Opinion

Start small and start right is the key to enterprise AI success

Amitabh Mudaliar | Updated on February 25, 2021

Any organisation that wants to implement AI on a larger scale should have a robust data pipeline that is derived from the large quantity of unstructured data that organisations have.

Artificial Intelligence (AI) has the potential to completely transform the CPG (consumer packaged goods) industry from what we see currently. Predictive analytics, recommendation engines, natural language processing, conversational interfaces, cognitive computing, and computer vision are some of the AI-based technologies which can have many applications across the CPG value chain.

Every firm around the globe is experimenting with AI and many pilots and proof of concepts are being tried out in small pockets. For example, an upcoming CPG firm can use AI to inform its business decisions in deciding the next product or details of the packaging. The firm can use sentiment analysis to understand what its consumers think across the social media landscape and further use it to analyse its strength and weaknesses. AI can be used to streamline supply chain functions by tracking the movement of goods. It can also be used to personalise the online shopping experience based on preferences.

The above scenario just demonstrates how AI can transform the operations of a CPG business to be more efficient. Yet this has shown few good results in terms of enterprise-level implementations. There are a lot of reasons why many large-scale AI-based implementations fail. The right implementation of AI has brought in immense benefits in terms of optimisation, automation, and efficiency to organisations. However, these success stories have been far and few in between.

There are many reasons why most AI at scale projects fail and it is important to understand how to make the enterprise AI imperative in your organisation a success. It could be identifying the wrong problems to adopting the wrong use case to leveraging a bad data set to using an inefficient training model. And in a post-Covid world the need for an AI-driven organisation will be more pertinent than ever. Any organisation that exists would sooner or later would have to be an AI organisation which is continuously deriving insights from their data, optimising processes, deriving efficiency at every level or risk losing out to their competitors. Let us dive into examining factors that make AI successful in an enterprise.

Steps for success

The foremost step for a successful AI project is to identify and solve the problem with the apt AI technique and data. AI revolution was preceded by the data revolution and the reason for this is obvious: for a successful AI program, you need the right data, and a large quantity of it. A lot of effort goes into cleaning and organising the data pipeline to render it useful for AI implementations.

Any organisation that wants to implement AI on a larger scale should have a robust data pipeline that is derived from the large quantity of unstructured data that organisations have. This can be plugged directly into AI systems to curate and cleanse AI-ready data. It is important to use appropriate predictive analytics models and algorithm to reap full benefits of AI.

The final yet the most important element in successful AI projects is having the right resources. Even among those who claim to be qualified, there is a sore lack of data, modelling, and AI programming skills. Having a talented team who can contribute positively to the entire AI journey is of utmost importance.

Another key factor that is emerging and will be crucial in AI adoption in the future is the explain-ability of AI algorithms and models. Currently, AI models are a black box with very little to no explanations of the results produced which is leading to mistrust of AI outcomes. Organisations should start considering the incorporation of explainable AI models/software in their existing AI infrastructure.

An explainable AI model explains the factors and variables that go into an AI decision thereby making the entire process transparent. It is the moral responsibility of an organisation that the AI models deployed in their digital landscape are unbiased, transparent, and accountable. This will lead to easy and widespread adoption of AI among its customers and stakeholders. Although explainable AI is still a nascent field, a concerted effort should already be underway in this direction if they want to use AI on a large scale.

When used in the right way, AI can deliver personalised recommendations to consumers, transform customer journeys, predict outcomes, augment inventory management, optimise functions, and streamline logistics processes across the CPG industry. In addition to AI, natural language processing can be leveraged to aid customers through digital assistants and chatbots. Intelligent automation used in conjunction with rule-based automation can transform the landscape for robotic process automation. AI can disrupt the way you do things but to make it successful start small with the right success factors and then go big. It suffices to say AI can disrupt the way you do things but to make it successful start small with the right success factors and then go big.

The writer is VP and Region Head, Retail, CPG, Logistics, Infosys

Published on February 25, 2021

Follow us on Telegram, Facebook, Twitter, Instagram, YouTube and Linkedin. You can also download our Android App or IOS App.

This article is closed for comments.
Please Email the Editor