The Ghost in the Service Machine: Designing with AI in the Loop

As AI becomes both design partner and service participant, how do we create experiences that feel human, adaptive, and truly collaborative?

As Lee wrote about last week, with AI tools lowering the barriers to software development, service and experience design become more important than ever. Service design is about understanding human needs, mapping journeys, and orchestrating seamless experiences. But whilst AI is increasing the need for service design, it is also changing the practice in fundamental ways.

First, AI is no longer just a tool for automation. It’s an active design partner. From generative AI to predictive analytics, these technologies can help us not only optimise existing services, but also imagine new possibilities that weren’t feasible before.

Second, we are no longer designing services only for human users. AI itself is becoming a participant within services, handling inquiries, making decisions, orchestrating workflows, and even collaborating with humans in real time. This means service design must now account for how humans and AI interact, collaborate, and co-create value together.

Some fear that AI will strip away human warmth, reducing services to cold efficiency. Others see a future where it enhances personalisation and creates more meaningful interactions. The reality depends on how we design this interplay and how we embed AI both as a partner and a participant in the services we create.

AI as Service Design Thought Partner

In 2025, AI is moving beyond theoretical discussions. Enterprises are embedding AI into customer journeys, supply chains, and internal processes, not just to automate, but to augment. The question is no longer if AI will transform service design, but how we ensure it enhances human experiences while also integrating AI as a participant in its own right.

This edition of ShiftAcademy explores AI-augmented service design, breaking it down into three levels:

• Optimising Existing Services – AI for improving efficiency and reducing friction.

• Enhancing Service Experience – AI for hyper-personalisation and adaptive interactions.

• Reimagining Service Models – AI as a co-designer, shaping entirely new service paradigms.

For many organisations, the first step in AI-enhanced service design is not reinvention, but optimisation. Before we create new models, we can use AI to refine the services we already deliver - reducing friction, improving reliability, and unlocking efficiencies at scale.

Traditionally, service optimisation has relied on human-driven process improvement: analysing workflows, identifying bottlenecks, and incrementally refining operations. But AI changes the game. Machine learning models can surface inefficiencies in real-time, predictive analytics can anticipate service disruptions before they happen, and intelligent automation can streamline complex multi-step processes.

At this level, AI acts as an efficiency engine, improving service operations in three core ways:

1. Reducing Friction – AI helps eliminate pain points in service delivery, from slow customer response times to manual back-office tasks.

2. Predicting and Preventing – AI anticipates service failures or bottlenecks, enabling proactive rather than reactive interventions.

3. Enhancing Responsiveness – AI-driven systems enable real-time adjustments, ensuring services dynamically adapt to demand and context.

🎯 Scenario: AI-Driven Service Recovery

One of the most powerful applications of AI in service design is predicting service disruptions before they impact users and dynamically adjusting in response.

💼 Case Study: AI-Powered Incident Resolution in IT Support

Fujitsu, a global tech services provider, used ServiceNow’s Predictive Intelligence to shift from reactive to proactive IT support. By analysing historical incidents and real-time data, their AI system could:

• Auto-categorise and route service requests

• Detect early warning signs of disruption

• Trigger alerts or initiate auto-remediation before users noticed issues

This approach significantly reduced manual triage, improved first-contact resolution, and enabled faster response times.

AI didn’t replace the service desk, it made it smarter, faster, and more responsive.

Full case study here

🔬 Experiment: Mapping AI Optimisation Opportunities

To explore how AI could optimise your existing services, start with a Service Friction Mapping exercise. GenAI tools can be your co-pilot in this process, helping you surface hidden pain points, reimagine workflows, and identify practical starting points for experimentation.

Here’s how to run the exercise with GenAI support:

1. Identify a service process that suffers from bottlenecks or inefficiencies.

🧠 GenAI Prompt to try:

“List common service processes in a [sector/department] where users typically experience delays or frustration. Include examples of bottlenecks or pain points.”

(e.g. HR onboarding, IT helpdesk)

2. Map out the current workflow, pinpointing where manual effort, slow response times, or unpredictability cause friction.

🧠 Prompt to try:

“Describe the typical steps involved in [chosen service process]. Highlight where delays, manual interventions, or user frustration are most likely.”

(You can also paste in your current process map and ask questions.)

3. Identify AI augmentation opportunities, where automation, prediction, or real-time analytics could smooth out pain points.

🧠 Prompt to try:

“Given the following service process and pain points, suggest AI use cases that could reduce friction or improve efficiency.”

(Then paste in your mapped workflow or friction summary.)

Or, more generally:

“What kinds of AI capabilities (e.g. automation, prediction, NLP) could be applied to improve the efficiency of [service process]?”

4. Prioritise quick-win experiments, starting with AI capabilities that require minimal system integration.

🧠 Prompt to try:

“From the list of AI opportunities above, which could be piloted quickly with minimal integration or training? Suggest a prioritised list of small experiments with high potential value.”

This is a lightweight way to introduce GenAI into your service improvement practice, not just as a solution enabler, but as a thinking tool. It also helps teams build confidence by showing how AI can support the design of change itself.

⚠️ What Not to Do When Mapping AI Optimisation Opportunities

• Don’t automate broken processes: Applying AI to a flawed workflow just speeds up dysfunction. Fix the friction first.

• Don’t ignore the human in the loop: Even in back-office tasks, humans often add judgment or context. Know where they’re essential.

• Don’t treat AI as plug-and-play: Many AI capabilities require quality data, contextual prompts, and iterative tuning to deliver value.

• Don’t start with the fanciest use case: Aim for small, boring wins that reduce effort or failure points. That’s where trust in AI starts to build.

Read further to discover how AI shifts from behind-the-scenes optimiser to a visible, responsive layer in your user journey.