CXOToday has engaged in an exclusive interview with Mr. Balakrishna D. R. (Bali), Executive Vice President – Global Head, AI and Automation and ECS, Infosys
- What are the factors affecting AI’s risk of failure? How can the involvement of human intelligence mitigate them?
Some of the factors are:
- Lack of stakeholder support and endorsement across the organization
- Misidentification of AI opportunities leads to an inability to scale from point solutions
- Insufficient amount of good quality training data leads to problems like overfitting
- Biases and discrimination present in the data and model
- Lack of transparency due to poor explainability
- Lack of reproducibility and drift control mechanisms due to poor MLOps
- Lack of proper control mechanisms to ensure that regulatory aspects are adhered to
- sLack of proper mechanisms to detect and mitigate security threats.
Human intervention is needed in critical stages of an automated decision-making system. Human actors should have the capability to override AI systems and raise alerts for preventing undesirable scenarios. Our Infosys Applied AI offerings are driven by diverse teams having representation from all disciplines, business functions, and demographics, who can identify and mitigate threats such as data quality issues, biases, compliance, and security issues, well ahead of time and take corrective measures. Different points of view are needed in different stages that can bridge gaps in the AI system.
- How would you outline Infosys’ strategy to identify valuable data sources and drive seamless AI transformation?
Our Applied AI strategy is powered by our 3D framework which tackles all the different complexities and roadblocks, right from the conception stage to the final delivery of AI projects. Let me elaborate on the various activities that are done in each stage to build robust AI systems delivering consistent business value.
Discover: In this stage, we embark on various exploratory activities for identifying potential opportunities for transformation. Coupled with our domain expertise and technical acumen, we assess, discover, and quantify the disruptive potential of each of the opportunities. Through design thinking workshops and idea exchanges, we leverage our consulting accelerators like AI Canvas, Verticalized Blueprint, AI Radar, AI Maturity assessments, etc., to identify the low-hanging fruits as well as the ones that will deliver maximum business impact.
Democratize: Our focus in this stage is to strategically leverage key AI assets to drive mainstream adoption of AI across the enterprise. These will be enabled by platforms offering a library of cognitive business solutions, pre-trained models and data sets, compute power, environments for experimentation, and rapid deployment. An AI service store that leads to the uberization of AI services coupled with MLOps and control tower capabilities helps scale AI solutions rapidly.
De-risk: Responsible AI initiatives lie at the heart of our AI strategy. We have tools, frameworks, and processes in place to develop AI systems that are secure, ethical, fair, explainable, and purposeful. Another focus area in this stage is digitally re-skilling talent for seamless human-machine co-working.
- For enterprises transitioning to live AI operation and business adoption, how can Infosys help them gain an AI-powered competitive edge and drive the digital shift?
We are helping enterprises transition to ‘live’ AI adoption through our Infosys applied AI. Infosys applied AI is an integrated offering that leverages the synergy of AI, analytics, and cloud to deliver differentiated offerings for businesses.
It unlocks unmatched efficiencies, future-proofs, and scales existing AI investments, while mitigating the reputational and performance risks of AI. Infosys applied AI cloud enables businesses to readily access AI hardware and software, across private and public clouds, as highly contextualized platform services. They can also process AI algorithms locally on any device using Infosys edge AI.
Infosys applied AI helps enterprises uncover actionable insights from their own data estates, open-source data, and curated data exchanges on the cloud to build new AI models and use cases for the organization. With our offering, businesses can harness the full potential of the Infosys AI partner ecosystem and create fit-for-purpose solutions in the AI Living Labs, orchestrating offerings from startups and over 30 leading providers of intelligent automation, AI solutions, data solutions, and enterprise security.
From state-of-the-art computer vision to natural language processing, Infosys applied AI cloud brings pre-curated deep learning libraries and accelerators that enable fast deployment and collaboration for data science teams.
Companies, working with Infosys applied AI, have ready access to a growing portfolio of indigenous AI solutions to solve business problems. For example, an American bank used one of these solutions to create an NLP-based expense claims management mobile app.
Businesses can rely on Infosys applied AI and Infosys’ rich ecosystem of consortiums and regulatory bodies driving AI directives for implementations that are compliant with regulatory and security norms. Analytics model interpretability, bias detection, and continuous performance monitoring are built into various stages of the lifecycle, from development to deployment and use. For example, using Infosys applied AI, a machinery manufacturer analyzed warranty claims patterns to eliminate bias, from the data set and process, before reengineering and automating the claims approval process.
The applied AI cloud offers the following key services:
- Can work with any hyperscaler
- Provides a wide array of AI services for video, image, and language processing
- Ready to deploy
- Scales AI from point solutions to enterprise-wide
- Democratizes and makes AI available to all, fostering innovation through experimentation
- Accelerates AI adoption by providing AI engineering, technology, and industry-specific services
4. How is Infosys futureproofing and efficiently scaling AI investments while managing reputational risks?
As one of the leading IT services and consulting companies, Infosys is constantly investing in state-of-the-art AI to remain at the forefront of innovation and meet evolving customer needs. Here are some of the ways Infosys is futureproofing and scaling its AI investments while managing reputational risks:
Developing next-gen AI first platforms: Infosys has developed its own Applied AI Cloud that has pre-curated assets and models and data sets for applications in video and image analytics, natural language processing, etc. These next-gen AI capabilities are also baked into all our vertical-based solutions in retail, finance, manufacturing, and other industries for use cases such as predictive maintenance, asset protection, quality inspection, and others.
Incubating experts in AI: We also invest heavily in training our workforce on AI and its applications. Our workforce is proficient in developing AI systems in the latest cutting-edge deep learning models, large language models and Generative AI, physics-informed neural networks, metaverse, and all.
Collaborating with leading AI vendors and partners: Infosys has formed partnerships with leading technology vendors to stay abreast of the latest AI developments and leverage their expertise in AI. Infosys was one of the first investors in Open AI. Through the Infosys Innovation Network, Infosys also collaborates with start-ups in the AI space to tap into new ideas and innovations and co-create point solutions.
Infosys Responsible AI: Infosys is committed to ensuring that its AI solutions are ethical, transparent, and fair. We have systems, frameworks, and tools for ensuring ethical, bias-free AI development, deployment, and use. Infosys also adheres and enables enterprises to comply with data privacy regulations and best practices to safeguard sensitive data through secure and explainable AI platforms.
Infosys Living Labs: Infosys invests heavily in research and development to stay ahead of the curve in AI and related technologies. The company has established several living labs and innovation centers that focus on AI, machine learning and other emerging technologies.
- What is the impact of Infosys applied AI across the value chain, from development to deployment and usage?
The applied AI brings transformation across the AI value chain by strategically bringing together various enablers across each stages.
- A shared infrastructure for machine learning and deep learning operations, and being able to support a large number of projects on cloud infrastructure at any point in time
- Private AI infrastructure supporting all the deep learning and machine learning needs
- Public AI cloud used for specialty workloads, generate training data, and using highly mature services
- An AI store as part of the applied AI cloud with pre-curated and ever-increasing assets, including datasets required for training the AI models, catalog of the latest models, codes for AI services, and cognitive services for rapid experimentation and innovation
- A platform for AI engineering lifecycle to scale AI projects from pilot or experimentation to enterprise scale deployments
- Superior MLOps capabilities and tools and frameworks for model validation and model health monitoring
- Could you share some thoughts on the best AI adoption practices from a practitioner’s perspective?
Here are some best AI adoption practices from a practitioner’s perspective:
Conduct proper discovery exercises: Before starting an AI project, it’s important to have a clear understanding of the business objectives that we want to achieve. An in-depth assessment of the requirements and the boundary conditions as well as overriding conditions need to be blue-printed. This will help us to identify the right use cases for AI and prioritize the efforts accordingly.
Building a strong team: AI projects require a multidisciplinary team, including data scientists, engineers, product managers, and business analysts. It is imperative to have the right talent in place with the necessary skills and experience to deliver the project.
Invest in data and compute infrastructure: AI models require large computational power to train and operate effectively. It’s important to invest in the necessary data infrastructure, including data storage, processing, and quality control.
Focus on data quality: The quality of your data is critical to the success of your AI project. Make sure to invest in data cleaning, data labeling, and data verification to ensure that the models are accurate and reliable.
Empowering through Responsible AI: AI models can have unintended consequences, so it’s important to ensure that they are used ethically and responsibly. This includes considering issues like bias, fairness, and privacy when designing and deploying AI systems. The AI models should be explainable and reproducible and should not infringe on regulatory guidelines. AI testing and model validation must be rigorous and proper standards should be in place.
Continuous learning framework and MLOps: Once the AI system is deployed, it’s important to continuously monitor and improve its performance. This includes collecting feedback from business users, tracking performance metrics, and refining the models over time.
Infosys Knowledge Institute’s inaugural Data+AI Radar identifies why AI fails to deliver on heightened expectations and recommends three areas for improvement: develop data practices that encourage sharing, bind explanations into advanced AI, and focus AI teams on business. If companies improve on these fronts, they can add up to $467 billion in profit growth, collectively, and increase internal satisfaction with data and AI.
