Designing success with generative AI

The buzz around generative AI is undeniable, but many organisations struggle to unlock its full potential.

Having worked on a variety of generative AI projects—from proof of concepts (PoCs) to full-scale implementations—I’ve found that success always comes down to building a strong foundation. It starts with clearly identifying the business problem, understanding where generative AI can deliver the most value, and methodically working towards a practical and impactful solution.

In this blog post, I’ll walk you through the essential steps for leveraging generative AI successfully, using industry-leading frameworks and tools to maximise impact.

Start with business representation

Before implementing generative AI, you need to establish a solid understanding of your business’s current state. What are the key challenges? Where are the opportunities? And most importantly, how can generative AI provide value?

Begin by gathering all relevant stakeholders. Conduct a thorough current state analysis to explore existing processes, tools, and challenges, and pinpoint where users encounter friction. This will help you identify opportunities for generative AI solutions, whether through operational efficiency, cost reduction, or a new service offering.

For instance, in a recent project with an energy client, their IoT device was collecting valuable data on household power usage but failed to provide actionable insights. By mapping their current customer journey, we highlighted generative AI-driven opportunities, such as personalised energy-saving tips and improved solar performance monitoring.

Understand what generative AI is good at

Generative AI has revolutionised industries, but it’s not a one-size-fits-all solution. In our experience, we’ve found it excels in the following areas:

• Content generation, such as automated product descriptions or personalised marketing materials.
• Complex reasoning and summarisation, including chatbot support and document summarisation.
• Pattern recognition and anomaly detection within unstructured data, like analysing customer feedback for trends.

To determine if generative AI fits your needs, consider the type of data and task complexity. If the task involves creating new content or interpreting complex data relationships, generative AI is likely a strong candidate. For more straightforward, structured data tasks, traditional machine learning might be a better fit.

The role of experience design in generative AI

One of the most critical—and often overlooked—components of successful generative AI projects is experience design (XD). While it’s easy to focus on the technology itself, XD ensures that the solutions we build are both functional and meaningful to the people who use them.

In the context of generative AI, XD plays a pivotal role in:

• Aligning generative AI outputs with user needs: Whether it’s a chatbot, a recommendation engine, or a content generator, the success of generative AI hinges on how well it meets the expectations of its users.

• Ensuring usability and accessibility: Complex generative AI solutions can be daunting. XD helps simplify the user interface and experience, making advanced generative AI tools accessible to a wider audience.

• Bridging the gap between business goals and technical implementation: XD ensures that the technology not only works but also delivers measurable value to users and the business.

A strong focus on XD can be the difference between a generative AI solution that feels clunky and one that fits seamlessly into the user’s workflow. During a recent project, we designed an AI-driven internal tool for a client’s operations team. By incorporating user feedback early and often, we ensured the tool was intuitive and delivered exactly what the team needed to make faster, more informed decisions.

Applying the Desirability, Viability, Feasibility Framework

The Desirability, Viability, Feasibility (DVF) framework is a widely used tool for evaluating and prioritising innovative projects, including generative AI initiatives. It helps teams focus on opportunities that balance user needs, business goals, and technical feasibility.

• Desirability: Does the solution address a real user problem?
• Viability: Will it deliver measurable business value?
• Feasibility: Can it be implemented with current resources and technical capabilities?

For one energy client, applying the DVF framework highlighted a generative AI feature designed to optimise solar energy usage. This solution was highly desirable (users wanted actionable insights), viable (it aligned with business goals to increase customer engagement), and feasible (the required data was readily available).

Leverage AI canvases for structured thinking

AI canvases are powerful tools that provide structure and clarity throughout your generative AI project. They help teams visualise the problem, design solutions, and account for critical factors like data and ethics.

Depending on your project’s phase, there are different canvases you can leverage:

• AI canvas: Offers a high-level overview, including the business problem, AI solution, data requirements, success metrics, and risks.
• AI project canvas: Focuses on solution design, covering outputs, customer experience, and data integration.
• AI ethics canvas: Explores critical ethical considerations such as fairness, transparency, accountability, and privacy.

For a retail client, we used the AI project canvas to map out a recommendation engine. This exercise helped the team understand how generative AI would integrate with their e-commerce platform and defined key outputs, such as personalised product suggestions. By visualising these elements, we could pinpoint exactly how generative AI would enhance the user experience and drive business value.

Practical tips for running canvas sessions:

• Prepare thoroughly: Before the session, gather relevant data and define clear objectives.
• Engage the right stakeholders: Ensure representation from business, technical, and user experience teams.
• Facilitate effectively: Encourage open discussion but keep the team focused on the canvas sections.
• Avoid common pitfalls: Don’t spend too much time on one section—aim for a balance between depth and progress.

While canvases are excellent for guiding early exploration and fostering alignment, they’re just the beginning. Once the initial groundwork is laid, transitioning to a more comprehensive framework is crucial to move from planning to execution.

Implementation challenges and mitigation strategies

Even with robust canvases and frameworks, generative AI projects can face several common challenges:

• Integration with legacy systems: generative AI solutions often need to fit into existing infrastructures. This can be a complex, time-consuming process.
• Data governance and quality: Poor data quality can derail a generative AI initiative. Ensure early-stage data audits and regular quality checks.
• Change management: Introducing generative AI can disrupt existing workflows and roles. Engage teams early and provide training to ensure smooth adoption.

Mitigating these challenges requires proactive planning. Start with small, manageable PoCs and expand gradually, ensuring each stage builds confidence and demonstrates value. Additionally, use our generative AI project checklist as a guide to navigate these challenges and ensure all critical steps are covered.

My go to tools for generative AI success

When it comes to implementing generative AI projects, having the right tools at your disposal can make all the difference. At DiUS, we’ve seen first-hand how these tools can streamline collaboration, enhance decision-making, and keep projects on track. Whether you’re aligning stakeholders, assessing data, or building generative AI models, the right technology can save time and reduce complexity.

Here are some of the tools we turn to time and time again for successful generative AI implementations:

• Miro or MURAL: These digital whiteboards are invaluable for running collaborative canvas sessions, especially in hybrid or remote environments. They allow teams to map out ideas, visualise frameworks, and align on project goals in real time, no matter where they are.
• Data visualisation platforms (e.g., Tableau, Power BI): Understanding your data is key to any generative AI initiative. These platforms help you assess data readiness, identify quality issues, and communicate insights clearly to stakeholders.
• Machine learning platforms (e.g., Azure ML, Google Cloud AI): Once your data is ready, these platforms make it easier to build, train, and iterate on generative AI models. They also offer tools for deploying and monitoring models in production.
• Documentation tools (e.g., Confluence): With so many moving parts in an generative AI project, keeping track of decisions, insights, and progress is critical. Documentation platforms provide a centralised space for your team to collaborate and maintain a record of every phase of the project.

These tools don’t just support execution, they enhance collaboration and ensure every team member stays aligned, helping turn complex generative AI initiatives into practical, impactful solutions.

Final thoughts

Generative AI holds incredible promise, but as I’ve learned from working on countless projects, success doesn’t come from the technology alone. It’s about taking a structured, thoughtful approach. By starting with a clear understanding of your business challenges and leveraging frameworks like the DVF framework, you can ensure every step of your generative AI journey is focused on delivering real value.

At DiUS, we’ve seen how combining the right tools, a strong foundation in data, and iterative development can turn ambitious ideas into practical, impactful solutions. Generative AI isn’t a silver bullet, but when used thoughtfully, it can transform the way you work and the value you deliver to your customers.

Whether you’re just starting out or refining an existing approach, remember it’s about solving the right problems in the right way.