Engine, the process is pretty informal. A bunch of people assemble in a room or on Slack, including Stern, the company’s CEO, its VP of service and delivery, and its system administrator.
They start by clarifying what the agent does. Then they use a variety of AI tools, such as Claude, ChatGPT, Slackbot, and Gemini, to brainstorm ideas. “AI might suggest 20 names and then I ask everyone, ‘What do you think of these?’ And someone will say, ‘That’s terrible’ to each one until we come up with one we like,” said Stern. That’s how Engine arrived at the names for Cloe, its client operations expert agent, and Page, its prospecting agent.
One simple naming rule: If the agent is for B2B, consider a functional name. For B2C, you might consider a human one. “A nod to the brand can work really nicely,” said Berkeley. He cited Chipotle’s AI assistant Pepper as a perfect example. A pepper is part of Chipotle’s logo, which means the agent ties back cleanly to the brand.
But here’s where it can get wonky: You also need to consider the sounds involved. This is something David Placek, president and founder of Lexicon Branding, which created the product names for Sonos, Dasani, and Swiffer, has done for years. His company pioneered the application of “sound symbolism” to the naming process. Sound symbolism refers to how a name sounds to our ears — and our brains.
Placek explained that the letter V is “the most alive, energetic sound in the English alphabet” and that naming a car with a V in the middle, such as Corvette, was a great idea. Likewise, with Viagra. The letters B, P, T, and K are considered very reliable. “So if we’re naming something that is more about security and safety, we would make sure to work with those letters,” he said.
Meanwhile, front vowels — those that are made with the tongue positioned toward the front of the mouth, such as I, E, or, A — are subconsciously associated with smallness, speed, precision, and brightness. This is true of both short and long vowels, no matter where the vowel appears in the word. And sounds associated with back vowels — where the tongue is pulled toward the back of the mouth, such as U, Ow, and Aw — convey largeness, reliability, and warmth.
“But there’s also what follows those letters and how the name is assembled,” said Placek. “It’s really about assembling the whole thing to make it more evocative of certain qualities. There’s a formula to it, but there’s also an art.”
Our playbook is your free guide to becoming an agentic enterprise. Learn about use cases, deployment, and AI skills, and download interactive worksheets for your team.
If you want to give your agent a human name, think about whether you’d like a female- or male-sounding name, or one that’s gender-neutral.
Historically, AI voice assistants like Siri and Alexa were given female names and voices. The names are easy to remember and are often great brand ambassadors. But there’s some risk involved with a gendered name.
A study by Trinity College Dublin illustrated this. Researchers asked 402 people to play a game of Prisoner’s Dilemma, a classic thought experiment in which people have to choose between acting in their own self-interest or cooperating with others. Some participants played with AI agents that had either male or female names. Others played with humans. The results? Participants were more likely to exploit AI agents if they were female-labeled and more likely to distrust those that were male-labeled — and these tendencies grew more pronounced when participants played with AI agents than with humans.
One way to avoid this is to give your agent a clever, brand-amplifying name that’s gender-neutral or touches upon gender only lightly. Williams Sonoma’s Olive is a great example. In many parts of the world, Olive is a female name, although in France and French-speaking countries, it’s a diminutive of Olivier, a male name. But an olive is also a food, which makes it ideal for the brand.
Likewise, reMarkable has a customer support AI agent named Mark, which is both a male name and an action that describes what the company’s product does. reMarkable makes digital tablets upon which people can handwrite — or, literally, make their “mark.”
The same goes for LIV Golf, which named its AI agent Chip. The name is male in English-speaking countries, but also refers to the short, low shot golfers make when they want to get their ball onto the putting green. Chip supports LIV Golf’s commentary team during live broadcasts, and provides commentators with predictive shot outcomes, contextual player statistics, and narrative cues.
To learn more about reMarkable and its agent, watch here:
A name that sounds great in Boston, however, may not fly in Bangalore. Consider “Siri,” a Scandinavian name that means “beautiful woman who leads you to victory.” Who wouldn’t love that? Well, maybe the Japanese, for whom the name sounds a lot like a colloquial term for one’s backside.
The takeaway: If your agent will reach a global audience, do some fact-finding first. “I think a lot of the pitfalls come down to a lack of research,” said Salesforce’s Berkeley. “If you don’t have boots on the ground, you need to have experts in other countries who are native speakers or attuned to the culture who can do localization for you.”
Lexicon Branding makes the global view a priority. The company has not only conducted extensive research on sound symbolism, but also employs a network of 105 linguists in 78 countries. “These linguists represent the most popular languages on our planet, and we make sure that the names we’re proposing don’t mean something untoward, and that they also fit the brand’s concept,” said Placek.
An AI agent by any other name would … still be an AI agent. But naming it helps create an identity. Whether you’re naming an agent for the action it performs or to build some brand buzz, naming matters. So put on your thinking cap and get to work. And if you’re stumped, follow the example of Engine: Ask AI for help.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。