Global Forum For Indurstrial Devlopment is SPONSORED BY ICO INDIA
- MP Society Registration (Act. 1973 No. 44) 03/27/01/21857/19 (MSME Forum Established Since-2009)
India is home to about 55-60 million MSMEs (Micro, Small and Medium Enterprises). MSMEs play a lead role in generating employment and contributing to India’s GDP. But MSME’s lack access to formal credit. As a result, many of these companies operate well below their potential.
Banks and other financial institutions often deploy manual processes for sourcing, underwriting and servicing credit. These firms are therefore unable to service the credit demand that MSMEs have. Reports estimate that the MSME credit demand-supply gap to be about USD 350 Bn.
The proliferation of cashless transactions, GST and Smartphone penetration have built an alternate data trail for the otherwise thin-disclosure/filed MSME sector.
In this webinar, Shalabh Singhal, CFA, Co-Founder, ZipLoan will discuss how Digital MSME Lenders are using alternate data and ML(machine learning) to solve the “low credit supply to MSMEs” problem that India has.
The webinar will address the following learning outcomes. Understanding:
1. Size of the Digital MSME Lending opportunity in India and its impact on overall economy
2. Key enablers for the Digital MSME Lending ecosystem
3. Key Machine Learning applications (credit decisioning and beyond) in Digital MSME Lending
4. Popular Machine Learning models in Digital MSME Lending
5. Key risks and control measures in ML powered Digital MSME Lending
Picture this: Your marketing software provides actionable guidance on writing innovative, personalized content that will steadily improve conversion rates. Your employees are more engaged and happier, spending less time working on tedious, repetitive tasks such as summarizing articles, tagging content and adding descriptive metadata to images in the CMS. Instead, they are working on creative marketing campaigns and producing fresh, well-received content. Your teams are more productive, translating to company-wide efficiency.
This scenario is an emerging reality thanks to the practical application of AI in digital marketing. There is no shortage of hype describing how AI will fundamentally change our work and lives in the future. When it comes to AI, it can be hard to separate fact from fiction, and in many cases marketing buzz from practical applications. To understand how AI-powered technologies can impact your business, it’s first essential to understand how the technology is being used today to deliver value.
Marketing technology has been evolving rapidly over the past decade. We have seen Content Management Systems transform into full-featured Customer Experience Platforms. Alongside this platform growth, we have seen an explosion of marketing point solutions. This sheer number of marketing systems has led to a subsequent focus on feature and data interoperability. If you have an email system, a CMS, and a CRM, it’s now much easier to pull that data together, and then to find correlations and insights that can help your business. At the same time computing has advanced to the point where the algorithms and computing power needed to run neural networks or crunch machine learning models are available to any business. There is a confluence here of these two pieces — the affordability of computing power intersecting with massive amounts of data.
Top tech companies including Microsoft, IBM, Alphabet (Google’s parent), and Amazon (just to name a few), are in a race to develop publicly-available cognitive APIs. This category of intelligent services is collectively referred to as Machine Learning as a Service (MLaaS). It’s worth noting that according to Stratistics MRC (global market research company), the MLaaS market is expected to grow to 7.6 billion dollars by 2023. These companies invest in large teams of top computer scientists, linguists, and data scientists, and the work they do provides massive value to all of their customers, allowing other companies to tap into this expertise without the cost of developing the tools themselves.
The massive tech companies are not only racing to be first-to-market to provide these services, but they also have tremendous use for these technologies internally for their operations, not to overlook the immense potential of not-yet-realized applications and enterprises. This continuing trend of ubiquitous and inexpensive cognitive services is spurring adoption, allowing software providers, systems integrators, and internal development teams to infuse advanced AI-powered features into their software for a fraction of the investment it would have taken even five years ago.
While I’m guilty of using ML and AI interchangeably, it’s important to note the distinctions between them. AI is a general classification of programming machines to display characteristics of intelligent behavior or to “think.” Machine learning is a subset of AI in which a system automatically learns and improves with access to data and without explicit programming. The more data provided, the more the system improves performing specific tasks.
Machine learning and AI more broadly have the potential to provide exponential value to businesses over the long-term, but there is no lack of benefits in the short-term. Any organization with departments of content authors who produce large volumes of content (publishing, media, B2B, nonprofits) can benefit from the efficiency of being able to automate a variety of frequent tasks. In the content management space, this includes automated tagging of images and text or the creation of text summaries.
As far back as several years ago, several media giants including the Associated Press (AP), Fox News, and Yahoo were already using AI-powered software to generate stories and recaps. Though none of this works entirely without human involvement, the tasks AI-powered software can cut down on eliminates employee time spent on such work, availing them to take on higher-value tasks. AI is best viewed as a tool for enhancing human capabilities instead of replacing humans for the time being.
Another area where machine learning and AI are compelling is in the automation of analytical activities such as segmentation, optimization, and predictive modeling. The race to deliver this capability in off- the-shelf products and services offers the potential of unlocking customer insights that would not have been previously possible without a small army of data scientists.
Recommendation engines are a great example of predictive analytics and one which we are all familiar with through the casual use of popular services. Companies like Amazon and Netflix invest heavily and train machine learning and AI models to make surprisingly relevant and accurate predictions. They have no shortage of data to pull from, given the number of variables that can be tagged, and the various dimensions that can be categorized.
Netflix, for example, can make recommendations based on the description of the content itself (content-based filtering) but also on a user’s previously viewed titles in relation to his or her similarity to other users (collaborative filtering). The combination of these two models known as the hybrid recommendation system is prevalent, and in the case of Netflix, allows the company to provide their customers with personalized content recommendations.
One downside of these technologies being so easy to adopt is that it becomes easier to fall into the trap of selecting a technology in search of a problem. It’s essential to focus first on the business and use case that will derive the most benefit, and then enlist the right technology to solve that problem. The list below describes the categories of machine learning services that are ready for use today, and potential business use cases that they are particularly well suited to solve.
The burgeoning field of computer vision includes image, object, and facial recognition. It’s possible to use inexpensive APIs to derive information such as the topic, a text summary, recognized faces, emotion or sentiment, inappropriate or lewd content, or the sex and age of individuals in the photo. If your digital media library has thousands or tens of thousands of images, this can significantly ease manual maintenance of the library, and also provide more helpful ways to search your date when retrieving the content. For example, you could easily find photos of a daytime city street with happy people walking down it.
Natural Language Processing (NLP) is a branch of AI in which systems are developed to understand, interpret, and manipulate human language. Deriving concept or intent of a sentence when accounting for the ambiguities of language is a complicated problem. While this field has existed for over 40 years, factors such as a demand for stronger customer experience through human-machine interaction, advances in computing power, and access to large sets of language data has resulted in a massive boon to the field. Today this area is practically applied to voice interfaces (Alexa), chat interfaces (chatbots), and text mining applications like social media sentiment analysis or comment moderation.
Many of our clients have vast reserves of content that have been meticulously tagged by humans for decades, which is a perfect data set to train a new predictive model. Right now, we are teaching Watson to auto-tag new articles based on a library of previously tagged content. We are also using off-the-shelf models from Microsoft and others to tag content with the sentiment, tone, grade-level, and other aspects relevant to content writing. This application is especially suitable for media or publishing clients.
Given the vast stores of data many companies have, machine learning and AI-powered services offer a powerful capability to surface actionable insights. By feeding data describing your content or products along with analytics collected from numerous sources and touchpoints with customers, it’s possible to mine the data for correlations and patterns. The system scours the data to discover new customer segments, valuable customer journeys, or recommend content that is more likely to lead to the desired conversion. Amazon, Microsoft, and IBM are deploying a quickly-expanding toolset that allows businesses to integrate with this technology without having to learn and master complex ML algorithms. With the right data and a moderate amount of knowledge about machine learning, you can start building recommendation systems powered by similar technologies that Netflix or Amazon use to personalize content.
It’s worth mentioning that every time you make an API call, it costs a fraction of a cent. Large companies are providing these cognitive services at little cost as part of their enhanced suite of cloud offerings. For the most part, you are not required to sign up for lengthy contract terms, meaning that the barrier to entry for using these technologies is extremely low.
While there is a great deal of hype around the application of AI for improving customer experience and marketing programs, hopefully, this article has helped to show there are many practical uses of this technology available today. We’re taking small, measurable steps right now, so clients can see the positive impact of these solutions for their specific business challenges. We have focused on the use of readily available and cost-effective cognitive services and machine learning APIs that clients can use to automate routine tasks, so their teams can focus on higher-value work like personalized campaigns and content.