What It Takes to Build Brand Personality Into AI Agents

It’s no secret that as AI becomes accessible, its presence across the industry will continue trending upward. As it stands, creatives, marketers and everyone in between are already leaning into its capabilities en masse – a permanent change to advertising as we know it.

However, what has stayed consistent is the fact that not all use cases are created equal. As with everything, some people are seeing more success than others, especially when it comes to using AI-powered brand agents to transform brand worlds into conversations. Warmth, humour, consistency, and even edge are all characteristics we’ve come to expect in our interactions, but implementing these features into a chatbot, virtual assistant or voice interface? That’s a whole different ballgame where, according to some of the industry’s sharpest minds, the rules of engagement may require careful consideration.

Meet the New Boss

When it comes to adventuring into what Ryan Lynch, co-chief executive officer of Beardwood&Co, calls the “branded AI agentic future”, the first thing today’s marketers and brand owners must consider is looking beyond the allure of branded agents and avatars.

“It all starts by building a compelling brand world with a clearly established and consistent identity, and a powerful tone of voice to match it,” he explains. “Imagine an emotionally intelligent AI capable of creating stories customised to consumers' real-time behaviours and moods. Beyond branded content, a food brand avatar could make recommendations on travel, dining, and culinary exploration in every major city you visit. Or, what about an oral care brand avatar generating episodic romance novels about winning smiles? These agents work cross-platform, create personalisation, and build awareness.”

Of course, getting there is far easier said than done. While digging into the distinctly human traits of a brand – and what makes them relatable – is a good start, and not a concept creatives and strategists are unfamiliar with, to fully unlock Ryan’s vision, focus with a limited scope must be acknowledged as a red herring.

Specifically, for Instrument’s executive strategy director, Avi Couillard, it’s important that people working in this space recognise writing for but one moment at a time as a pitfall best avoided. Instead, refining a broader approach may just be the key to seeing success in the long run.

“It’s about building a more holistic and systematic approach that helps your brand show up consistently, no matter the context,” he says. “We’re moving toward a world where more customer interactions are AI-driven, but your customers don’t care who made it, just that it’s great and aligned with their expectations. That’s why it’s critical to set clear guidelines for how your brand should show up across different touchpoints, whether you’re launching a new product, answering a support question, or engaging on social.”

Yes, distinct moments require distinct voices. But, as Avi emphasises, the work still needs to have a familiar and consistent voice. It’s for this reason that before anything fancy happens, a strong, flexible system must be developed – one capable of ensuring the AI receives the right inputs in the first place, and alerts the team if something is off.

Same as the Old Boss

Leaning into the tried and true strength of consistent brand voice isn’t the only suggestion that adland may find familiar when delving into the realm of AI application. Another crucial point that cannot be overlooked is the fact that, ultimately, all this work is done in the interest of effective marketing, which means consumers will still want to feel valued within the process.

To this end, Brandon Murphy, head of strategy and innovation at Trade School, wants the industry to remember that establishing a personality that aligns with the brand and its customers alike is the key to cracking the code.

“First and foremost, know your audience,” he says. “This may seem overly simplified, but before you establish a brand personality for AI, you must understand who the agent is speaking to. The AI is centred around the consumer, and the consumer has to feel connected to the brand identity.”

Equally so, the head of strategy believes that Avi’s point about consistent narratives is spot on, adding that an extra benefit to this methodology is the fact that it helps establish credibility. Citing that brands are still expected to act like people, AI or not, he emphasises that those who are inconsistent are on the fastest track to distrust and losing customer loyalty.

So, what does this look like when done well? For Brandon, two of the most reliable means for success are ensuring brand-led AI accumulates experience with users, and making sure that its voice feels genuinely human – not just ‘how we sound’.

“Remembering the users’ tone, past behaviours, and preferred style of interaction are all beneficial in making a customer feel appreciated and connected,” he continues, explaining the former. “Meanwhile, responsiveness to complex questions and tones of voice give the illusion of personality – of a back and forth. The AI should behave like a living entity speaking to another.”

This is something which Ryan is quick to echo. Observing that agencies should already be shaping their work around what’s distinctly human about a brand, and how people relate to it, he adds that the same should be applied here – it just takes time to do so.

“To train an AI to perform consistently, you need more data for it to perform than you would for a human copywriter. The example messages, dos and don'ts, and the language principles the brand employs will shape each and every response. Nuanced outputs require nuanced inputs. Without a human interface, you have to build humanity into the interface for it to succeed.”

Teaching Human Behaviour, or Should We?

While designing brand personalities capable of speaking and responding with all the warmth and wit of a real person may sound like a daunting task, several different agencies have found effective entry points, with a – perhaps – surprising number of approaches proving viable.

The first, and perhaps most obvious, as Laurence Cornwall-Watkins, managing director at Bright Blue Day, notes, is just taking the time to make sure the kitchen isn’t filled with too many cooks, or too many dishes.

“Don’t throw your entire content archive at your model and hope for coherence,” he says. “Instead, identify one to four people who embody your brand – the founder, the CMO, or the intern who just gets it – and then train the AI through them. Let it learn their opinions, humour, rhythms, and language. Personality is a human trait, so to build it into AI, you need a human to lead it.”

Pointing out that brand guidelines alone aren’t enough to build an AI agent’s personality, Laurence questions why people who obsess over the tone-of-voice decks and purpose statements are baffled when their AI turns out like a “dull chatbot in a logo hoodie.”

“You can’t prompt your way to brand personality,” he continues. “If we want AI agents that sound like us, we need to stop feeding them lifeless documents, and start training them with living brand behaviour. When you do this right, the result isn’t just more authentic; it’s more scalable. You get brand consistency, content velocity, and a voice that doesn’t just respond, but connects. That’s not a prompt. That’s a presence.”

Meanwhile, according to Left Field Labs executive creative director Rich Foster, one of the pivotal realisations at his agency was that their precise focus shouldn’t just be on tone, but on conversations more holistically, including the likes of flow, timing, visual alignment, and intent.

“A strong conversational identity doesn’t stop at what’s said,” he explains. “It needs to look like the brand too. For those like Nike or Prada, whose visual presence is instantly recognisable, that signature style must extend into AI-driven interactions. If the tone is on-brand, but the design is generic, the experience falls flat.”

This is where, as the ECD calls it, “dynamic art direction” comes into play. Through the use of modular, flexible systems (visual components, typographic rules, colour, motion, etc.), the agency’s creative teams are able to build conversations that assemble in real time – living design languages that flex with context, rather than static templates.

“Define your AI like you’d cast a role,” Rich suggests, breaking down the process. “Is it playful? Wise? A hypebeast? A skincare bestie? Character drives consistency in tone, phrasing, and intent.”

Another valuable consideration is designing for tone shifts, as it turns out. Notably, by building a tonal spectrum, and opting to not lock into a singular voice, these AI tools gain the capability to respond on a need by need basis, such as with clarity and calm for support requests, or casually for discovery requests.

“Guide the model, and don’t just cage it,” Rich says. “Use smart prompts and rules to stay on-brand, but give space for surprise and spontaneity. That, and know when to end the conversation. When you structure interactions with a sense of closure, it helps users feel confident, and not caught in an infinite loop. The best conversations know when they’re done.”

Meanwhile, under Michelle Higa Fox, group creative director at BUCK, the team has found a few additional paths to success. Although the GCD affirms that consistency should be the core of any AI brand assistant – something effectively demonstrated by Notion AI’s assistant – part of the secret to designing a thoughtfully articulated interface lies in understanding the type of tool you’re designing, and consistently thinking about dialogue the same way one might think about briefing an actor or an artist.

Specifically, the team has found that elements like a character’s personality traits, backstory, motivation, and quirks – and communicating these – can be critical to the success of developing the behavioural context and system roles of chatbots, virtual assistants and voice interfaces.

However, when writing for AI, Michelle clarifies, the process is slightly different. “In addition to describing personality traits or tone-of-voice, writing for AI often involves thinking about dialogue in a way that is different from traditional narrative,” she elaborates. “A large sampling of speech by a character or brand is invaluable, and is always the base of the personality, be it from narrative content or a backlog of social media posts. However, a chatbot or virtual assistant is almost always in dialogue with a user. This means we spend significant time anticipating a range of expected and unexpected questions and using those insights to iteratively refine the chatbot’s personality and responses.”

To this end, when the team at BUCK is writing for AI, they’re encouraged to creatively direct by identifying clear North Stars and inspiration, and allowing artists to ideate in the space of possibility. But, when working within the framework of an open-ended system, it becomes about establishing clear sandbox walls and defining boundaries, rather than setting up some broad creative aspirations.

“Crafting these walls requires nuanced skill and understanding of what the assistant should and shouldn't say, and how it should respond across countless contexts,” Michelle specifies.

With all that said, not everyone believes that introducing humanity to such technological endeavours should be of primary importance. In fact, as Momentum Worldwide’s chief creative officer of North America, James Robinson, puts it, “humanity might be overrated.”

“Humour, charm, and warmth are becoming table stakes in AI, often easily implemented just by tech alone,” he says. “Check out the AI podcasts put out by Eleven Labs or Speechify. They could not be more charming. They’d fool anyone into thinking they are listening to two live people kidding around, and maybe that is the problem. Maybe what customers want from AI isn’t humanity. Maybe what they want is honesty.”

Tapping into his agency’s internal research, the CCO explains that, in fact, customers don’t seem to mind talking to an AI, provided they know it’s one. But, if that AI tries to fool them into thinking it’s something it’s not… that’s a huge, huge problem.

This need for radical transparency, in fact, was the guiding principle when Momentum built its proprietary AI. While including an engaging personality was still something the team opted to do, more important was that every answer, no matter how charmingly delivered, was footnoted and annotated to clearly lead back to brand materials and tested sources.

“When you get down to it, really, what most people are looking for is an AI they can trust,” James says. “And, we find that after a couple of checks, the customer never checks again, because we have built trust. Then, just for good measure, our whole experience is built with AI oversight to make sure there are no hallucinations or fabrications. Honest AI answers beat a charming AI personality any day.”

The Right Approach for Your Business Is…

Whether you’re for or against the prioritisation of humanity in AI agents, at the end of the day, the safe takeaway seems to be that good, functional design trumps everything else. In a world where people have distinct preferences for what they want out of interactions with such creations, there’ll always be those that favour certain models over others. So, the most important thing to get right? Making sure they can actually capture a brand’s behaviour effectively.

This is what John Jamison, director of engineering and AI at H/L advocates for. While the agency builds its agents around three core pillars – voice (how they sound), values (what they stand for), and volition (how they make decisions) – none of those elements end up mattering if these creations aren’t able to respond to the unpredictability and messiness of day-to-day life.

“We think hard about how our agents handle the weird stuff: ambiguity, surprise, and the unpredictability of human interaction itself,” he notes. “People are emotional, irrational, and wonderfully inconsistent. Our agents have to be ready to navigate that – not just the neat edges of a controlled system. Since we work with some of the most powerful foundation models out there, the craft isn’t about retraining them from scratch. It’s about building the scaffolding around them with smart prompts, system instructions, guardrails, and interaction patterns.”

With this in mind, the director would even go so far as to say that the ‘I don’t know’ moments agents face actually matter more than any of the perfect answers they might provide. Something he considers to be often overlooked by others in this line of work, he reiterates James’ point – the best brand agents aren’t trying to impersonate people – but adds these are the moments when humour, grace, or even a little edge may be required.

“[The best agents] are carrying the brand’s gut and soul into conversation, flexing when they need to, standing firm when it matters, and leaving just enough imperfection in the mix to feel alive,” John sums up.

That need to cut through is echoed by OLIVER’s Tom Claridge, who serves as head of strategy for its U-Studio. Citing the present as the “age of digital landfill”, he criticises poor use of gen AI which floods the market with results that fail to deliver nuance, humanity, or the ability to actually get a brand recognised, before echoing the recurring point – capturing brand voice and behaviour is essential to cutting through.

“Without guidance, gen AI-authored content can brilliantly deliver surface benefits, but often misses deeper customer motivations,” he continues. “Meanwhile, effective brand voices connect features and attributes to emotional outcomes that reflect how people experience the needs your brand serves.”

To this end, Tom wraps up by offering several warnings about the limitations of poorly briefed and prompted content, which he hopes people will avoid in their pursuit of strong future AI offerings.

First up is realising that while algorithms can approximate, it’s rare that they actually capture the distinctive personality that makes your brand recognisable. “Generic language lacks the vocabulary choices and expressions unique to your brand identity,” he explains. “Those formerly ignorable, baseline, forgettable tone-of-voice documents are now becoming like bad code for your brand… repeated ad infinitum. Fix it or be ignored. If it reads like it could come from any of your competitors, is it helping your brand?”

Following this, the head of strategy points out that vanilla LLMs struggle to genuinely evoke emotions, even if they understand which ones to target. Combined with gen AI’s tendency to produce predictable patterns based on common language constructions, and this is exactly what, as he pointed out, is presently cluttering the space.

“We need to break these patterns to stand out,” he concludes. “True brand distinctiveness comes from conceptual collisions and pattern interruptions that grab attention. For example, if 10 people prompt for 10 headlines to sell an EV, how many of these headlines will be the same? Finding ways to be original, to do something differently, will always help you stand out in a sea of sameness.”