Vinod GC

Ethical Dilemma Situation:
A Customer Service BPO has its existing customers categorized as VVIP, VIP, Important, Regular and Blacklisted. They use an AI agent which is modelled to prioritize incoming calls / text messages of VVIP and VIP customer category and record their queries / complaints while other category customers are subject to long waiting queues.
Dilemma:
VVIP / VIP: This category is kept happy with an intent to maximize revenues and achieve SLA KPI’s.
Others: Unethical as they are subject to long waiting periods even if the nature of the call could be to address an emergency.
As the AI agent follows the applied logic of prioritization, such dilemma situations could arise leading to dissatisfied customer base, reputational damage to the company or even legal disputes.
Approach for guided decision:
1.       Define ethical principles in KB
Create a list of ethical principles aligned to the organizational objectives, add to the internal knowledge base and train the AI model to adhere to the principles. This will make the agent strike a balance between guiding principles and rigid rules.
2.       Weighted priority:
Build in weightages of priorities between customer category and query urgency. This will help allocate a reasonable balance between category and urgency.
3.       Human in the loop (HITL) system:
In case of conflicting situations which the agent is not able to handle, create a mechanism to escalate to human agents for intervention and decision making. The AI agent must be capable of learning and improving referring to the human decision.
4.       Compliance audits:
Conduct periodic compliance audits to review the agent’s decision compliance in accordance with the guiding principles.
Drawing the line:
What should AI do?
Preliminary prioritization, escalate in case of conflict, follow guiding principles, provide explanation for decision and continuously learn & improve.
What should AI not do?
Decision in conflicting situations, override guiding principles.
 
 
 

For this discussion, let us consider the automated process of maintenance request management in a Facilities Management (FM) company which caters to B2B and B2C clients. This process is automated involving multi-agent AI collaboration. Described below, the AI agents that shall be used and their roles.
 
I.            Agents & Roles:
 
1.       Agent A (Conversational AI) - This agent interacts with the client (through omni channel platforms including, web portal, mobile app, chatbots, WhatsApp or voice), gathers all required (defined) variables, understands and categorizes the request.
2.       Agent B (Classification & Prioritization AI) - The role of this agent is to analyze the request considering all input variables and classify the urgency level (Critical / High / Moderate / Low) of the request.
3.       Agent C (Scheduling & Optimizing AI) - Based on the urgency level classified by Agent B, this agent optimizes, and schedules technicians based on their availability, skill and location and communicates available slots back to the client through Agent A.
4.       Agent D (Analyzer AI) - This agent checks, if the asset mentioned in the client request has IoT sensors, gathers log data, fetches historical maintenance records and analyzes them to validate the fault described by the client and possibly identify its root cause(s). The agent also provides fault & potential remedy insights to the technician prior to the site visit.
5.       Agent E (SLA Compliance AI) - The role of this agent is to monitor and track the workflow and escalate risks and potential SLA non-compliances proactively.
6.       Agent F (Feedback AI) - This agent captures client / technician feedback, collates overall workflow performance and feedback insights for other agents to learn and improve their performance continuously.
II.            High-level Workflow:
 


 
 
III.            Potential challenges in coordination between agents:
a.       Conflicts:
Agent C (Scheduling & Optimizing AI) could schedule over / underestimated duration prior to Agent D (Analyzer AI) validating the complaint and finding the root causes. Likewise, Agent B (Classification & Prioritization AI) could misclassify the priority prior to validation by Agent D (Analyzer AI).
b.       Time Delays:
If all technicians are busy and the company doesn't have adequate resources, Agent C (Scheduling & Optimizing AI) could fail to schedule allocation of technician for a critical job leading in delays to addressing the priority.
c.        Data Consistency:
Formats of varied input data used across the Agents must be normalized, else might lead to misinterpretation leading to incorrect agent outputs.
d.       Error Dissemination:
Logical error caused by the agent at any stage in the workflow could have a cascading effect on subsequent decisions and actions.
e.       Explainability:
Both the agents D (Analyzer AI) and E (SLA Compliance AI) must have the capability to explain the rationale behind their findings about the root cause(s) and non-compliance(s) respectively.
 
IV.            Strategies for smooth AI agents' collaboration:
a.       Central Orchestrator AI
Introduce a central workflow manager agent to ensure the workflow progresses in the right sequence with adequate information to resolve conflicts. This will help avoid time delays and avoid conflicts.
b.       Shared Memory
Build a central repository that stores real-time data along the workflow. This helps break data silos.
c.        Explainability
Agents must have the ability to record the rationale behind each action / decision. Based on the flow of work the agent must have the ability to provide real-time alerts such as "the work is delayed due to the complexity of the problem" etc.
d.       Fallback Protocol
Define clear fallback protocols such as escalation mechanisms to alert delays, disputes, SLA noncompliance and unresolved issues.
e.       Secure design:
Firmly controls the exchange of various information from knowledge base(s) and between agents. Map exchange of required information across agents. Doing this shall eliminate conflicting decisions.
 
There could be more strategies applied depending on the type of applications, architecture and technology used, considering their limitations and the application purpose.
 
 

What is trust and why is it crucial in organizational setting:
Trust can be defined as a strong belief in the character, capability, or truth of a person or something. In other words, trust is a byproduct of reliable consistent and transparent behavior of someone or something overtime.
Trust is one of the essential characteristics of a person within any setting, that determines dependability which influences success. Likewise, gaining trust with the introduction of AI agents within organizations is paramount to its success.
 
Scenario:
Let us imagine the Facilities Management (FM) company introduces an AI agent to help its management by analyzing technicians’ performance data and developing performance evaluation reports. The objective is to minimize manual administrative efforts and produce fast and reliable reports.
This is a very sensitive area and hence producing a more consistent, accurate, transparent and unbiased report is paramount to gaining trust from the management and technicians that shall determine success of the rollout.
 
Strategies for creating AI agents that inspire trust:
Clear understanding of the use case (business requirement) and selecting the appropriate AI application / platform incorporating the following capabilities is a core necessity.
 
1.      Incremental progress
Start small by identifying and implementing small portions of evaluation and progress steadily by incrementing scope. This helps to focus and edify the agent to increase accuracy and build trust overtime.
 
2.      Intent clarity
During the initial setup, clearly defining the intent of the application using prompt engineering concepts will eliminate ambiguity. This will lead to the agent providing relevant and contextual insights eliminating errors during the initial stages and moving forward.
 
3.      Data reliability
Ensure to provide and use quality data to train the AI system and develop. Inclusivity of all possible scenarios will help eliminate bias in the system output.
 
4.      Explainability
Develop system capability to be able to clearly explain the inputs, logical steps and the rationale behind the insights. This will help eliminate the black box effect in the AI system.
 
5.      Traceability
Establish AI system performance metrics record logs and action steps to understand where and how the system went wrong. Continuous performance monitoring shall help us understand how the system is performing and maturing over time.
 
6.      Accountability
Decide and communicate through policies, key responsibilities and accountabilities for the inputs, rules, insights and the ones responsible for the overall system performance.
 
7.      Feedback looping
Build capabilities in the system to be able to accept feedback from users, experts and developers for the systems ongoing learning and fine tuning. Effective feedback loops place a significant role in the progressive performance of the AI system.
 
8.      Security
All data used to train, develop and deploy the system must be secure and private to avoid the spread of sensitive information.
 
Following and adhering to these strategies shall be beneficial to building trust over the system and for successful deployment and application.
 
 

Rapid application development abbreviated as RAD, is an agile software development methodology popularly followed in software development. This model was originally conceived in the 1980’s.  A significant benefit of using this methodology is to get a faster project turnaround which makes it an appealing choice for companies and software developers working in a fast-paced environment. Typically, any software development project that can be divided into small modules, and which can be assigned to different developers from different teams can be developed using the RAD model.
At the high level, RAD model contains 4 distinct phases namely;

I.      Requirements planning:
Generically software development starts with a detailed planning phase spending a large amount of time with the users. However, RAD starts by defining a very brief set of requirements, as the methodology allows any change in the requirements during any stage in the development cycle.
 
II.      User description / user design:
This is the important phase that sets RAD methodology apart from other methods where, the product is built through several prototype iterations. During this phase the developers work very closely with the clients to ensure that their requirements are incorporated into the product being developed and the expectations are met. The developer develops the prototype, the client tests it and both the parties get together to discuss what worked and what did not repeatedly until all or most issues are resolved, and requirements met.
 
III.      Construction:
In this phase, the prototype or the beta model is converted into a full-scale working model. As a vast majority of the problems and requirements were addressed during the previous phase, the construction of the product can be quicker. Typically, in this phase, construction preparation, application development, coding and integration & testing activities are done. Even at this phase, clients can suggest changes and modifications or even new ideas to resolve problems.
 
IV.      Cutover:
In this phase, the finished and finalized product is launched which includes data conversion, testing and switching over to the new system. Even during this phase, developers and clients work hand-on-hand to continue to identify and address problems.
 
Reasons for laying emphasis on prototyping than planning:
The prime reason why the RAD approach lays emphasis on prototyping than planning is to accelerate the process of system development. It also allows enhanced flexibility to make any adjustment as required by the clients during any stage of the development process. Prototyping also provides the ability to explore and understand the concepts more quickly. Most importantly it has a huge impact on cost.
 
This table briefs the advantages and disadvantages of the RAD methodology:
 
Advantages
Disadvantages
Overall cost of project is amazingly less as lesser number of developers are required
Involvement of the client during the entire development life cycle is mandatory
Customer feedback is prevalent during all the crucial stages
Managing the project is a bit more complex compared to other development models
It becomes easier to accommodate changes during any stage of the development due to the shorter iterations of prototyping
Projects that cannot be broken down to modules cannot be developed using this approach
Progress of product development is easy to gauge through different stages
May not be suitable for small-scale projects due to the deployment of powerful automated tools and techniques which turns out to be costly
Usage of reusable components significantly reduces the total development time of software projects
Team leaders must coordinate more closely with both the developers and clients to meet deadlines
Better product quality is achieved due to the usage of powerful development tools
Usage of powerful tools necessitates involvement of highly skilled professionals
 
 
Is this approach suitable outside software development sector?
Yes, this approach is suitable and can be used widely in various sectors other than software development. This approach can be used in scenarios when a project needs to be done quickly. Upon research we find that the application of RAD approach has been successful and beneficial in the Banking, Financial Services and Insurance (BFSI), Automobile and Online Retail industries.
 

 
With the growth in technology and usage of internet, there is a huge increase in customers opting for mobile banking platforms for financial transactions. With such a demand, BFSI institutions must move onto the mobile space to keep up with the demand. In order to meet this expectation from clients, a lot of solutions are developed and provided using the RAD approach.
 

 
The rapid prototyping a derivative of RAD is widely used in the automobile industry in developing physical prototypes and scaled models for their designs. Thanks to the 3D printing technology that makes the process efficient and effective for developing the prototypes at the required pace. These prototypes are subject to various experiments to identify issues / flaws and discover areas of improvement prior to commercial production.
 

 
Online retailing was getting popular even before the pandemic and has significantly grown and will grow in the future. In 2018, it was found that around 40% of the e-commerce transactions were using mobile devices. This has grown upto 54% in 2021 which describes the growing demand and requirement of mobile e-commerce applications. With the application of the RAD approach, retailers can develop and put to use simple, secure mobile applications in the shortest possible time to meet the demand.
 
 
 
 

Prior to comparing RAPID and DACI matrices with RACI matrix, let us briefly understand what an RACI matrix is. RACI is a short form for Responsible, Accountable, Consulted and Informed. It is a tool quiet extensively used in project management to define individuals’ roles and responsibilities in the execution of projects while removing ambiguities and provides clarity to all stakeholders. This topic is discussed in detail in the “benchmark six sigma forum” and can be accessed using this link.

 
What is RAPID matrix?
RAPID is short for Recommend, Agree, Perform, Input and Decide. It is yet another tool used in project management. While RACI matrix is used to define high level roles and responsibilities, RAPID matrix is used to define more lower-level project planning and decision-making responsibilities.

RAPID matrix example

 
What is DACI matrix?
A DACI matrix, expanded as Driver, Approver, Contributor and Informed is also a tool used in project management to define / identify key roles and responsibilities for making group decisions for every major task within a project.

DACI matrix example

Comparison between RACI, RAPID and DACI matrices:
Matrix
RACI
RAPID
DACI
Expanded as
Responsible, Accountable, Consulted, Informed
Recommend, Agree, Perform, Input, Decide
Driver, Approver, Contributor, Informer
Year
1970
After 1980
1980
Founders
Kristoffer Grude, Tor Haug and Erling Andersen
Bain & Company
Intuit
Application
Tool used to clarify & define roles & responsibilities to people within a project / team.
Tool used to agree and assign decision making roles and accountabilities.
A variant of RACI used to assign roles to personnel who drive projects to conclusion.
Merits
þ  Well known and used throughout organizations
þ  Once you assign the ‘A’ person, leave it to the team to make it work
þ  Preferred by 3rd party companies (e.g. consulting companies) as it has a specific person (A role) to deal with
þ  Enforce a recommendation or a decision, which is typically the blocker
þ  Role allocation is clear to manage the activities from beginning to end (entire lifecycle)
þ  Preferred by teams that want to get things done
þ  Eliminates ambiguities by giving full authority over definite aspects of decision making to a specific person
þ  Representing every aspect of decision-making with a specific role
þ  Removes collective responsibility and corrects team imbalance by giving authority to a single person / role
þ  Eliminate probable disagreements and speeds up decision-making
Demerits
ý  Everything is dependent on ‘A’ role
ý  No strong focus on teamwork
ý  Confusion may arise between Consult / Inform roles & Responsible / Accountable roles
ý  May lead to shortcut and not well-thought through options
ý  Not easy to implement as everything is not just link to one person
ý  Trademark limits use due to legal concerns
ý  Model only works if the rest of the team members are prepared to accept the decision made by the approver