Stage 2: build
This module helps business leaders, project managers, and team leaders select suitable technology focusing on the role of human oversight in your automation solutions.
Evaluating technology considerations
Stage 2: build, involves planning and designing the chosen automation use-case in detail, with the objective of producing an implementation plan that includes the technical architecture, workflow, and user interfaces or applications designs. It is critical to focus on the suitability of the chosen technology to the automation processes. The automation team must evaluate the technology capability suitable for each use-case in the pipeline and engage enterprise architects to determine the optimal technology. Standardised processes, or those with predictability and consistency are good candidates for automation.
Selecting automation technology
Choosing between RPA and IA depends on your organisation’s specific automation needs. Consider the nature of the error-prone processes within an organisation and process characteristics. RPA is valuable for automating straightforward, repetitive, rule-based tasks. It is less effective in handling error-prone processes without human intervention. However, IA brings advanced capabilities like machine learning and artificial intelligence, making it better suited for error detection, handling, and prevention in complex, dynamic tasks, and processes.
AI models should be designed and developed in a way that enables continuous monitoring and auditing to ensure adherence to AI Ethics framework.2 Use development best practices to ensure that AI models maintain transparency, reliability, compliance, accountability, and fairness.
Utilise the Artificial Intelligence Assessment Framework (AIAF) to ensure risk and impact assessments in designing and developing any AI component.3 To ensure accountability, justification and contestability, document in detail, every aspect of an AI component’s development including:
- roles and responsibilities
- development process
- data sources
- model architecture
- decision-making process
- results.
Expand the accordion below, to distinguish between processes suitable for either RPA or IA based on their use-case characteristics.20,21,22,23,24
| Characteristics | RPA suitable | IA suitable |
|---|---|---|
| Process level | Task, sub-process level | E2E process level |
| Cognitive ability | No, repetitive, rule-based, and high volume of tasks | Yes, decision-making, or human judgement |
| Volume or occurrences | Medium to high | Low to medium |
| Ambiguity | No ambiguity | Ambiguous |
| Predictability and stability | Highly predictable/unchangeable | Unpredictable and requires cognitive abilities |
| Complexity | Low to moderate | Moderate to high |
| Exception management | Not capable | Capable |
| Error-prone scenarios | Prevent human errors yet cannot correct errors | Detect, anticipate, and proactively fix |
| Adapt and learn | Not capable | Capable |
| Data handling | Structured data | Unstructured data |
| Suitable run time | In/out of working hour | In working hour |
| Human-in-the-loop | Zero-to-low intervention | Human-bot collaboration |
| Actions involved |
|
|
| Technology complexity | Relatively simple | Relatively complex |
| Users’ skill level | Relatively low to moderate | Relatively moderate to high |
| Governance required | Low | High |
| Implementation timeframe | Quick, non-invasive integration | Slow, invasive integration |
| Cost considerations | License and development efforts. | AI training, data preparation and integration, and skills acquisition. |
Human-in-the-Loop (HITL)
It is critical to consider the role of human oversight in your solution design. Human-in-the-loop (HITL) involves integrating human interaction, judgment, and oversight in automated systems. It capitalises on the unique strengths of our human cognition, such as contextual understanding, ethical decision-making, legal compliance, and adaptability to complex scenarios.
Decisions in government organisations often have profound societal implications, and fostering community trust is critical. HITL provides accountability, regulatory compliance, and adherence to ethical standards essential to building and maintaining trust in government. It also provides a contingency plan in case of technology failure.
By incorporating the HITL model during your automation design phase, government can build systems that strike a harmonious balance between the efficiency of automation and the indispensable unique qualities of human judgement. HITL supports effective, trustworthy, and citizen-centric government services.
Human value in automation use-cases
Expand the accordions below to learn about the human value in automation use-cases.
Technology
RPA and IA.
Use-case
Automated systems that involve critical decision-making, such as in healthcare diagnosis, legal proceedings, or financial decisions, must consider incorporating HITL.
Human value
Humans bring ethical considerations, empathy, and contextual understanding to complex decisions, ensuring that the outcomes align with community values and legal standards.
Technology
RPA and IA.
Use-case
Exception handling scenarios, such as unexpected events or anomalies in data.
Human value
Humans excel at handling edge cases, bringing adaptability and creativity to situations that may not be explicitly covered by automation technology.
Technology
RPA and IA.
Use-case
Systems that involve ethical considerations, particularly in fields like AI and machine learning where biases might emerge and must be carefully managed.
Human value
Humans can evaluate decisions from an ethical standpoint, ensuring fairness, transparency, and adherence to ethical guidelines.
Technology
IA.
Use-case
Automated systems that interact directly with end-users, such as customer service chatbots or AI-driven interfaces.
Human value
HITL fosters user trust by providing a mechanism for users to understand decisions made by automated systems and to challenge or seek clarification when needed.
Technology
RPA and IA.
Use-case
Processes where adherence to legal and regulatory requirements is vital, such as data handling related to privacy and human rights.
Human value
Incorporating human oversight ensures that automated processes align with legal standards, protecting privacy and human rights, and reducing the risk of non-compliance.
Technology
IA.
Use-case
Situations where the context is subject to rapid changes or evolving conditions.
Human value
Humans are adept at adapting to unforeseen changes, ensuring that decisions remain relevant and appropriate in dynamic environments.
Technology
IA.
Use-case
Tasks that require creative thinking or innovative problem-solving.
Human value
Human ingenuity is crucial in situations where solutions may not be readily apparent or where a novel approach is needed.
Technology
RPA.
Use-case
Automated processes that may have errors or inaccuracies.
Human value
Human validation helps identify and correct errors, ensuring the accuracy and reliability of automated systems.
Technology
IA.
Use-case
Decisions involving complex social, cultural, or emotional factors.
Human value
Humans can better understand and navigate nuances in decision-making, especially in contexts where emotional intelligence is essential.
Technology
IA.
Use-case
Data labelling and model training in machine learning.
Human value
Human involvement creates high-quality training datasets and improves the performance of machine learning models.
Next lifecycle stage
Stage 3: run
Operationalise the implementation plan after rigorous testing then deploy it into production environment.