AI Design Principles

Make consistently effective design decisions & improve user experience of AI services by adopting these AI Design Principles within your team and organisation.

These principles draw on Service Design, Agile, Lean, Conversation Design and other disciplines and have been put together as a result of more than a decade of working with organisations on AI enablement and transformation programmes, as well as interviewing over 300 expert practitioners in the field of applied AI.

These are the principles we adopt at VUX World when working with clients and teams to deliver transformative change with AI services, and we encourage you and your teams to adopt them too.

First, there’s the list of principles. Then, a downloadable poster of the principles. Finally, a deeper explanation of each principle, why it’s important and the kind of activities you should adopt to follow the principle.

AI Design Principles: Listed

1. Human-centred. Consider and design for the needs of all users, regardless of ability. Research your users and prioritise their needs.
2. Data-driven. Use data to derive use cases, validate designs, fuel interactions and improve outcomes. Start and end with data.
3. Goal-orientated. Be in service of goal completion, always making sure that the user can achieve their outcome successfully.
4. Collaborative. All stakeholders should be actively engaged in the process. Motivated champions make light work of change.
5. Holistic. Integrate solutions into existing business services and infrastructure. Design services, not chatbots.
6. Easy to use. Do the work to make the solution easy to use. Redesign process, technology and data to enable change.
7. Agile. Embody the agile manifesto and improve continuously. Going live is the beginning, not the end.
8. Ethical. Design responsibly, ensuring fairness, transparency, accountability, and respect for privacy.
9. Cooperative: Design to cooperate with the user to reach mutual understanding and goals. Design with Grice’s Maxims.
10. Personalised: Be proactive and pre-emptive. Tailor services to individuals based on history, familiarity and preferences.
11. Multimodal: Use the right modality for the right use case. channel and interaction stage. Not everything has to be dialogue.
12. Prepared: Plan for issues. Use human teams for support & ensure business continuity if service is offline.

 

AI Design Principles: Explained

1. Human-centred

Consider and design for the needs of all users, regardless of ability. Research your users and prioritise their needs.

Why it’s important

This principle is focused on having an outside-in view of your business. Rather than working on assumed problems, based on salience or the highest paid person’s whim; instead, conduct research to understand who your users are; all of your users; everyone that touches the service. That’s external users and customers, internal staff members, partners or suppliers and anyone else that uses the service. Research to find what priority needs, goals, expectations and pains they have. Make this your priority to serve.

Relentlessly focusing on your users will ensure you solve the most important problems in the right way, from finding the right use cases, to designing the right features and delivering the right results.

What it means

Being human-centred means:

• Conducting user research to inform use cases and design decisions.
• Prioritising user needs ahead of perceived business needs. If you solve the user’s need, you’ll solve the other by proxy.
• Designing for accessibility of all kinds; temporary and permanent impairments or disabilities in touching, seeing, hearing and speaking.
• Engaging with a diverse user base, all users of the service, back-to-front, to gather a broad range of perspectives and needs.
• Prioritising product features and project plans based on which items will best meet the needs of your users.

You should develop an obsession with your users and develop a culture where user needs are the primary focus.

2. Data-driven

Use data to derive use cases, validate designs, fuel interactions and improve outcomes. Start and end with data.

Why it’s important

This principle is about relinquishing assumptions of all kinds and developing a culture of data-oriented decision making. That’s using data at the outset of projects and use case scoping to inform what to use case to prioritise. Through design to test and validate assumptions. Within interactions; both pulling data into interactions to serve the use case, and pushing data out of interactions into business systems to enable transformative use cases. And post-interaction for analysis, improvement and value recognition.

Being data-driven is important because it removes opinion and assumptions out of your process and enables you to make decisions based on the reality of your situation.

What it means

Being data-driven means teams should:

• Decide what problem or use case to tackle next, based on customer conversations, feedback, behaviour, demand and intent.
• Validate product and service feasibility with usability or soft-market testing.
• Iterate and improve interactions through test-and-learn practices such as multivariate testing.
• Provide the basis for interactions through injecting relevant data into services to meet customer needs, either through content-based information or system integration.
• Enable users to complete tasks by gathering and moving data from interactions into appropriate sequenced systems.
• Analyse interactions at a granular level to improve interaction, journey and business outcomes.
• Understand and demonstrate the value of the solution on business outcomes.
• Gather intelligence from interactions to improve wider business and service performance.

This also includes the activities surrounding:

• Data discovery: understanding the data you have and its usefulness in enabling AI use cases.
• Data design: creating, massaging and cleansing data in preparation for chunking, ingestion or integration.
• Data delivery: how you feed data into AI systems, as well as extract data out, either in design or training or models and services, in real time during interactions or post-interaction for analysis.

Making data a core part of how you decide, design and improve services will ensure you’re always optimising for what matters in reality, leading to more effective services.

3. Goal-oriented

Be in service of goal completion, always making sure that the user can achieve their outcome successfully.

Why it’s important

This principle focuses on ensuring that every design decision aligns with user goals, and that interactions are designed to focus on outcomes. This means helping users get things done. You should alway aim to design your service to fulfil a user’s need from end-to-end. If your service plays a role in the user’s journey, always design to take the user toward their goal effectively. Monitor and measure the usefulness of the solution in service of goal completion. Optimise journeys and continuously validate that each step helps the user achieve their objectives.

By being goal oriented, you’ll ensure that users accomplish their outcomes, which in turn means that the business will achieve its goals.

What it means

Teams should be goal-oriented by:

• Understanding and documenting user goals through research and user feedback. User stories are a good way to document and stay focused.
• Designing workflows that streamline goal completion and eliminate unnecessary steps.
• Being clear about what the user can accomplish with your service.
• Ensuring that every interaction guides the user toward their goal in a clear and intuitive manner.
• Effectively grounding interactions to bring the user back on track.
• Monitoring and measuring success based on goal achievement rates and user satisfaction.

This focus on user goals and outcomes ensures that your AI services remain effective, valuable and deliver the outcomes users and your business need.

4. Collaborative

All stakeholders should be actively engaged in the process. Motivated champions make light work of change.

Why it’s important

Collaboration is key to successful AI design. To make meaningful change happen, you’ll require support and engagement from across the business. The changing of business processes, job roles, technologies and, eventually, operating models, means that all stakeholders from across relevant departments need to be involved and motivated to ensure deployment success.

What it means

Teams should be collaborative by:

• Involving cross-functional teams in design and decision-making processes.
• Encouraging open communication and idea sharing among stakeholders.
• Maintaining a RACI matrix to understand who needs to be engaged, for what and how often.
• Maintaining a stakeholder engagement plan to ensure the right people are invested and stay supporting the service.
• Finding and building relationships with service champions and key enablers.
• Aligning goals and expectations across different business units to ensure cohesive design efforts.
• Fostering a culture of cross-organisational teamwork where every member feels valued and heard, has a shared common goal and works cross-functionally to disband organisational silos.

Collaborative efforts lead to more innovative solutions, smoother implementation processes and greater adoption.

5. Holistic

Integrate solutions into existing business services and infrastructure. Design services, not chatbots.

Why it’s important

A holistic approach ensures that AI solutions complement and enhance existing business services rather than being a stand-alone appendage that gathers dust over time. For AI to truly impact business operation, transformation and customer experience, it should be integral to the way the business operates and should fuel transformation efforts. Therefore, it needs to integrate into existing business systems, sequence those systems together and integrate into the customer journey seamlessly. AI services will also inform the procurement or development of new systems and technologies required to deliver the capabilities needed to implement services.

What it means

Teams should be holistic by:

• Considering the entire user journey and how AI fits within it.
• Integrating AI services seamlessly into existing infrastructure and workflows.
• Designing solutions that enhance overall service delivery, not just isolated interactions.
• Ensuring that AI services add value across multiple touchpoints and business processes.
• Finding areas where consolidation or new capabilities will enhance business operations.
• Integrate new capabilities within existing services seamlessly.
• Tracking and monitoring that AI is playing a valuable role in the customer journey.

This comprehensive approach ensures a cohesive and effective service experience for users and enhances business outcomes.

6. Easy to use

Do the work to make the solution easy to use. Redesign process, technology and data to enable change.

Why it’s important

Ease of use is critical for user adoption and satisfaction. To make things simple on the front-end often requires substantial effort on the back-end. Redesigning business processes, rewriting company policies, rearchitecting systems and data, implementing new technologies and sunsetting others, bringing in new providers, making new integrations and changing the status quo. Just because it’s hard or complex, doesn’t mean the user journey should suffer. Always do the work to simplify interactions to make them as intuitive and accessible as possible. This is where you’ll find the biggest business and customer outcomes.

What it means

Teams should ensure ease of use by:

• Starting with the ideal experience and working backward to the process and technology requirements.
• Using journey mapping and lean methodologies to remove friction and waste from experiences.
• Challenging existing processes and behaviours, redesigning processes to remove waste and deliver faster value to users, knowing that means the requirement to change systems, operations and roles.
• Working with persistence to push through complex back-end changes that benefit front-end experiences.
• Conducting usability testing to identify and eliminate pain points.
• Designing interfaces and interactions that are intuitive and require minimal effort to use.
• Continuously iterating based on user feedback to enhance usability.

Prioritising simplicity and the reduction of user effort ensures that users can effectively interact with your AI services and accomplish their goals. In turn, it means that your business becomes more efficient, differentiated and future proofed.

7. Agile

Embody the agile manifesto and improve continuously. Going live is the beginning, not the end.

Why it’s important

An agile approach allows for the prioritisation of features that will have the biggest impact, putting services in front of users early to gather feedback and iterate, with continuous improvement and adaptation to changing user needs and market conditions.

The agile manifesto and principles are to be adopted to foster collaboration, reduce the risk of launching the wrong service and to ensure the performance of solutions meet user and business needs.

What it means

Teams should be agile by:

• Valuing individuals and interactions over processes and tools, working software over comprehensive documentation, customer collaboration over contract negotiation and responding to change over following a plan.
• Adopting iterative design and development cycles with regular feedback loops.
• Emphasising flexibility and responsiveness to user feedback and data insights.
• Testing the riskiest assumptions of a service first and validating design decisions.
• Encouraging experimentation and learning from both successes and failures.
• Continuously refining and enhancing your AI services post-launch.

This approach ensures that your AI solutions remain relevant and effective over time, reduce the risk of designing the wrong thing or a poor experience, and maximise business impact.

8. Ethical

Design responsibly, ensuring fairness, transparency, accountability, and respect for privacy.

Why it’s important

Designing ethically means to design artificial intelligence services in ways that align with moral and ethical principles. The goal is to ensure that AI technologies are designed and deployed responsibly, considering their potential impacts on individuals, society, and the environment. Ethical considerations are paramount in AI design to build trust with users, protect against brand damage and ensure responsible use of technology.

What it means

Teams should be ethical by:

• Ensuring transparency in how AI decisions are made and communicated to users and stakeholders.
• Implementing safeguards to protect user privacy and data security, making sure services operate reliably and minimise risks and harm to users.
• Designing AI services to be fair and unbiased, avoiding discriminatory outcomes and treating all groups equitably.
• Establishing clear responsibility for AI systems’ outcomes, including mechanisms for addressing grievances and correcting errors.
• Considering the environmental impact of AI development and deployment, and striving for sustainable practices.

Adhering to ethical principles fosters user trust and long-term sustainability of AI services.

9. Cooperative

Design to cooperate with the user to reach mutual understanding and goals. Design with Grice’s Maxims.

Why it’s important

Cooperation with users ensures that AI services work harmoniously with user intentions and needs. Through cooperating with users, we ensure the AI systems effectively collaborate with the user to achieve mutual understanding throughout the interaction, and that two two, user and system, work together to accomplish goals. This fosters a more intuitive, efficient, and user-friendly interaction, increasing user happiness and satisfaction.

What it means

Teams should be cooperative by:

• Ensuring that all information provided by the system is true and backed by reliable sources.
• Delivering exactly as much information as needed in an interaction—no more, no less—to satisfy the user’s inquiry or need.
• Keeping responses relevant to the user’s queries and context of the interaction.
• Presenting information in a clear, orderly, and straightforward manner, avoiding ambiguity and obscurity.
• Providing mechanisms for users to give feedback and adjust interactions as needed.

This cooperative approach leads to more effective and user-friendly AI services that are more natural and enable greater chances of success.

10. Personalised

Be proactive and pre-emptive. Tailor services to individuals based on history, familiarity and preferences.

Why it’s important

Personalisation is focused on treating each user as an individual. By being pre-emptive, we reduce the need for users to contact the business whatsoever, reducing contact volume and costs. Being proactive and personalising interactions, we develop deeper relationships and create more delightful experiences which improves journey outcomes and loyalty.

What it means

Teams should personalise interactions by:

• Using data, events and triggers to spot emerging issues and resolve them before they arise, or fix them before users develop a need to reach out.
• Initiating interactions with users based on issues detected and fixed to reduce contact.
• Using previous interaction and journey history to begin interactions in contextually relevant places.
• Leveraging user data to tailor responses and recommendations.
• Designing adaptive interfaces that adjust based on familiarity, user behaviour and preferences.
• Providing continuity across interactions by remembering user context and history.
• Ensuring that personalisation enhances rather than complicates the user experience.

Personalised interactions make users feel valued and understood, increasing satisfaction and loyalty.

11. Multimodal

Use the right modality for the right use case. channel and interaction stage. Not everything has to be dialogue.

Why it’s important

To deliver the best experience means utilising the right tools for the job. Making sure you identify the most appropriate channel and surface for your use case means that more users will use it, increasing adoption and engagement. Designing the right blend of UI components and modalities for the channel, and for each turn and stage of the journey, enhances the effectiveness of the service and increases usability, resulting in more successful interactions and journeys.

What it means

Teams should use multimodal approaches by:

• Designing the most effective end-to-end experience, regardless of whether it utilises AI.
• Identifying the best modality (text, voice, visual, tap, swipe) or blend of modalities for each use case.
• Designing each stage of the journey and conversational turn, where relevant, to utilise the most effective UI modality to increase progression through the service and goal completion.
• Integrating various modalities seamlessly to create a cohesive user experience.
• Adapting interactions based on the context, user needs and preferences.

A multimodal approach ensures that users can interact with AI services in the most effective and convenient way. It means that more users will adopt your service and have successful experiences with it.

12. Prepared

Plan for issues. Use human teams for support & ensure business continuity if service is offline.

Why it’s important

Things can go wrong at a service or interaction level. At an interaction level, the interaction itself may fail if the user has trouble using the interface, and so human support should always be available and never denied. At a service level, many things can happen that would take the service offline, from API failures to wider business issues.

Being prepared for when things go wrong ensures continuity and reliability of AI services. It prevents sudden spikes in contact or decreases in satisfaction and outcomes in unexpected situations.

What it means

Teams should be prepared by:

• Developing contingency plans for common failure scenarios, such as misunderstandings or journey failures.
• Implementing fallback mechanisms to handle interaction failures gracefully.
• Ensuring human support is available to assist users when needed.
• Developing business continuity plans for emergency situations where the service is offline.
• Tracking and monitoring key failure point risks, such as API performance, interaction bottlenecks and system uptime.
• Testing and refining disaster recovery plans regularly.

Preparation like this ensures that users can accomplish their goals, even in the face of unexpected issues.

Putting the AI Design Principles into practice

Any team can begin incorporating these principles into their organisation today. If you’d like some help, through training or coaching, feel free to reach out to one of our experts.