Note: This is the Third Part of a three-part series of our study ‘State of Enterprise AI In India 2019’ brought to you in association with BRIDGEi2i.
Subscribe to our Newsletter
Join our editors every weekday evening as they steer you through the most significant news of the day, introduce you to fresh perspectives, and provide unexpected moments of joy
Evolving AI Delivery Models
Gradually, we shall see the rise of AI as a separate function, and it will be tightly coupled with the solutions/services a particular enterprise offers. AI has given rise to three distinct delivery engagement models — AI-as-a-Service, AI-as-a-Solutions & AI-as-a-Product.
The modern AI stack consists of — infrastructure components that include computer hardware, algorithms, and data. From managing the building blocks to implementing production-level AI solutions that can generate results within a period of 7-8 months, AI delivery models will significantly change the enterprise AI landscape.
a) AI-as-a-Service: Industry experts forecast that AIaaS will soon evolve as the preferred delivery model that enables rapid, cost-saving onboarding of AI without being too heavily reliant on AI experts in-house. AIaaS consumption model enables enterprises with readily available cognitive capabilities and accelerators, allowing their team to focus on the business problem without having to worry about the underlying AI hardware/infra components. Another instance of AIaaS is when Solutions Providers list several of their Deep Learning and Machine Learning algorithms through a tie-up with AWS Machine Learning Marketplace.
b) AI-as-a-Solution: Solution providers deliver production-level AI solutions, custom-built around narrow business problems. In this delivery model, AI solutions providers and boutique vendors follow a collaborative approach — co-development of solutions that involve industry domain expertise. The solutions are deployed on-premise or on cloud infrastructure. By following iterative agile methodology and making the build cycles more iterative, solutions solution providers co-create solutions that deliver business value.
c) AI-as-a-Product: AI-as-a-Product is when an AI software product can be configured according to the needs of an enterprise. An example of AI as a product would be BRIDGEi2i’s Watchtower & Recommender that provides granular insights with real-time alerts. These products can be configured as per the specific needs of an organization and will also have to work seamlessly with other software products on the enterprise shelf.
The critical decision point will be choosing the right AI partner with domain experts, analysts, AI solutions engineering teams who can build the best solution in the shortest span of time. We believe with the shift in the scale of adoption, the role of AI Solution Delivery Leader will evolve as the one who enables the creation of production-grade AI-based automation solutions and lend business value.
Rise Of AI-As-A-Service Economy
According to 14Dell Technologies’ Digital Transformation Index, India is the most digitally mature country in the world. With the third-largest startup ecosystem and a strong developer base, India is on the cusp of a massive digital transformation. As digital organizations move further up the ladder to harness the potential of AI across enterprises, the AI-as-a-Service model (AIaaS) model will become a necessity in the near future from provisioning pre-built accelerators, data access, right AI tools and APIs as self self-service trend gathers momentum.
Veteran IT leader Kris Gopalakrishnan posited that AI and machine learning could be as big as 15$177bn IT services industry. Given how the AI disruption is here to stay, we see India playing a more significant role in strengthening the global AI ecosystem. India is the third-largest startup ecosystem across the globe, with 40,000 AI developers. We are also the youngest country in the world, which means that not only do we have the talent base to fuel transformation, we can also upskill and align the talent to harness the potential of AI. Home to some of the largest service providers, global system integrators and consulting companies, India is poised to become a global AI hub.
The burgeoning AI Services market, led by global consulting majors like Accenture, Deloitte, PwC, KPMG, and EY is complemented by mid-size and niche AI Service providers like Mu Sigma, BRIDGEi2i, Cartesian Consulting and Fractal Analytics that are offering high quality AI expertise and in-built accelerators — pre-designed and pre-validated solutions which can accelerate “on-ramp” to AI effectively. With stronger competencies, AI talent base, and competencies in specific verticals — mid-sized firms are well-positioned to provide more value to larger enterprise customers.
Which Sectors Are Frontrunners in AI Adoption & Where’s The Momentum Building
BFSI, due to its sheer size, is the largest adopter of AI. We see Machine Learning, Computer Vision, and robotic processing getting very widely adopted in BFSI. Telecom, Retail. Healthcare & Manufacturing are the next two sectors digitizing their processes that will be the torchbearers soon.
1. Banks Are Detecting Fraud and Managing Risk With AI
FSI industry generates enormous amounts of data mostly in a transactional form, which can be analyzed in real-time to make smart decisions. For banks, one primary application of AI is the automated underwriting of loans based on a customer’s entire history of transactions and credit scores. This would also eliminate human bias and errors that usually occur in loan approvals. AI is on top when it comes to security and fraud identification. By analyzing millions of transactions, machine learning systems are helping financial organizations identify anomalous patterns in transactions, which is reducing cases of fraud and strengthening trust among parties.
2. Telecommunication Players Are Using AI For Network Optimisation
Telecommunications is another sector that is leading in the adoption of AI. This is because there will be 20.4 billion connected devices16 across the globe by 2020, and CSPs understand that untapped value can be generated with the data being generated. CSPs are adopting AI/ML for purposes like network optimization, virtual assistants, and process automation. AI is essential for CSPs to create self-optimizing networks (SONs), which gives telecom operators the capability of automatically optimizing network quality for a particular geography and time.
3. AI Is Critical For Customer Experience In Retail
For retail companies, AI creates an opportunity to BRIDGE the gap between virtual and physical sales channels. From daily task management to gaining customer insights, AI is a key technology in a retail setting. The AI market in the global retail market size is expected to exceed$8bn by 202417, according to Global Market Insights. Factors like demand for supply chain optimization, enhanced business decision making, and forecasting among retailers are proliferating the use of AI in the retail market. Retail organizations emphasize the interaction between the business and customers is critical for the success of the business to create customer loyalty. As most retail businesses today are omnichannel, the use of AI helps them optimize their processes across different platforms, be it web, app, or the physical store.
4. AI is Driving Personalized Healthcare
The growth of artificial intelligence in the healthcare market is mainly driven by fast-rising demand for precision medicine, predictive diagnostics. Apart from providing better healthcare services, AI can also help with effective cost reduction in healthcare expenditure. Healthcare personalization is crucial due to its use in medical diagnostics, where a patient’s present and historical data is used to detect and predict serious health conditions. In addition, the growing need for accurate and early diagnosis of chronic diseases and disorders further supports the growth of this market.
5. Manufacturing Is Leveraging Sensor Data For Predictive Automation
In manufacturing, production processes are being automated, monitored, and integrated to create optimum use of resources. The staggering amount of data available in manufacturing processes through IoT sensors create the ideal environment to help train AI models. One major use case where AI is being leveraged is predictive maintenance of machines, where the analysis of various parameters of AI systems can alert companies of impending failures. AI algorithms are also being used to optimize manufacturing supply chains, helping companies adapting to market variables.
Impact of AI On Business – Use Cases
1. Enabling Data-driven Digital Transformation for FS firm
A leading Financial Services firm in India with over 8 million active customers and 15,000+ merchant locations across the country wanted to leverage data to enhance their digital transformation journey, including understanding their customer profiles and underlying personas, creating personalized recommendations and offers and reducing their fraud rates.
- Need to identify the next best/cross-sell offers, personalized recommendations for customers based on life stage and affluence to enhance customer experience
- Reduce fraud rates for first EMI default
- Improve IVR leakage and drop-off rates
- Developing an Accurate Booking Forecasting Engine
What BRIDGEi2i did?
- Use machine learning model to improve first EMI default scorecard
- Identify classes of information available in data and change in the mix by seasonal months, use sparsely populated but important variables to drive higher 20% higher lift in model
- Drop off & Leakage analysis and recommendations to improve the existing IVR menu
BRIDGEi2i Explored multiple recommendation algorithms for identifying the next best product recommendation. By leveraging BRIDGEi2i’s assortment recommender engine with its Gradient Boosting Technique, the team recommended the next most probable products for each customer. The team mapped life-stage product recommendations for each customer micro-segment, stamped for each customer.
- Customer Life Cycle: 75% accurate top two loan recommendations & 33 cross-sell profiles identified across 3 segments
- Affluence Segmentation: Migrated to 17 bands, reduced concentration in low affluent segments
- IVR Optimization: Quantified IVR drop off (65%) – key nodes for improvement identified
- Fraud: 70% outlier frauds detected; 2 investigated fraud captured & 40% reduction in First EMI frauds
2. AI-enabled recruitment solution for a low-cost carrier that caters to global ground management and air transportation.
The client wanted a robust and efficient recruitment solution that is capable of handling vast quantities of data and process massive applications. The client wanted to digitize the process of recruitment with AI-based stack rankings of the applicants with profiles and performance parameters.
What BRIDGEi2i Did?
BRIDGEi2i, in partnership with a technology vendor, created a single platform for all requirements with an intuitive, user-friendly design. The solution is a one-stop for requisition to onboarding. Bots are deployed for basic transactions and responding to frequently asked questions. API linkage and contact with Job Boards, online Document Management, Background Screening, and Medical integration was also carried out.
- The incorporation of the BRIDGEi2i solution ensured that the client found the process to become much more efficient with instant access to the talent pipeline.
- Overall, the client got the Dashboard view, which provided better management and also reduced the candidate acquisition costs.
- The client was able to re-use the profiles for relevant roles and focus solely on sourcing and selection.
Challenges & Opportunities in Enterprises
India is primed for Enterprise AI owing to the huge base of consumers and a sheer number of use cases. The global and Indian landscape is fast evolving to introduce AI across enterprises. Over the next few years, we expect AI-as-a-Service & AI-as-a-Solution to flourish AI as a Services and AI as solutions, both to flourish. We also see an extended ecosystem with the Open Source community, AI Consultancies, and Service Partners who build their own assets so that the package can be offered to end-users as a service.
But challenges still exist — as compared to the consumer world, AI in enterprises has to work on smaller amounts of data; for example, clickstream data in the consumer world versus user transaction details in the banking world. Hence, the accuracy of the predictions that AI provides is mission-critical. As compared to typical IT-based projects, uncertainty in outputs is inherent in AI and ML projects.
Explainable AI will also be a key factor for enterprise AI adoption. A well-defined enterprise IT solution for marketing-to lead works in a deterministic manner, but an AI solution that predicts new customer acquisition would give very different and unexpected results based on the training data.
Key growth enablers for enterprise AI in India are:
- Availability of usable structured data across various domains
- Availability of tech expertise in the talent market
- Market demand for cheaper options of automation
- Mobile and IoT availability of the high-end technology
- A growing number of sector-specific use cases across India, APAC & North America
- New business models and AI-based solution will drive synergies
Topmost bottlenecks holding back AI adoption
- There’s a lot of confusion in the market with non-AI solutions getting mislabeled as AI solutions.
- Lower awareness at the CXO level on how to make investments in AI and drive the ROI.
- Data protection laws in India are maturing, and enterprises have to implement privacy. Explainable AI will be a crucial factor for widespread adoption. As opposed to typical IT-based projects, AI solutions are probabilistic, not deterministic. Hence the results expectations need to account for its nuances.
- A strong absence of industry-academia ecosystem.
- An acute need for AI talent and skill augmentation.
C-suite needs to meet certain criteria before implementing AI solutions at scale:
- Identify situations and use cases where AI makes can deliver the most value
- Need to have access to computing power that can process and explore these massive amounts of data
- Build a company culture that recognizes the need for AI
- Put in place data governance policies to manage data securely
At one level, BRIDGEi2i has AI labs and Smart Apps, a committed CoE of over 100 people researching, analyzing, and deploying solutions to business problems. BRIDGEi2i doesn’t believe confining these results to the labs; with its knowledge community SCalA, the employees are taken through a learning path disseminating that knowledge and expertise to apply it to the real world.
By virtue of working closely with businesses and understanding the issues that concern them the most, BRIDGEi2i has been able to devise and pin-point the four troublesome areas that most businesses struggle with:
- Monitoring extensive data and real-time alerts
- Aiding Decisions
- Planning and Optimization
- Interactive overlays
We can solve some of the most complex business problems through contextual solutions that leverage consulting expertise, advanced data engineering, and our four proprietary AI accelerators.
Here are some scenarios where we feel AI capabilities increasingly find usage:
Data Extraction: Data is now collected in many types, handwritten notes, excel sheets, images, the video that is almost impossible to parse in a short time manually. So, Image processing and Computer Vision being used to make sense of the data.
Identity recognition with Computer Vision: Customers are being on-boarded with a video recording that records their facial features and used to identify them at PoS, access points, etc. to confirm identity.
Insurance Underwriting: Customers record a video of the scratches and dents on a car damaged by accident, and that is being analyzed to ascertain claim reimbursement in insurance.
NLP for topic mining and chatbots: A chatbot or voice-assistant is mining the data with NLP and provide customer resolution.
Anomaly Detection: Finding out anomalies in patterns about what is not normal and flagging it off as “Risk” or “Likely Fraud.”Preventive Maintenance: Monitoring machine performance through the linked sensor and being able to predict when preventive maintenance is required so that shutdowns can be avoided.
Conclusion & Way-Forward
Looking ahead, the AI-as-a-Service (AAS) model offers a multi-billion dollar opportunity to service providers. AI consultancies are poised to ride the growth wave and compete for a bigger market share by providing an accessible path to AI, deep AI talent bench, a more responsive relationship coupled with the best-of-the-breed AI technologies and lower cost as compared to large vendors. Some of the major differentiators of AI Service providers are sizeable AI workforce, best-in-class solutions for specific domains, and presence across multiple sectors and geographies.
Today, many organizations have more data now than in the past. However, the key challenge for building relevant AI applications is a learning data set. Most organizations taking their first steps in AI seek solutions around specific business problems that can deliver tangible returns against KPIs. AI consultancies with strong AI delivery competencies are highly valued for providing a swift “on-ramp” to AI tech through pre-built accelerators that can be easily integrated with existing IT systems and provide returns in 7-8 months. As PoCs mature into broader deployments, in the longer term, we’ll see AI Service firms becoming valuable partners in the digital transformational journey, and help enterprises deliver early wins in AI, even in the test and learn phase. In the next few years, we shall see more companies outsourcing AI initiatives. Some of the dynamics shaping the AI-related outsourcing market are a lack of talent, fear of going all alone on AI initiatives, and the rise of managed AI delivery models.
As captured in the report, the AI Services market is dominated by the Big 4, mid-sized AI firms, and boutique vendors. To stay ahead of the pack, AI consultancies will have to integrate strong AI teams, bolster outsourcing capabilities, and build sector-specific capabilities. The focus will also shift on acquiring tech assets that can augment in-house capabilities and reduce the cost to serve.
On the other hand, before onboarding AI vendors, buyers should understand how PoCs can deliver tangible ROI against the KPIs or specific business functions, the deployment methodologies, and how outsourcers can provide advanced capabilities.