Define the decisions
Agree audiences, tasks, system boundaries and the questions research needs to address.
AI & Automation Experience
Understand when AI helps people, when it creates friction and where human control matters.
Products, Services & Environments
Understand when AI helps people, when it creates friction and where human control matters.
An AI feature can produce an answer without helping someone complete their goal. AI and automation experience research examines how people understand, use, check and respond to automated systems. It helps your team identify useful assistance, confusing interactions and situations where users need more control or support.
What we look at
Task success with AI assistance- Trust, transparency and control
- Fallback when the model is wrong
- Prompt and conversation friction
How we study it
- Human-AI moderated task tests
- Chatbot conversation evaluation
- Trust and explainability interviews
- Automation UX recommendations
What we investigate
Questions this research can answer
Do people understand what the system can do? Can they provide the information it needs? Do they recognise uncertainty or an incorrect response? Can they check, edit, reject or reverse an action? These questions connect the AI interaction to a practical user goal.
Research can also explore onboarding, conversation flow, handoff to human support and the effort needed to review outputs. In workplace tools, we examine how assistance fits existing roles and processes rather than treating willingness to try a feature as proof of productivity.
What we investigate
Experiences we can investigate
Potential study areas include customer-service chatbots, employee assistants, search and recommendation interfaces, automated approvals and AI-assisted content workflows. The research scope defines the tasks, permissions and systems involved. Sensitive or consequential scenarios need additional safeguards and relevant specialist input.
Early concepts can be explored through interviews and prototype testing. Working pilots support task-based observation. Diary studies or follow-up interviews may help investigate repeated use where access and duration allow. Product development and model evaluation are separate from the user research scope.
Our approach
Research grounded in evidence
UX Pacific studies the human experience around chatbots, assistants, recommendations and automated workflows. We scope research around your audience, product readiness and the decisions ahead, whether you are exploring an idea, preparing a pilot or improving an existing experience.
From context to opportunity
Our research process
Review readiness and risks
Check access, test data, safeguards and known limitations.
Recruit suitable users
Consider roles, domain knowledge, AI familiarity, language and access needs.
Test realistic scenarios
Observe ordinary use and agreed failure or recovery situations, with neutral follow-up questions.
Analyse evidence
Separate user difficulties, system behaviour, domain correctness and test constraints.
Agree next steps
Discuss priorities, unresolved questions and suitable follow-up evaluation.
The output
What your team can receive
Depending on scope, outputs can include task findings, conversation or workflow issues, evidence-backed priorities and recommendations for onboarding, explanations, controls and recovery. Approved examples illustrate what participants encountered and how they responded.
Reports document participant coverage, tested scenarios, system configuration and limitations. A findings discussion helps product, design and operations teams decide what to change, what to test again and where technical or domain expertise is needed. Dashboards, implementation and repeat monitoring are separately agreed.
Research focus
Trust that matches the situation
The goal is not simply to make people trust the system more. Research explores whether their confidence fits its demonstrated capabilities and limits. We examine what users believe an answer means, whether they check important outputs and when they choose not to rely on automation.
Explanations, source references and uncertainty messages can be tested for comprehension and usefulness. An explanation that sounds convincing is not evidence that the output is correct. Domain-specific accuracy needs an appropriate reference and qualified assessment, agreed separately where required.
Research focus
Control, mistakes and recovery
Useful test scenarios include incomplete requests, misunderstood intent, unavailable information and incorrect or conflicting responses. We observe whether people notice the problem and can recover through clarification, correction, cancellation or human support.
For systems that take actions, research can examine whether users understand permissions, previews and confirmation steps. Simulated tasks or a controlled environment help avoid unintended messages, purchases or account changes. Any deliberately introduced errors are planned with appropriate safeguards and participant debriefing.
What we investigate
Understanding effort and task outcomes
Depending on scope, measures can include task completion, assistance, correction effort and reported ease or confidence. Reviewing and fixing an output is part of the workflow, not an invisible extra step. We distinguish system response time from the total effort needed to reach a usable result.
A comparison with an existing workflow needs suitable tasks and fair conditions. Product version, configuration and study dates are documented because system behaviour may change. Small qualitative studies identify issues; they do not establish organisation-wide productivity gains.
Global research
Language, inclusion and cross-market use
Different users may interpret prompts, tone and explanations differently. Cross-market studies agree local recruitment feasibility, language support and comparable scenarios. Inclusion research can explore relevant access needs, digital confidence and interaction formats without assuming one audience represents everyone.
The study can surface concerning experiences or inconsistent treatment within its coverage. It is not a comprehensive fairness assessment of a model or proof that all demographic groups receive equivalent outcomes.
Research focus
Privacy and research safeguards
Before testing, agree what participants may enter, which systems receive the data and how study material is handled. Use synthetic or appropriately de-identified information where possible. Avoid real credentials and confidential records in unapproved tools. Recording, observer access and any external analysis tools need agreed permissions and handling arrangements.
What we investigate
Research boundaries and common questions
Before we begin
Frequently asked questions
No. Research can explore current experiences and needs before a concept exists. We still define the decision your team needs to make so the study remains focused.
Timing depends on the question, method, participant profile and review process. After an initial discussion, we define scope, milestones and a realistic schedule.
Tell us which people and experiences you want to understand, what your team is considering and what evidence you already have.
Patient Experience
Understand patients' experiences of access, communication and care-related services.
Explore
Citizen Experience
Citizen journey mapping, public service usability testing, and form testing for services people cannot opt out.
Explore
Partner / Dealer Experience
Partner interviews, dealer journey mapping, and portal usability testing for the channels that sell and suppor.
Explore
Retail & Shopper Experience
Understand discovery, comparison and shopping across channels.
Explore
Let us help shape the next step
Understand the human side of automation.
Make your next step better informed.
Next research service Journey Experience Understand the human side of automation. Make your next step better informed. Contact Us Share your AI experience and the decisions ahead.