Empowering vulnerable employees with technology

Client
Research conducted with real client as part of
Services
Research
Samhall
Research conducted with real client as part of
Hyper Island MA
Services
Research
Strategy
The Challenge
Samhall is a Swedish organisation that employs people facing barriers to the job market - cognitive, mental and physical disabilities included. During a two-year work placement phase, employees move across different roles and sectors, often without consistent guidance or support.
The question
How could AI be used to improve the daily work experience of these employees?

Research Approach
Access to Samhall employees was limited, which shaped our methodology significantly. We conducted desk research and interviewed a special needs teacher, a deliberate choice to ground our understanding in the real experiences of people supporting this population, even when we couldn't speak to the employees directly.
Midway through the project, a client check-in refocused our direction. Rather than exploring AI automation for operations, we pivoted to designing an employee-facing chatbot, Sammy, aimed at supporting workers with daily practicalities: scheduling, location guidance, break reminders, and access to benefits information.
This pivot was the right call. The research had shown us that what employees needed wasn't efficiency optimisation - it was reliable, low-friction support that reduced their cognitive load and freed managers from repetitive administrative tasks to focus on genuinely human work.


Design & Prototyping
We built personas and service journeys to guide the design, created a functional database structure, and tested the chatbot concept using Claude, ChatGPT and no-code tools including Landbot.
We hit real technical constraints, managing four integrated databases proved too complex for the project scope, and Landbot's responses were unreliable. We made the decision not to invest further in a tool we couldn't trust for this user group. That decision mattered: when your users are vulnerable, an unreliable product is worse than no product.
We also made a principled design decision to keep Sammy's scope deliberately narrow - scheduling, transport, breaks, benefits. Conflict resolution, personal information and sensitive HR matters were explicitly out of scope. Human support should always take precedence in those moments.

Outcome
The project delivered a validated concept, service blueprints, personas and interaction designs for Sammy. The technical prototype was not production-ready - a limitation we were honest about. The value was in the research direction, the design principles established, and the clear articulation of where AI can genuinely help vulnerable workers without overstepping.
Research conducted with Samhall as a live client partner as part of an MA in Digital Management at Hyper Island.