Designing AI-enhanced solutions, a new design process
In the rapid pace of AI advancement, it’s tempting to jump right into developing AI-enhanced products. However, successful innovation always comes back to the user. As Steve Jobs wisely advised, we must start with the customer experience (CX) and work back toward the technology.
But how does the design process adapt when AI comes into play? How do we ensure that we’re not just adding AI for the sake of it, but truly enhancing user experience?
While aligned with established practices like Design Thinking, the introduction of AI requires some unique alterations to the design process. In this article, we cut through the AI hype to focus on these critical changes, delivering actionable insights for anyone venturing into the world of AI-enhanced product development. We’ll help you keep your focus where it belongs – on creating solutions that genuinely improve the user’s life.
Where do the modern design processes fit into this new picture?
Design processes today are customer-centric in their nature, regardless of whether they’re tailored for products or services. Creating solutions that solve your customers’ problems is at the heart of them all. At a glance, the CX design processes we all know and love seem perfectly applicable to designing solutions that use AI. However, the more you scratch the surface, the more you start to see where things need to change.
Let’s use Design Thinking as an example.
The process starts with empathizing with your customer as well as the business to understand their pains and objectives.
You then synthesize the data and define the problem you are trying to address in a clear, user-centered, and actionable problem statement.
Next, you gather stakeholders from various teams to ideate a wide range of possible solutions to the problem statement. You collectively decide on which one to prototype, and you move to the next step. This phase – in its current form – doesn’t cater for the complexities of the vast possibilities that AI introduces. (I go into detail about this later.)
The prototyping is often light and requires little financial investment. User testing validates your hypotheses and sets you up with just enough data to justify advancing into production (or testing another idea).
This phase is vulnerable to disruption by the introduction of AI and its inherent unpredictability. AI models can behave in ways not anticipated during their training, and machine learning can make it difficult to predict how an AI-enhanced product will respond to certain user inputs.
So how does design need to change to support the creation of AI-enhanced solutions?
While I’m not suggesting a complete overhaul of familiar processes, significant changes are necessary, such as:
UX best practices
AI systems learn patterns from the data they are given. If the data is poor quality, incomplete, or biased, the AI system will generate outputs that mirror these deficiencies. This could manifest as inaccurate predictions, irrelevant recommendations, or even offensive or discriminatory behavior. Such outcomes will lead to a poor user experience.
Take the smartwatch (wearable) for example. They can sense your movement and then prompt you with a question like, “Would you like to start recording your outdoor run”. This is potentially a great and really simple example of AI in use today.
In a non-AI user journey, the user experience would be deterministic in that the user would have had to manually choose from a menu with a series of options that they wanted to log their run. The UX team knew the user’s intention and designed a clear path to the outcome. Effective, but very linear and predictable. In stark contrast to the next example.
Today, with the help of AI, that experience could be just as seamless. So much so, that it’s almost invisible. The sensor detects my movements, and the software works out the probability that I am running vs cycling vs walking vs swimming and prompts me to look at the screen with a subtle haptics vibration, where the question is there asking me if I would like to record my run. It’s 100x more convenient than the old way, and I never thought I needed this feature until I tried it. That is a UX win.
On the flip side, here is an example of a UX failure involving an AI-enhances service. The dreaded help desk chatbot. Recently, and more often than I care for, I’ve found myself at the mercy of an automated chatbot that has been unable to determine my issues, and instead of being helpful and leaving me loving the experience, it raised my blood pressure by repeatedly asking me the same questions, sending me into this eternal loop of frustration and resentment towards the brand.
It’s a lengthy chapter, so I’ll summarize it in one sentence. Designing AI solutions requires a probabilistic approach, as the outputs are much less predictable than the traditional (deterministic) and linear paths we’ve been designing up until now.
Expand the team of collaborators
Another significant alteration to the traditional CX design process involves expanding the range of collaborators. Historically, this group included customers, researchers, UX designers, stakeholders, and software engineers. Today, designing AI-enhanced solutions necessitates the inclusion of additional key players for success. These include AI specialists, and Content Designers who shape the user interface, providing clear guidance and feedback to users. These new roles augment the familiar roster of collaborators, enriching the process and outcome.
CX design principles must focus on establishing trust
The next significant change involves redefining our CX design principles. With AI-enhanced solutions, we need to focus on establishing trust between the users and the technology. These systems, such as large language models (LLMs), can occasionally make mistakes and fail to acknowledge them. Our solutions must help our end-users identify inaccurate or potentially harmful AI-generated content, and should focus on the principles like the following examples:
Transparency
Be clear about what data the AI uses, how it’s being processed, and how decisions are made. For instance, we could show users a simplified flow of how the AI processes their data or briefly explain how the AI makes specific predictions or recommendations.
Explainability
People are more likely to trust a system if they understand how it works. Therefore, we should explain the AI’s processes in simple terms whenever possible. For an AI recommendation system, we might include a feature that explains why a particular recommendation was made.
Reliability
Ensuring the AI performs consistently and accurately over time helps build trust. An AI chatbot, for instance, should provide accurate and relevant responses to user queries consistently.
Privacy and Security
Emphasize that user data is safe, and their privacy is respected. Clear and user-friendly privacy policies and secure data handling practices will play a key role in this aspect.
Accountability
Users need to know who is responsible if something goes wrong with the AI. In the user agreement or terms of service, we must clearly state who is accountable for any issues or damages caused by the AI.
Bias Mitigation
Lastly, we must address potential bias in AI. Outlining efforts to ensure the AI’s algorithms are tested and audited to mitigate any discriminatory or unfair practices can go a long way in establishing trust.
Failing to do this could severely damage the customer relationship and the business’s brand reputation behind the solution.
Ethics & implications
Addressing ethical considerations and bias is an often overlooked yet critical aspect. Despite their objective facade, AI systems can unintentionally mirror and propagate existing biases in their training data. A CX designer’s mandate is to design with an intent for fairness, accountability, and transparency.
Less UI, more UX
Next, when creating AI-enhanced solutions, the focus often shifts away from the Graphical User Interface (UI) and towards crafting the User Experience (UX). The complexity lies not in UI elements, but in designing the communication between the user and the AI. UX designers in AI-driven projects must design how the system communicates its functions, limitations, and decisions to the user. They guide the user in providing clear inputs and handling the AI’s output effectively.
This shift is largely due to the nature of AI solutions, for example solutions where interaction happens through natural language processing rather than graphical elements. This creates a deceptively simple experience for the user. Most of the complexity is hidden and user only experiences a swift response from the AI. Traditional digital experiences rely on graphical user interfaces. Some AI experiences rely on voice or text as the interface. This makes it even more crucial for UX to bridge the gap between the complex AI tech and a seamless user experience.
Context is key
When designing AI-enhanced solutions, we must be aware that while the foundational principles of CX design continue to hold true. Understanding user needs, conducting task analysis, and iterating based on feedback – AI introduces unique aspects that need explicit attention.
AI systems come with a certain degree of unpredictability due to their adaptive nature, a marked shift from traditional deterministic design. CX Designers, now grapple with creating interfaces that cater to complex, multi-modal interactions (text, voice, touch, etc.) that AI systems are capable of.
Moreover, AI systems are data-dependent and the quality, quantity, and nature of data used can have profound effects on the user experience. AI systems learn patterns from the data they are trained on. If the training data is poor quality, incomplete, or biased, the AI system will generate outputs that mirror these deficiencies. This could manifest as inaccurate predictions, irrelevant recommendations, or even offensive or discriminatory behavior. Such outcomes will definitely lead to a poor user experience. As such, considerations around data privacy, security, and even data representation become central to our design principles.
Designers also need to help users comprehend the capabilities and limitations of AI – it’s a fine balance to strike between promoting user reliance on the system and fostering critical awareness about when to question the AI’s outputs.
We touched on ethics in the CX Design Principles chapter above. In essence, determining the appropriate level of AI involvement in a solution isn’t a one-size-fits-all answer. It relies heavily on the context – user needs, task complexity, environmental constraints, and ethical considerations. By taking these into account, we can create AI-enhanced solutions that are not only effective but also responsible and respectful of our users.
Performance = UX
In the world of digital solutions, the performance of the solution is a critical component of the user experience. This is particularly important for AI-enhanced solutions, where user expectations might be even higher due to the implicit promise of advanced technology.
Performance in the context of AI-enhanced solutions isn’t just about speed, though that’s certainly a significant factor. Nobody wants to wait for a slow-loading application or deal with laggy interactions. However, performance also extends to how accurately the AI interprets user inputs, and the relevance and utility of the outputs it provides.
For instance, consider a voice assistant application. If the AI doesn’t accurately interpret the user’s spoken commands or takes too long to respond, the user experience will be negatively impacted. Similarly, if the AI doesn’t provide a useful or relevant response, the user is likely to be frustrated.
AI can enhance performance in many ways. It can process vast amounts of data more quickly than a human could, “learn” from past interactions to provide more relevant responses, and adapt to a user’s changing needs over time.
However, it’s essential to remember that good performance isn’t just about what happens when everything goes right. It’s also about what happens when things go wrong. AI-enhanced solutions need to be designed to handle errors gracefully, providing clear and useful feedback to the user when things don’t go as planned. In summary, when designing AI-enhanced solutions, remember that performance isn’t just a technical concern—it’s a fundamental aspect of the user experience. When performance is good, the solution feels effortless and enjoyable to use. When performance is poor, the user’s trust in the solution—and the organization behind it—can be tarnished.
Expansive thinking vs. limiting possibilities
Creating AI products requires a delicate equilibrium between innovative thinking and practicality. Here’s how to maintain this balance:
- Define objectives and establish clear goals for your AI solution from the outset, focusing on user problems it needs to solve.
- Prioritize users and ensure their needs and expectations guide the CX design process, limiting scope to truly beneficial features.
- Encourage diverse collaboration by inviting perspectives from varying fields such as machine learning, UX design, and data science to identify opportunities and constraints.
- Prototype early and validate your assumptions and gauge the practicality of your solution by testing a minimum viable product or prototype with users early on.
- Iterate continually post-launch to refine the solution based on user feedback and performance data, enhancing capabilities over time.
- Educate stakeholders and align expectations by informing everyone about AI’s potential and limitations.
Balancing expansive thinking and limiting scope can be challenging with AI projects. However, clear objectives, user-centric design, interdisciplinary collaboration, and continuous iteration make it achievable.
Conclusion
Designing for AI demands a shift in our thinking and processes. We need to accommodate the unique attributes and constraints AI brings into the CX design equation, from a probabilistic customer journey, establishing trust, and including new roles in our collaboration, to balancing expansive thinking with practical limitations.
But you don’t have to navigate these changes alone. Our team at A-CX brings a wealth of experience and expertise to the table. We’ve been at the forefront of AI solution design, wrestling with these very challenges, and finding innovative ways to create effective and user-centric AI-enhanced solutions.
We understand the criticality of trust, the balance between UI and UX, the importance of performance, and the need for iterative feedback loops in AI solution design. Our tried and tested methodologies are built upon these very principles.
Whether you are at the ideation stage, looking to improve an existing product, or anywhere in between, we can provide the guidance and support you need to successfully incorporate AI into your solution.
Our aim is to empower your business with AI capabilities, ensuring a seamless, enjoyable, and efficient experience for your users, leading to improved customer satisfaction and business success.
Reach out to us today, and let’s shape the future of AI design together!
