Deloitte on GenAI in the Enterprise: Balancing Value and Risk

Generative AI (GenAI) is revolutionizing enterprise operations by enhancing productivity and creating new revenue streams. Alongside these opportunities are challenges related to data biases, security, and regulatory delays. The latest Deloitte: Decision Points series addresses these complexities, offering practical insights into integrating GenAI into business strategies. Post this discussion which happened in Nov 2024, Deloitte continues to do deep client work in AI, GenAI and Agentic AI across different industries and in varied areas of the business. Duleesha Kulasooriya (Innovation Leader, Deloitte Asia Pacific. Managing Director, Deloitte Centre for the Edge), Vijay Gopalakrishnan (Partner, Offering Leader for AI/GenAI, Automation, BPM, and Strategy, Deloitte India) and Haresh Perera (Enterprise Technology and Performance Leader at Deloitte Sri Lanka and Maldives)

discuss the transformative potential of GenAI and its implications for boards and leadership teams. 

The experts concentrate on pinpointing high-value use cases, managing associated risks, and ensuring that AI initiatives align with organizational goals. Key areas of discussion include enhancing operational efficiency, improving customer engagement, and transforming decision-making processes. They stress the importance of starting with projects that have clear, measurable outcomes, establishing robust governance frameworks, and preparing teams to adapt to AI-driven workflows. 

What are the best use cases for GenAI in an enterprise setting, and how can it drive value across the different business functions?

Duleesha: I spend a lot of time working with Boards and CEOs, and today’s GenAI tools are helping leaders by dramatically improving how they access, process, and synthesize information. Early applications include summarizing hundreds of pages of documents into key insights and enhancing tools like chatbots to better support customer and employee interactions. This is crucial for leaders who must absorb large volumes of information quickly without relying on multiple layers of staff. Those who embrace these tools early are gaining a tangible edge and, seeing the benefits firsthand, are encouraging broader adoption across their organizations. The key takeaway: GenAI can significantly streamline decision-making, making leaders faster, more informed, and more effective.

Vijay: Over the past year, my team and I have worked on over 100 Generative AI (GenAI) client engagements across many industries, including BFSI, government, life sciences, pharma, telecom, media, technology, CPG, retail, industrials, and manufacturing. These projects have spanned the entire business value chain, positively impacting front-office functions (sales, marketing, customer service, etc.), mid-office operations (pricing, finance, procurement, etc.), and back-office functions like IT, HR, manufacturing, and supply chain.  

The use cases in GenAI can fall into four main categories. The first and most widespread use case is search and summarization, where GenAI scans large volumes of information and documents, providing summaries and insights on a  near real time basis. 

The second set of use cases for Generative AI is content generation, which includes creating images, videos, and audio. It has proved particularly useful in digital marketing, making processes more efficient and saving time.

The third set of use cases involves coding and IT/technology optimization, where GenAI helps to automate code generation and creates huge synthetic data sets for performance testing. 

The fourth category covers various business applications, such as sentiment analysis, where GenAI scans customer reviews across social media and categorizes them as positive, negative, or neutral. 

While many of these GenAI use cases started as proof-of-concepts, about 25-30% and more have moved into production in the last few months.

Duleesha: From a board and CEO standpoint, the key questions are whether GenAI will help reduce costs (by improving efficiency) and whether it can drive top-line growth (by enabling new business opportunities). Some of the use cases we have seen directly address these concerns. For example, cost-saving use cases focus on streamlining operations, automating tasks, and enhancing productivity. Top-line, growth-driven use cases can open up new revenue streams, such as improving customer engagement, enabling personalized marketing, or creating new products and services. As we dive deeper into these use cases, we find that leveraging GenAI can optimize existing processes and unlock new business potential.

Which use cases are boards most familiar with now, and what should they be more familiar with further down the road?

Duleesha: The initial focus for most organizations adopting AI is improving efficiency and productivity, as these are the low-hanging fruit delivering quick, tangible results, faster, better, and cheaper outcomes. However, after this initial spike in interest, companies often seek to move beyond basic improvements. The next level of interest is in using AI to achieve things previously thought impossible, finding new ways to grow the top line and explore opportunities beyond traditional business models. This shift from incremental improvements to groundbreaking innovations is where the most excitement and potential are.

Vijay: While boards have initially focused on AI for productivity and cost savings, there is growing interest in GenAI applications that can directly impact sales and revenue. One example would be a global jewellery manufacturer, where GenAI allows B2C customers to input their preferences (e.g., gems, design, colour etc.) as prompts into a Generative AI tool which creates a near-real-time, personalized image of the jewellery piece. This level of hyper-personalization boosts sales and increases customer loyalty, as customers can further adjust designs in the image to their satisfaction and receive the piece of jewellery made as per that image which has all of what they want. 

The second example would be from the manufacturing and pharma space, where GenAI’s image creation features were used to rapidly generate marketing collateral tailored to new countries and markets where a drug is being launched, including images and content. It reduces the time required for marketing launches by 50%, speeding up entry into new countries. 

These innovative uses of GenAI deliver tangible business outcomes like faster time-to-market and enhanced customer engagement, with immediate positive impact on sales and revenue, and thus capturing the attention and excitement of boards and top management.

How does GenAI increase productivity? What specific areas from your experience with clients are showing significant interest?

 

Haresh: GenAI is a revolutionary tool that fundamentally changes how organizations work and think. It is not just about efficiency improvements. It is about transforming workflows and enabling new ways of thinking, collaborating, and innovating. One of the benefits for employees is how GenAI can support ideating and collaboration. It can help employees explore new possibilities, test new ideas and gauge market reactions quickly to help teams iterate faster and be more responsive to market needs.

An interesting use case is how GenAI can automate routine tasks, like meeting notes and action items. GenAI can instantly generate summaries, extract key action points and suggest follow-ups. It saves time and ensures accuracy and consistency in documenting meetings, allowing staff to focus on higher-value tasks rather than the mundane administrative work.

Moreover, GenAI can help organizations go beyond their internal knowledge base. By analyzing structured (e.g., databases, spreadsheets) and unstructured data (e.g., emails, documents, social media), GenAI can provide insights into what other organizations are doing, essentially offering a window into industry trends, competitor strategies, or innovative approaches. It can inspire new ideas and techniques that organizations may not have considered.

In short, GenAI is not only a tool for improving existing processes. It is a powerful enabler of new ideas and approaches. It brings continuous inspiration, helping teams become more creative, efficient and connected with the world outside their organization. GenAI can lead to deeper insights and drive continuous innovation within a business.

Vijay: GenAI has a profound impact even in traditional industries like manufacturing and logistics, where efficiency and precision are critical. Let’s break down a couple of powerful examples that illustrate how GenAI enhances productivity and quality in these areas.

In a manufacturing environment, for high-end luxury cars, engineers on the assembly line needed a system with near real-time access to engineering specs, quality documents, and other critical information to help automated and fast generation of assembly instructions to ensure every vehicle meets precise specifications, and also faster assembly times with the process being error-free. To support this, we built a GenAI-powered application for the engineers which can instantly access the appropriate information like data on car parts alongside additional assembly descriptions, and use that to quickly and automatically generate the error free automated assembly instructions . It ensures that the highest quality cars are produced and boosts productivity by reducing downtime and improving decision-making speed on the shop floor. 

GenAI can be a game-changer for improving response times and consistency in customer service. When a customer calls the contact centre of one of our clients from the logistics industry, the GenAI system we built for the client already knows the customer’s information and the nature of their inquiry. GenAI then automatically drafts an email response based on the data. The contact centre agent only reviews the email draft and makes slight modifications before sending it to the customer, significantly reducing response time, ensuring consistency in messaging, and improving the productivity of customer service teams. By automating much of the manual work, agents can focus on more complex or personalized interactions, improving both efficiency and quality of service.

Duleesha: I have another example of a global shipping company. The challenge was managing lading prices and the fees charged for shipping containers across different countries. Traditionally, the process was slow and inefficient: the company had to collect data from various sources and synthesize it, and by the time they had the complete picture, they were already lagging by two to three weeks. This delay meant they often worked with outdated pricing information, which put them at a competitive disadvantage. 

With GenAI, the situation changed dramatically. Now, the company can access real-time competitive data, including spot trading prices and pricing trends from various countries, almost instantaneously. The real-time processing and analysis of all this information led to immediate decisions, ensuring they always work with the most up-to-date data. It gives them a massive competitive edge, as they can adjust their pricing on the fly and respond to market conditions faster than their competitors.

In this scenario, GenAI is used for operational efficiency and intelligence gathering, allowing the company to make strategic decisions that directly impact its bottom line. The ability to stay ahead of competitors, adjust pricing dynamically, and optimize their business model based on real-time data is a perfect example of how GenAI can create competitive advantages in areas that might not initially seem like obvious targets for AI. This use case is a reminder that GenAI can help unlock value in almost any aspect of business, improving internal workflows, enhancing customer interactions, or transforming strategic decision-making. The possibilities are far-reaching, and this example helps illustrate just how powerful it can be to gain a market advantage in traditional industries like shipping.

Vijay: Absolutely, the potential of GenAI extends across every sector and profession, and the public sector is no exception, especially in regions like South Asia, where civil servants and bureaucrats often face time-consuming administrative tasks. One of the most common challenges in government is the time spent on routine yet critical tasks like drafting letters, preparing reports, or conducting research. These tasks require manual effort and can drain valuable time from higher-level decision-making and strategy. GenAI helps automate these tasks for the civil servants and helps get those tasks done in a comprehensive and very quick manner, using features of search and summarization and content generation. 

Duleesha: A powerful aspect of GenAI is it empowers leaders to be much more self-sufficient and proactive in their decision-making and strategic thinking. A particular tech-forward CEO I knew began using GenAI for routine tasks and as a brainstorming tool to challenge existing assumptions and push boundaries. It allowed him to accelerate his decision-making and gain deeper insights into various aspects of the business without waiting for reports or depending on others to gather the information for him.

Data is the new oil, and managing this data and all these large language models to generate AI applications presents challenges for data quality, privacy and reliability. How do we address this?

Haresh: Technology needs to be business-led, and that is especially true for AI and data strategies.  A business-led approach means that technology teams must work closely with business stakeholders to understand the strategic goals and ensure that the AI and data initiatives focus on the right objectives, from improving customer engagement to optimizing operations, increasing revenue, or enhancing decision-making. Alignment between business strategy and technology ensures that AI is applied in the most relevant and valuable ways, creating a tangible impact. Additionally, having a solid framework is crucial and includes defining policies and establishing guardrails to ensure technology usage is responsible, ethical and secure.

A critical aspect of this is trustworthy AI that ensures that AI models and systems are fair, transparent and accountable and that they operate in a way that builds trust among users, stakeholders, and customers. Data governance is at the heart of this framework. AI systems cannot operate effectively or responsibly without proper governance protocols. Data privacy, data labelling, and data quality are all crucial components. No organization has perfect data, but acknowledging that and building a data governance team to manage and improve data quality over time is an important step. 

The Data Protection Act will come into force a few months from now. What legal obligations will it present?

Haresh: An organization does not have to solve every challenge by itself. A wealth of external expertise, tools, and frameworks exists to help. However, understanding and addressing the foundational aspects of data governance, legal compliance, and AI ethics is essential, as it sets the right tone for the entire organization and establishes a solid operating foundation for future innovation. A solid framework with clear policies and guardrails ensures the responsible use of AI tools and that the organization can mitigate risks associated with data breaches, unauthorized access, and potential misuse. One significant risk is the potential for employees to inadvertently or deliberately expose the most sensitive or proprietary data by feeding it into AI models without understanding the privacy or security implications. 

Vijay: Your point about distinguishing between GenAI and traditional AI is crucial because GenAI is a subcategory of AI, but traditional AI (e.g., predictive models) remains relevant. GenAI often complements traditional AI by handling tasks like data search and summarization, to help input of data into predictive traditional AI models. Principles like data governance, data quality and privacy have always been crucial in AI. While some industries may prefer expensive on-premise solutions, many companies use API calls to cost-effective and stable cloud-based LLMs from Amazon AWS, Google GCP, Microsoft Azure/OpenAI etc. These cloud solutions are secure and can operate within virtual private clouds (VPCs) with proper vulnerability testing having been done. A human-in-the-loop review process is essential to mitigate IP risks such as violations from using LLMs trained on public data. 

For example, if we develop a GenAI application using LLMs trained on public data for a media company to generate articles, we need to use techniques like Retrieval Augmented Generation

(RAG) to restrict access to write data to the LLM instance being deployed to prevent hallucination causing wrong outputs of the GenAI application, and also a human needs to review the output of the GenAI created news articles  to ensure that there are no potential IP violation/copyright infringement issues due to public data used to train the LLM which was customized and deployed.  Despite using cloud-based GenAI solutions on enterprise virtual private cloud environments , security measures, such as traditional cybersecurity techniques and enterprise-specific safeguards, are in place. With over 100 client engagements in GenAI,  my team at Deloitte has not encountered any security breaches.  So, GenAI, combined with traditional AI and robust governance frameworks, ensures that businesses can leverage these technologies securely and effectively.

Duleesha: To get the most out of GenAI, organizations need to trust the technology and be willing to approach things differently, like how they handle data. One of the unique strengths of GenAI is its ability to extrapolate from incomplete or imperfect data. Traditional systems, especially in structured environments, often require complete, accurate datasets to function without a glitch. In contrast, GenAI can handle gaps in data or missing pieces, filling in those gaps by drawing from many sources and generating plausible outputs based on patterns it has learned.

From different perspectives in the boardroom and end-user cases, how do we prepare people to utilize the capacity of AI in the workforce?  

Duleesha: The biggest challenge to AI adoption is a shift in mindset across different generations. Younger employees are adopting AI tools more quickly because they recognize the value of using them. They can accomplish tasks that once took weeks in just days. If older leaders in an organization do not provide access to these tools, younger workers may seek them out as they see AI as a way to work faster, better, and cheaper.

Faster adoption often happens when senior leadership pushes the mandate down, but the middle layer of employees, those in managerial roles, tend to resist. These workers feel threatened because their role as intermediaries between staff and leadership is diminishing, and the introduction of AI makes them feel less relevant and less secure in their positions. This resistance is more about human and organizational dynamics than a technical barrier to adoption.

Vijay: The technical challenge of AI adoption also ties into how younger employees, including students and early-career workers, already use GenAI extensively outside of work, to do things quickly and more efficiently. Corporate leaders at the mid and senior levels must keep up with this shift and support their teams to leverage the same at work, or risk becoming less attractive employers. Providing employees with tools like Co-Pilot licenses, as one client did, can boost adoption, as younger workers are already familiar with and using GenAI tools on their own. If companies do not offer these tools, employees will likely use them independently.

Haresh: To successfully integrate AI tools into the workforce, collaboration between the CHRO (Chief Human Resources Officer) and CIO (Chief Information Officer) is crucial for developing a solid upskilling programme. Train every employee to use GenAI tools effectively. A structured training programme can help employees write better prompts, understand the right AI tools for the organization, and avoid mistakes, such as using incorrect prompts or mishandling data. Additionally, organizations must educate employees on the rules and guidelines for using AI, including data security. Hands-on workshops and practical experience, as seen with Deloitte and other organizations, build comfort and confidence in using these tools. Training is more than giving employees new skills—it also helps demonstrate that senior leaders value and accept AI usage, which fosters broader adoption across the organization.

Duleesha: The shift from a Centre of Excellence to a Centre of Adoption is a critical change in how organizations approach AI integration. The centre of excellence, typically led by tech experts, highlights the use cases and proof of concepts. In contrast, the centre of adoption combines the efforts of the CHRO and CIO to drive enterprise-wide AI adoption. The CIO provides the necessary technical infrastructure, while the CHRO takes a structured, agile approach to implementing AI tools across the organization. This collaborative model is considered best practice for achieving full organizational adoption and realizing the value of Generative AI.

Vijay: In AI adoption including GenAI adoption, a common pushback is when organizations use GenAI without seeing a return on investment (ROI). The key to addressing this lies in choosing the appropriate use cases and ensuring proper design. A simple two-by-two matrix can help select the use case that delivers the maximum business outcome, while also being technologically feasible to be deployed considering factors like data readiness. Additionally, designing for cost-effectiveness is essential. Using the right platform leveraging API calls to LLMs etc, and designing and implementing using the right techniques around embedding, chunking etc. to minimize costs can be effective.

From a technical perspective, efficiency in tasks like document searches is crucial. Techniques such as chunking and vectoring allow for more precise and cost-efficient searches within large documents. Data privacy and ethical concerns are also vital. Ensuring that AI models get trained on unbiased and appropriate datasets helps avoid issues like hallucinations or incorrect outputs and protects data privacy. 

How can boards and end-users effectively manage the risks of deploying Generative AI, given its rapid advancement, inherent biases in data and models, and the lagging pace of regulation?

Duleesha: At the recent AI Summit, there was a question about the rapid pace of technology and why invest in one solution when something better may come out in a few months. The response emphasized that waiting is not an option because technology is evolving too quickly. Even if a solution may soon be outdated, participating and gaining experience is crucial. The key is to design with flexibility, ensuring that solutions can be adapted or transferred to new platforms as they emerge.

Balancing innovation with regulation requires more than just meeting compliance. It involves developing an ethical framework. By adhering to this ethical approach, organizations can stay ahead of regulatory changes, ensuring they comply and lead in responsible AI usage.

Haresh: Companies need to collaborate with regulators and provide feedback. It begins with demonstrating that they have taken steps towards data governance and responsible use of technology, which can help shape a positive view from regulators. However, risks arise when companies rush into AI adoption without considering proper guardrails, which can lead to issues later on. Organizations can show regulators they are committed to responsible AI deployment by establishing clear governance frameworks. Regulators are there to support innovation, not stifle it, and will likely be more open to collaboration if they see value. Additionally, CEOs in some regions, like Sri Lanka, are working closely with regulators, and learning from more advanced countries and their regulatory approaches could help accelerate the process of getting up to date with the rapidly evolving landscape of GenAI.

Any inputs on managing the risks?

Vijay: The risks of adopting Generative AI (GenAI) are like the risks of driving vehicles on the road. There is always a small probability of something going wrong, but that does not stop you from driving vehicles on the road. Similarly risks exist for GenAI, just as with traditional AI. Organizations do implement design reviews, testing, proper model training on the right test data, and well-established risk advisory and cybersecurity practices including testing to prevent issues with GenAI implementation and adoption, and Deloitte has worked with these organizations to help them in this process. Following these principles has helped avoid major breaches and demonstrated the effectiveness of managing risks.

How should enterprises measure ROI for AI and GenAI?

Vijay: Measuring ROI for GenAI starts with identifying the most pressing business pain points that GenAI can address, ensuring immediate and impactful outcomes. Next, a feasibility check is crucial, ensuring that the necessary infrastructure and data are available to support the solution. Beyond that, the ROI measurement framework for GenAI should follow the same principles as any other technology solution, focusing on evaluating the effectiveness and value it delivers to the business.

But should you be more patient with AI because of its transformational nature? 

Haresh: When selecting a use case for Generative AI, choose one that addresses a clear business pain point without being too complex, as overly complicated projects can take too long. Simple process improvements to save time on repetitive tasks can offer immediate ROI. For example, reducing the time spent on tasks like job description creation or administrative work can lead to measurable efficiency gains. Additionally, GenAI can open up new revenue opportunities, such as improving campaign strategies and personalizing customer experiences, which can drive sales and increase the top line. While ROI is crucial, you must be open to experimentation. Organizations should start with measurable use cases and iterate, celebrating successes and learning from failures. A phased approach will build confidence and show boards that AI investments are worthwhile. Ultimately, with the right approach, ROI will follow. It is part of a longer-term journey.

Vijay: While senior management should be patient when adopting Generative AI, the teams deploying GenAI solutions must recognize that management has limited time and budgets. Both internal teams and vendors must manage expectations of the senior management and not stretch resources too thin. What helps is to do what Deloitte works on with clients on every engagement including GenAI engagements i.e. identifying use cases which address a key business pain point which is important for the client and its management team, and also use cases which have the right data and IT infrastructure to make the GenAI deployment technologically feasible. The good news is that AI adoption is progressing well, with many projects already reaching the production stage, indicating that the transition is happening successfully.

Duleesha: From a board perspective, the concern is whether investing in AI is the best use of resources. The recommended approach is to start with a phased strategy: begin with a proof of concept, demonstrate business value, and highlight savings. Next, expand the project to a production scale within a specific domain. Once there is confidence in the technology and a capable team, the focus can shift to scaling AI across the enterprise. Jumping straight into large-scale enterprise licenses is premature and could be risky. First, prove the value and ensure the team can manage and deploy the technology before driving broader adoption.

Look into your crystal balls. What does the future of GenAI in the enterprise look like 5, 10, 15 years down the road?

Duleesha: In the next few years, the role of AI will fundamentally change. AI is already advancing rapidly, with startups creating digital twins of humans, AI agents trained to take on specific roles like marketing, customer service, supply chain, or even chief of staff. These AI agents will be highly capable of gathering and processing vast information volumes. In the future, work will increasingly involve hybrid teams, with AI agents integrated alongside human employees, not just for incremental tasks but as full team members. AI will play a significant role in organizations, including in boards and management teams, reshaping how businesses operate.

Vijay: While it is difficult to predict what GenAI will achieve in the future, AI and GenAI are established and rapidly growing technologies which are here to stay. GenAI has made significant strides in the past year, complementing traditional AI. The full potential of GenAI is still unfolding. What seemed impossible a year ago has already been accomplished. GenAI will continue to surprise us with new capabilities, including areas where traditional AI or even current version of GenAI has limitations. 

Haresh: The technology landscape is evolving at an unprecedented pace, with innovations like robots, digital agents, and drones reshaping how work gets done. While these technologies lack human emotions, they are increasingly augmenting human capabilities — especially in high-risk roles and tasks people are unwilling or unable to perform. The real differentiator will be how businesses and individuals integrate these tools to build safer, more productive, and more resilient environments. Although the future remains difficult to predict, the constant pace of change makes this era of technological transformation one of the most dynamic — and full of opportunity — we have ever seen.