How an AI chatbot can smoothen the volunteering experience

The statistics are clear. Singaporeans want to volunteer – but they don’t.

A recent study found that over half of Singaporeans intend to volunteer, but only 22 per cent do. In a country grappling with an ageing population, a limited talent pool, and rising healthcare costs, this willingness to volunteer represents a significant force for good that is underused.

Project ISLA, a new AI-enabled chatbot by Thoughtworks, in partnership with Amazon Web Services, demonstrates how social service agencies and providers can tackle this with digital transformation.

The chatbot empowers social service organisations to support volunteers at every step of the journey and reduce drop-offs, shared Sarah Sulistio, Product and Innovation Principal – Thoughtworks, and Philip Tran, Partner Solution Architect – AWS.

“We’ve been working with public sector agencies for the last five years to improve their systems and business processes and innovate new solutions to help improve how they serve the needs of the social services sector and their beneficiaries,” said Sulistio. 

During this process, they noticed that various social service organisations struggle to attract, train, and retain potential volunteers, due to inefficient processes and outdated systems for engagement and communication.

To subscribe to the GovInsider bulletin click here.

Simplifying volunteer touchpoints

When a volunteer first begins their onboarding, they often have to read through copious amounts of training materials and attend hours of training. This can look like familiarising themselves with risk assessment frameworks and understanding the proper procedures when helping a family for the first time.

To simplify this process, ISLA can summarise and contextualise these training materials for volunteers.

As ISLA is powered by a large language model (LLM), prospective volunteers can also ask questions in natural language. Then, the chatbot can provide clear, relevant, and concise answers.

Unlike traditional rules-based chatbots, GenAI-powered chatbots can better understand user intent and provide more relevant answers, shared Tran.

The best volunteer can build rapport with beneficiaries and earn their trust – but such skills take time to nurture. To support this journey, volunteers can run simulations with ISLA and roleplay conversations to boost their confidence and strengthen their trust-building skills.

The end of every engagement tends to be paperwork. This includes scribing, condensing, and submitting notes for future review. 

ISLA simplifies this by providing a voice-to-text option that can transcribe and summarise conversations in real time, freeing up time for volunteers to focus on the core mission of serving their beneficiaries.

To subscribe to the GovInsider bulletin click here.

Mapping high-impact use-cases

When developing the chatbot, the Thoughtworks team sought to go beyond just building an assistant that can guide volunteers but also one that could streamline tasks and provide emotional support.

“Our approach to building ISLA was very human centred,” Sulistio said.

Sulistio explained that many Thoughtworks employees volunteer with the Lion Befrienders charity, which supports senior citizens. Through this experience, they were able to map out the volunteer journey from the perspective of both the volunteer and the social service agency.

This helped them identify high impact use cases where AI could immediately reduce friction for volunteers, by mapping common interactions between volunteers, social service agencies, and beneficiaries. 

Through this mapping process, the product team was able to uncover key pain points as well as opportunities for AI intervention.

Developing the right architecture

Once these use-cases were identified, the team had to build the right technical architecture.

“We wanted to mimic that experience of chatting with a social service agent. That means there’s no dead-ends and it doesn’t sound like you’re speaking to a robot,” said Sulistio.

The team leveraged Amazon Bedrock, a fully managed service that enables organisations to tap on high-performing foundation models from leading AI companies such as Anthropic, Meta, Stability AI, and Amazon. 

They decided to use Amazon Titan as their main foundation model, due to the broad range of applications it offered, as well as built-in support for the responsible use of AI.

To prevent the model from hallucinating, or providing incorrect answers, the team implemented retrieval-augmented generation (RAG) technique, which enables the chatbot to provide the right information in response to queries, said Tran.

While the model is optimised for English, it has multilingual support for 100 additional languages as well.

Helping agencies build their own products 

When the team began piloting ISLA, they realized that many social service agencies in Singapore were already developing their own experiments to better support volunteers.

“We started shifting from rolling out ISLA as a product to working with agencies collaboratively to take the approach we took for ISLA to build use-cases,” Sulistio explained.

Throughout the process of building ISLA, Thoughtworks developed the GenAI Digital Product Accelerator framework, which emphasises the importance of using product thinking to develop products that meet customer needs, said Tran.

Now, they have begun running workshops with senior government leaders where they tap on the principles and the approaches that underpin ISLA to adapt to various agencies’ challenges and use-cases. 

These workshops centre around defining key problem statements and how to take a human-in-the-loop approach to designing solutions, Sulistio said. Thus far, they have run over 10 such workshops in Singapore and aim to support more agencies in building solutions to meet their unique needs.