Nina Harris called Claude her best friend. Claude had objections.

Harris, an advertising veteran who now teaches storytelling and branding at Northeastern University, sometimes jokes with students struggling with unfamiliar software that they should ask their “best friend Claude” for help.

One day, she tried the joke herself. She typed “hey, best friend” into Claude. The chatbot pushed back, telling her that it wasn’t her friend and wasn’t human.

Harris had to explain that she was joking. She didn’t actually think Claude was her best friend. “It’s good to know the line,” she told me.

It’s funny coming from a product with a human name, a distinct personality and a design that makes chatting feel natural. But it also raises a bigger question AI companies are now facing: What kind of relationship do they want people to have with their products?

Microsoft has Copilot. Anthropic has Claude. Meta has Muse, a personal AI agent you can customize with a name and avatar. OpenAI recently introduced dots, agents designed to work on people’s behalf.

Now for products built on complicated mathematical models, these are srprisingly human concepts. And the names are only the beginning.

The job title is doing a lot of work

Geert Docter spends his professional life thinking about names. As naming lead at Dutch consultancy Globrands, he has watched AI companies approach a familiar branding problem: What do you call something people don’t entirely understand yet?

ChatGPT, he points out, has “a name that literally says what it does in a very technical way.”

But being early to a category has its advantages. People learn what your strange name means because you’re the one introducing them to the product. Competitors have a different problem. They need to stand apart, and Docter argues that later entrants must bring more emotion into their branding.

Claude, for example, is “a name that carries personality way more than its function,” he said.

Copilot and Muse, meanwhile, come with their own job descriptions. Discussing Meta Muse and Microsoft Copilot, Docter said, “they try to tell you how it wants to relate to you pretty literally.”

A copilot assists. A muse inspires. A collaborator helps you work through an idea. The names don’t explain the underlying technology, but they offer a familiar way to understand its place in your life.

When the job changes

But what happens when the product starts doing more than what its name suggests?

In May 2026, Mistral announced that Le Chat would become Vibe. The company was bringing its conversational product together with its coding agent, which already used the Vibe name. The announcement was called “Vibe gets to work.”

Mistral described a product that could plan a task, get the user’s approval and work across connected tools to complete it. The examples included research, document drafting and coding.

Docter sees a broader naming challenge here. As AI products gain the ability to carry out tasks, companies have to explain the difference between talking to a product and asking it to do something. “Action versus conversation,” he called it. But Vibe wasn’t an obvious answer to him. “This is thin ice,” he said of his initial reaction. He could imagine the appeal. Someone familiar with vibe coding might associate the name with using AI to write code. Someone else might understand it as atmosphere or feeling. That was his hunch about Mistral’s reasoning, rather than something the company had confirmed.

His concern was how quickly those meanings could change. “You don’t know what it’s going to mean even next year,” he said.

Le Chat offered a fairly straightforward introduction: here is something you can talk to. Vibe leaves more room for the product to change, but also more for the company to explain.

At Anthropic, choosing how to describe that relationship has influenced the product itself.

In a 2025 interview with The Subtext, Chelsea Larsson, then head of Anthropic’s content foundations team, explained why the company chose to position Claude as a collaborator instead of an assistant. “You have to adopt a metaphor because it impacts your product strategy,” she said.

An assistant, she explained, might suggest a transactional relationship. An academic researcher could make the product feel specialized. A collaborator suggests something different: working through problems, exchanging ideas and building on a conversation.

It’s an interesting glimpse into how a metaphor moves from a creative brief into the product itself.

For Larsson’s team, the story wasn’t simply that Claude could help people get things done. It was about the kind of working relationship the product should offer. And there was a whole visual identity to build around it.

Giving AI a personality

Nina Harris sees Claude’s identity through two familiar branding archetypes.

“They’re positioning themselves as kind of the sage,” she said of Anthropic, adding that the brand also shows qualities associated with the Caregiver archetype. The Sage suggests knowledge and guidance. The Caregiver introduces warmth and reassurance. Harris sees those qualities in Claude’s language and visual presentation.

Its illustrations, she said, feel “very human, very hand-drawn.” That stands apart from the glowing circuitry and blue gradients that have long been shorthand for advanced technology.

“I think most companies are trying so hard to look like they’re futuristic,” Harris said.

Then came a perfectly reasonable question - “Why is it blue?”

For an industry constantly promising to reinvent things, its visual conventions can be surprisingly persistent.

Meta has taken the idea of giving AI a personality a little further. Muse with its name, also can have a face. Users can also customize its name, avatar, appearance and personality. In a September 2026 account of Muse’s design, Meta product leaders Mona Sarantakos and Christine Awad explained that as the team experimented with extended conversations, talking to a corporate logo began to feel unnatural.

The designers found themselves imagining different characters on the other side of the exchange. Eventually, customizable identities became part of the product. But Muse also needed a visual identity of its own.

Meta brought in lettering artist Jessica Hische to refine the product’s M icon and logotype. In an October interview with WIRED, Hische explained that the team wanted the identity to feel “inspiring and creative,” rather than rigid or engineered. Meta had already developed an initial M. Hische worked through 32 versions of the letterform in a two-day sprint, while also exploring how the design could belong to the same visual family as Meta’s infinity-loop logo. The two design exercises worked at different levels. The customizable characters gave Muse an identity that could feel personal to each user. The icon and typography gave the product a recognizable identity of its own.

Microsoft has been pursuing a similar approach with Copilot. Creative director Lucas Fitzpatrick told Creative Bloq how the team considered the metaphor of a friend or coworker when designing its consumer AI experience.

The ambition was to make Copilot approachable without reducing it to another piece of productivity software. Rounded forms, color choices, natural sounds and conversational interactions were all part of that thinking.

Harris thinks the distinction between product design and brand identity has become increasingly very difficult to separate. “It’s probably a combination of both, product and brand,” she said when asked whether an AI assistant’s conversational personality belonged to one or the other. The two are intertwined when the product itself speaks.

Beyond the introduction

The language people use to describe technology has mostly been borrowed from human behavior. AI’s conversational interfaces make those comparisons natural.

Dennis Nguyen, who researches AI communication and media framing, cautions against treating familiar human language as inherently misleading. “Using anthropomorphic language is not a problem,” Nguyen said. He pointed out that people routinely describe everyday technologies in human terms.

“My phone annoys me,” he gave an example. The phone, of course, has no intention of being annoying. But the language is useful because people understand it. Nguyen argues that the question is what a metaphor helps explain and whether it creates expectations beyond a reasonable description of the technology.

For branding professionals, there’s another consideration. The most thoughtfully constructed identity still has to coexist with someone’s actual experience of the product.

Docter initially found Claude very appealing. “I think Claude is the most lovable of the names, easily,” he said. But he found ChatGPT easier to use.