Mike

Evaporating Cloud Technique is a conflict resolution technique that is one of the processes of Theory of Constraints. It was introduced by Dr. Eliyahu Goldratt to solve difficult problems that have no clear solution. It resolves the conflict by stating both sides of the problem and reviewing the underlying assumptions that create the conflict. The EC diagram depicting the current problem statement is shown below.,

Here the conflict arises from two opposing needs, but the goal is the same to improve overall performance
NEED A: To deploy customer service chatbot to reduce cost immediately
NEED B: To delay deployment so that a well trained chatbot that gives high quality responses is deployed
 
Let us now state the underlying assumptions.,
a)      Deploying the bot quickly will save cost but stand the risk of low customer experience
b)      Delayed deployment of bot due to training ensures high quality responses but does not give immediate saving
c)      Reducing cost and getting high quality responses are mutually exclusive
d)      If chatbot is deployed it cannot learn quickly and improve on quality
We need to a challenge the above assumptions to find a win-win solution. Challenging the above we see that we can build a solution that will address both goals.
1)      To mitigate issues caused by delayed deployment and to also ensure high quality response, we can deploy the bot in phased manner to address such simpler queries to start off with and parallelly train the bot on much complex queries as it continuously learns over time. This ensures that quality is not compromised, and immediate savings are also realised
2)      We can also do a Pilot testing by deploying the bot for limited section of the customers thereby avoiding potential risks or at least minimizing it. This will help improve quality in the long run before full deployment and save cost immediately
3)      We can also use Human in loop to answer complex queries and let the chatbot handle much simpler queries ensuring customer trust
4)      We can also use the feedback loop post launch and improve the chatbot performance over a period making it handle simple and complex queries. This ensures continuous or real time learning
Hence the company can design a Phased and iterative solution to save reduce costs and sustain acceptable quality. This approach stated above will bring in a win-win situation to the company ensuring benefits in terms of reduced cost in the short run and better customer experience in the long run.
 
 

AI Hallucinations occur when LLM’s make up answers that are not factual. This mostly occurs when the model tries to be extra helpful by fabrication information when the model does not have a factual backing for the data presented. This also occurs when the model has not been trained on that topic.
Prompt engineering techniques focus on guiding models to produce truthful or factual data for any query. By refining these prompts, we can minimize the number of hallucinations produced by AI and increase the overall performance of the AI system. Some of the practices we can follow to reduce hallucinations ae listed below.,
 
1)      Give clear and precise prompts avoiding ambiguity that may provide irrelevant output
2)      Explicitly state the context with some additional background information
3)      Break down the tasks into smaller manageable steps ensuring that AI models understand it easily
4)      If need arises specify the output format required for the answer
5)      Incorporate examples to get the desired output
6)      Ask the model to differentiate between factual vs speculative answers
7)      Ask the model to prove the references for the source of the information and prove its limits or in simpler terms ask the model to validate its own responses
8)     Usage of domain specific terminology also helps to align responses reducing nonsensical information
 
When working with highly specialized domains, we can use some of the below mentioned methods to reduce hallucination.,
 
1)      RAG – Retrieval Augmented Generation: This method combines the model’s capability to generate information with reliable information sources like knowledge bases.
Example: Who is called the “Father of the nation” in India? This will make the model look up information in the knowledge base before generating a response. Likewise, if any question is asked pertaining to a specific domain the model will investigate the relevant knowledge bases before responding
2)      ReAct Prompting – Reasoning + Acting: This method is used reduce hallucination by incorporating a step-by-step reasoning and an external tool such as an API to ensure that the responses are grounded in factual data.
Example: In medical diagnosis, instead of asking for symptoms of a disease, we break down the question into smaller steps and narrow down to the specifics based on reliable medical data and then ask it to respond with prognosis
 
There are several other methods to reduce the impact of hallucination in generative AI but there is still a long way to go to build trustworthy intelligence as the knowledge bases are vast and susceptibility to nonfactual information which is a reality today.
 

Let’s understand what a knowledge management system is. Knowledge management system (KMS) can be seen as an entity like an IT system in an organization that implements knowledge management. It will deal with organising, storing and retrieving a lot of knowledge relevant to the various processes within the organization. It ensures that the necessary information is available on demand and can be used for process guidance.
Well, what is a knowledge base then? Knowledge base (KB) can be defined as a centralized repository of information that relate to policies and procedures. It also helps you create, store and share documents across your company.  It is made available to the employees for easy access in case any guidance is needed.
Now what are the common reason for a KB becomes unusable in a KMS? Some of the common reason that contribute to KB becoming unusable are listed below.
1)      Poor user experience
2)      Incomplete knowledge base
3)      Complex knowledge base
4)      Low user adoption
5)      Easy access
6)      Poor structure
7)      Outdated content
😎     Security issues
9)      Scalability issues
How do we ensure that these reasons are mitigated? What are some of the strategies that can be employed to improve the usability of the knowledge base are listed below.,
1)      Structuring a knowledge base – Using a proper hierarchy, tagging and linking content
2)      Optimizing a knowledge base – Employing proper search engines
3)      Knowing your audience – Understanding user groups and their search patterns
4)      Promoting the knowledge base – Ensuring that the KB is promoted to resolve FAQ’s and Complex searches
5)      Collecting feedback – User feedback on the KB usability and its effectiveness
6)      Scalability ready – Easy Integration with advanced technologies
7)      Easy Transition – Version control adaptations
😎     Broadcasting Updates – Ensure all users are notified about updates through various channels (Chat, Email, etc.,)
We can ensure that the usability and effectiveness of a KB can be drastically improved if the above are adopted.
With the future of technology focussed on Artificial Intelligence, Knowledge bases can be used extensively with AI to increase reachability, robustness, accuracy and effectiveness. In terms of Organizing the KB, AI can help in streamlining content. AI can use NLP to understand and interpret the user’s intent behind the queries asked. AI can help in organizing information effectively by employing Hierarchical structuring, identifying metadata and tagging and help in content linking. In terms of efficient retrieval of information, AI can help in optimization in searches. This enhances performance and improves user experience. AI also offers multilingual support. AI can be used to take the existing SOP’s into near accurate decision trees by analysing data, identifying patterns and generating new content. AI enhances data security. Bringing in AI allows us to incorporate Analytics of the highest order giving us insights on user interactions, common queries (FAQ’s) and thereby system performance. Over time this will lead to better and accurate answers.
As a closing statement, while a knowledge base as an entity is susceptible to failure, Artificial Intelligence can bring in higher robustness, greater effectiveness and a level of security to ensure that the Knowledge base remains usable and relevant across time.