According to McKinsey, artificial intelligence (AI) could add around $13 trillion to the global economy by 2030. While the excitement around AI is palpable, it’s essential to discern where its implementation can be genuinely beneficial. Drawing lessons from blockchain implementations, it becomes clear that the solution should align with the problem. AI is a powerful tool, but it’s essential not to overestimate its capabilities. A clear understanding of AI’s potential sets the foundation for a realistic and effective implementation strategy. We can’t force-fit AI solutions where they have no utility.
To successfully implement AI projects, it is crucial to follow a strategic approach. Firstly, rigorous technical and business due diligence is necessary. The proposed AI project must be technically feasible, supported by the availability of pertinent data and skilled Software/Machine Learning engineers. Furthermore, the AI system should genuinely contribute value to the business. There should be clarity on what AI can or cannot do. According to Andrew Ng, a pioneer in AI and a professor at Stanford University, a general guideline is that if a task typically takes a second or so for humans to complete, it possesses substantial potential for AI application. For instance, AI can effectively classify restaurant reviews as positive, negative or neutral or identify objects like cars and pedestrians.
Also Read
Learnings from 2023 and trends in 2024
Secondly, it is important to understand the difference between Machine Learning and Data Science. Machine learning and data science are often used interchangeably, but they serve distinct purposes. Machine learning primarily is used for automation, while data science delves into deciphering patterns and insights within vast datasets. Acknowledging this difference is crucial for a nuanced approach to AI implementation.
Also Read Cycle of dependency: The desperate yearning for more has consumed many businessmen like Naresh Goyal Poor lessons: Through its dodgy accounting, Byju’s has queered the pitch for India’s start-ups Macron in India: New ties for renewables Evaluating Mintzberg’s 10 schools of thoughts for strategy formulation Thirdly, it is imperative to grasp that AI doesn’t automate jobs but tasks within jobs. A job is made up of several tasks. Factors such as complexity, repetition and the amount of data involved should be considered in recognizing AI potential of a task. For example the job of a customer service representative is composed of tasks such as answering inbound phone calls from customers, answering customer chat queries, checking status of customer orders and keeping records of customer interactions. While the second and fourth tasks have high automation potential, the other two may not be the ideal candidates. Therefore, it is advised to break down a job into tasks and assign them a rating (high, medium and low) based on their AI potential. This granular approach helps in identifying specific areas where AI can optimize processes and enhance overall productivity.
Generative AI based on Large Language Models is a sub-category of AI which is finding a lot of applications in organisations. They can generate high-quality text, images, speech, and codes, drastically reducing the time required for specific tasks. For example, implementing an AI project without LLMs might take several months, whereas with them, it can be accomplished in days.
There can be several applications of Generative AI. It can assist a surgeon in doing research on medical procedure faster so that the surgeon can devote more time to the surgical procedure. Similarly, it can be useful in the process of legal documentation review by reducing the time taken to gather information and review documents. Generative AI tools can also help marketers to write multiple versions of a website copy and test them online.
The implementation of AI often necessitates workflow changes and process re-engineering, potentially altering how tasks are executed within an organization. Therefore, beyond task automation, AI also has the capacity to augment output quality and reduce processing time by reshaping workflows.
Starting with a small pilot project can be a good way to understand AI systems and also demonstrate its credibility to the organisation at large. Building trust through small wins is crucial for broader acceptance. The success of a small project may be more important than the goals it is intended to achieve.
Assembling a team with diverse expertise is crucial for AI initiatives. A well-rounded team should include data scientists, machine learning engineers, domain experts and project managers. To begin with a team of one or two members can also be sufficient. It is important to regularly solicit feedback from all stakeholders involved in the AI implementation process. This inclusive approach helps in addressing concerns, refining strategies and ensuring that the integration aligns with the organization’s overarching goals.
Beyond traditional AI techniques, advanced methods like fine-tuning and Retrieval-augmented generation (RAG) can be valuable. Fine-tuning assists LLMs by providing additional relevant data for better understanding of the context. For example an LLM can be shown how to write text in the legal language by giving it access to a sample legal document. The LLM then uses this context-rich knowledge to generate more accurate and contextually fitting responses. RAG improves the accuracy of LLMs by providing them with access to external knowledge bases. For example, if a user asks an LLM “What is the capital of France?”, the LLM might generate the response “Paris”. However, if the LLM is using RAG, it can first retrieve the information from an external knowledge base that the capital of France is Paris. This allows the LLM to generate a more accurate and informative response to the user’s query.
Implementing AI in an organisation is a strategic endeavour that requires careful planning and execution. By understanding AI’s capabilities, identifying suitable tasks, utilizing advanced tools, and fostering a collaborative team environment, organizations can navigate the AI landscape successfully. Start small, learn from pilot projects, and continuously refine your approach to unlock the full potential of AI within your organizational framework.
Ravi Singh – Author and IRS officer. Views are personal
According to McKinsey, artificial intelligence (AI) could add around $13 trillion to the global economy by 2030. While the excitement around AI is palpable, it’s essential to discern where its implementation can be genuinely beneficial. Drawing lessons from blockchain implementations, it becomes clear that the solution should align with the problem. AI is a powerful tool, but it’s essential not to overestimate its capabilities. A clear understanding of AI’s potential sets the foundation for a realistic and effective implementation strategy. We can’t force-fit AI solutions where they have no utility.
To successfully implement AI projects, it is crucial to follow a strategic approach. Firstly, rigorous technical and business due diligence is necessary. The proposed AI project must be technically feasible, supported by the availability of pertinent data and skilled Software/Machine Learning engineers. Furthermore, the AI system should genuinely contribute value to the business. There should be clarity on what AI can or cannot do. According to Andrew Ng, a pioneer in AI and a professor at Stanford University, a general guideline is that if a task typically takes a second or so for humans to complete, it possesses substantial potential for AI application. For instance, AI can effectively classify restaurant reviews as positive, negative or neutral or identify objects like cars and pedestrians.
Secondly, it is important to understand the difference between Machine Learning and Data Science. Machine learning and data science are often used interchangeably, but they serve distinct purposes. Machine learning primarily is used for automation, while data science delves into deciphering patterns and insights within vast datasets. Acknowledging this difference is crucial for a nuanced approach to AI implementation.
Thirdly, it is imperative to grasp that AI doesn’t automate jobs but tasks within jobs. A job is made up of several tasks. Factors such as complexity, repetition and the amount of data involved should be considered in recognizing AI potential of a task. For example the job of a customer service representative is composed of tasks such as answering inbound phone calls from customers, answering customer chat queries, checking status of customer orders and keeping records of customer interactions. While the second and fourth tasks have high automation potential, the other two may not be the ideal candidates. Therefore, it is advised to break down a job into tasks and assign them a rating (high, medium and low) based on their AI potential. This granular approach helps in identifying specific areas where AI can optimize processes and enhance overall productivity.
Generative AI based on Large Language Models is a sub-category of AI which is finding a lot of applications in organisations. They can generate high-quality text, images, speech, and codes, drastically reducing the time required for specific tasks. For example, implementing an AI project without LLMs might take several months, whereas with them, it can be accomplished in days.
There can be several applications of Generative AI. It can assist a surgeon in doing research on medical procedure faster so that the surgeon can devote more time to the surgical procedure. Similarly, it can be useful in the process of legal documentation review by reducing the time taken to gather information and review documents. Generative AI tools can also help marketers to write multiple versions of a website copy and test them online.
The implementation of AI often necessitates workflow changes and process re-engineering, potentially altering how tasks are executed within an organization. Therefore, beyond task automation, AI also has the capacity to augment output quality and reduce processing time by reshaping workflows.
Starting with a small pilot project can be a good way to understand AI systems and also demonstrate its credibility to the organisation at large. Building trust through small wins is crucial for broader acceptance. The success of a small project may be more important than the goals it is intended to achieve.
Assembling a team with diverse expertise is crucial for AI initiatives. A well-rounded team should include data scientists, machine learning engineers, domain experts and project managers. To begin with a team of one or two members can also be sufficient. It is important to regularly solicit feedback from all stakeholders involved in the AI implementation process. This inclusive approach helps in addressing concerns, refining strategies and ensuring that the integration aligns with the organization’s overarching goals.
Beyond traditional AI techniques, advanced methods like fine-tuning and Retrieval-augmented generation (RAG) can be valuable. Fine-tuning assists LLMs by providing additional relevant data for better understanding of the context. For example an LLM can be shown how to write text in the legal language by giving it access to a sample legal document. The LLM then uses this context-rich knowledge to generate more accurate and contextually fitting responses. RAG improves the accuracy of LLMs by providing them with access to external knowledge bases. For example, if a user asks an LLM “What is the capital of France?”, the LLM might generate the response “Paris”. However, if the LLM is using RAG, it can first retrieve the information from an external knowledge base that the capital of France is Paris. This allows the LLM to generate a more accurate and informative response to the user’s query.
Implementing AI in an organisation is a strategic endeavour that requires careful planning and execution. By understanding AI’s capabilities, identifying suitable tasks, utilizing advanced tools, and fostering a collaborative team environment, organizations can navigate the AI landscape successfully. Start small, learn from pilot projects, and continuously refine your approach to unlock the full potential of AI within your organizational framework.
Ravi Singh – Author and IRS officer. Views are personal
Get live Share Market updates, Stock Market Quotes, and the latest India News and business news on Financial Express. Download the Financial Express App for the latest finance news.
login
who are we?
contact us
qosheapp
Toi Staff
Gideon Levy
Belen Fernandez
Rami G Khouri
Mort Laitner
Donald Low
Ali Fathollah-Nejad
Nikkei Editorial
Michael Kwet
visit website
