Using LLMs to create a collaborative journey vision

Vibe coding – the practice of using conversational prompts in LLMs to build new tech – allows organizations to build the specific tools they need, without the in-depth technical expertise that would previously see such projects handled exclusively by coding and development teams. 

There are many reasons organizations – and individuals – take this approach. The freedom to build to requirements is just one. Others vibe code what they need because of the precision custom creation offers, or the appeal of easier maintenance. 

For Olivia Lucas, who currently holds the position of senior service designer at EA, the reason was different. She needed to create an in-depth experience vision for an online customer journey. The vision would be used to communicate a new account experience for current customers.

But instead of turning to the traditional CX tools to create a journey map – or dreaded slide deck – she used Claude and ChatGPT to vibe code an interactive website that could present various parts of the new ideas to the stakeholders who needed to see them. 

In this interview with CX Network, Lucas explains why and how she used LLMs, how to prompt and collaborate with LLMs, how vibe coding is already reshaping her role, and how LLMs could shape the future of journey management. 

Don't miss any news, updates or insider tips from CX Network by getting them delivered to your inbox. Sign up to our newsletter and join our community of experts. 

Learn More

Communicating a vision to multiple stakeholders

In a digital and connected environment, no journey exists in isolation. When Lucas set out to visualize a new account experience the goal was to accommodate the needs and ideas of different stakeholders in a way that would reduce duplicated work and efficiently communicate the relevant parts of the experience to the stakeholders who needed to sign off on them. 

She explains: "In my role as service designer, we focus on creating a cohesive strategy across lots of touch points and experiences, then trying to make the journey not just touch-point specific – for example they go to the homepage or they go to a specific website – but cohesive."

"A lot happens before and during a journey, users don't go straight into the game. It's about bringing it all together and knowing how the website fits in with what the user is experiencing in the whole scheme of things," she adds. 

In large organizations, different stakeholders – including some who may not be familiar with how journey design tools work – are involved with different parts of an account experience.

Each has their own objectives regarding how their piece should evolve and what that should look like. Allowing stakeholders to build out visions individually would duplicate work, but an effective and centralized experience vision drives alignment across the different groups. Holding a role that sits in the middle of these stakeholders, Lucas had the unique perspective of what each needs to see. 

"Instead of people saying their piece and pushing their perspectives in meetings, the experience vision brings more of a visual, a story, something to react to, and it's a good place to capture all the ideas," she explains. 

No two experience visions are the same, but at the core a strong vision will capture current challenges, explain why those challenges need tackling, and present an idea of what the future looks like. While different professionals have different methods, Lucas combines a storyboard narrative with a journey or blueprint using tools such as Miro and Figma. 

On this occasion, however, the density of the content and the need for different stakeholders to see different parts of the vision pushed her to try something different. 

"You're not going to go share the same details with the C-suite that you do a product partner," Lucas says. "With this experience vision I was trying to solve all of these usual hurdles in one go with an LLM," she adds. 

Selecting the right LLM for the job

With all the information to hand and access to both ChatGPT and Claude, Lucas started to experiment. She first turned to Claude, reporting the result was "a very cool experience". 

Lucas deliberately chose not to over prompt the tool to see what ideas it would add to the mix. After sharing minimal content and direction she says the initial result was "like a homepage that then took you into either the experience vision or more of a journey blueprint type detail of the current experience". 

It then connected all the concepts, such as all potential products, services and experiences that could be added. "Everything that comes from the story that's going to get us to that better future," she says. 

"Pulling out and then expanding those elements went into more detail, that we usually share with the product partners. For example, here are the behaviors we're addressing and what it's really going to do if we do this and how we can do this thing well. It did a really good job of helping me get to an outline," she adds. 

One of the benefits of Claude is the volume of information it can handle. Lucas had 16 concepts, each comprising one to two pages of information. While the ideas were formulating for years, once the project started the process of creating the experience vision using LLMs took less than three weeks. 

"I knew I wanted the vision to be grounded in the specific phases of the journey; what are users doing before they create an account, during account creation and then after. I gave the LLM an idea of what I'm thinking, then it mostly wrote the story," she shares. 

It's important to note this wasn't a case of information in experience vision out. Each output required editing and refinement, proving the point that AI – and LLMs in particular – are no replacement for human creativity and intelligence. They can, however, be incredibly powerful assistants when applied in the right way to the right scenarios.  

The secrets of a successful LLM prompt

Part of harnessing the full potential of LLMs lies in knowing how to prompt them. As AI becomes more embedded in workflows, AI literacy will likely demand more CX practitioners become proficient in this skill. 

There's no single formula for a great prompt. Generally speaking, users should explain the role they want the LLM to take, give context, state clear instructions, specify the structure and provide some examples. "There's definitely a science to it," Lucas says. 

In her experience, success didn't just come from the brief shared with the LLMs, but the settings applied to each tool and knowing the limitations and benefits of both ChatGPT and Claude. 

"Prompting is like another form of storytelling. The question is how do I tell the LLM enough to get the task done and give it the right context to do that task," she says. 

"I found what worked decently well is approaching it in a storyboard format and telling the LLM these are the different pieces of the story and this is how I see them fitting together. That helped a lot. The clearer I am with what I want in the beginning, the easier it is to get the output," she says. 

These tips are applicable to all LLM use cases, not just vibe coding. Often, Lucas wants the LLM to collaborate with her to ensure her ideas were the best they could be, which changes how and what she prompts.  

She says: "Sometimes I try to give the LLM space to create something for us. When I do that it's interesting for me to be challenged as to whether my idea is the right idea. Is there something out there that can inspire me? That process gives me that initial step.

"I still look to places like Dribbble, or even just like Google or Pinterest for inspiration, but the LLM has become another tool where I can say, 'how would you do this with this?' In general, the more specific you can be about what you want, the better it is," she adds.

Specifics don't always demand words. As an experience design specialist, sometimes Lucas prompts in pictures, sharing visuals and diagrams with the LLM as well as giving it specific written requirements. 

"In cases where I know exactly what I want the LLM to do and I don't want it to give me that additional inspiration, I might specify that I want a storyboard with six frames and I may even draw a diagram as part of the prompt. Any inspiration you can share with the LLM will improve the output. You can even share visuals you find online," she explains.

As LLMs continue to develop, so too does the art of prompting them. Lucas says only six months ago, some LLMs would struggle to produce a slide with a workable layout. "It wasn't even worth the prompting. But now I feel like it does a lot better with visuals," she says.

Predictive capabilities are another work in progress. Lucas shared multiple documents with Claude to create her experience vision. With current capabilities, Lucas says her biggest tip for practitioners is to "intimately know the information you're putting in". 

Until LLMs become better at predicting user needs, she says that combining the knowledge of what goes in with a strong idea of what you want from the model will create better outputs. 

One of the biggest considerations for Lucas is the number of prompts required to reach a workable output. Lucas is mindful about the environmental impact of AI and ensures she uses it responsibly. The more adept she becomes in prompting the tools, the more she can minimize her own personal impact. 

"I'm aware of how much energy AI needs and I don't have a good solution for that, but I avoid using LLMs for random queries and when I am using them for work I plan beforehand, so it can reduce that overall impact," she explains.

LLMs and future of journey design as a practice and vendor category 

Professionals from all areas of business are turning to LLMs to execute thousands of different tasks, from creative work to refining ideas, cleaning data, creating code, or analyzing numbers. Naturally, it's changing the jobs humans perform. 

Lucas says that her own role now includes more editing and content management than in the past; tasks that are essential safeguards when using an emerging technology that is known to drift and hallucinate. But in her experience, LLMs do an undeniably great job. "It's interesting to see how you could bring a lot more of this tooling internal potentially, then also really make them your own," she says. 

It isn't difficult to imagine a future where availability of LLMs sees traditional CX tools replaced in some organizations, while other organizations invest in a one-time enterprise LLM license to replace the need for multiple, specialist CX tools. Particularly in scenarios where it can take months to sign off the budget required for specialist tools. Both outcomes will drastically change the vendor landscape. 

"I'm sure a lot of people who work in this field want somewhere where they can post their insights, host their journeys, bring it all together and communicate things how they see them in their minds. I feel like especially in CX, UX, all these human-centered design roles, it's going to make our work that much better. It will be interesting to see how that goes," Lucas says.

Quick links 

• VIDEO: Listening smarter: Bringing customer understanding in-house
• VIDEO: Where is the AI vendor market going, and what will this mean for CXOs
• CX isn't in crisis. It's having its finest hour