5 best practices for scaling AI in the enterprise

AI has entered a new phase. There has been an explosion in generative AI in recent months. The ability to use text to automatically write stories and create art is developing very rapidly. Early applications of these new capabilities in co-authoring software, writing news articles and business reports, and creating commercials are already emerging. We can expect entire industries – from software engineering to creative marketing – to be disrupted.

At its core, AI has become the best prediction machine possible. We’ve seen AI built not only into large applications like autonomous driving, but into hundreds of tools and utilities for everyday use. AI has reached the right tipping point on the maturity curve to drive mainstream, important and diverse business applications. While AI is disrupting the way we live and work, true innovation for most companies comes not from experimentation, but from industrializing AI at scale.

Here are five best practices for making the most of emerging AI opportunities across the enterprise.

Start with the question, not the answer

One of the main challenges of implementing AI is defining the business problem the enterprise is trying to solve. As the saying goes, don’t end up with an answer looking for a question. Simply deploying new forms of technology is not the right approach.

Then examine the issues and determine if AI is the best way to address the problem. There are other digital technologies that are well adapted to simple problems. To ensure success, clearly define the business problem and determine the course to take from the start – some may not need AI.

Plan for AI-based transformation to be different from automation

In automation, the end-to-end process is broken down and broken down into smaller parts. Each part is then digitized and the parts are then reassembled in the value chain. Automation ensures efficiency, time-to-market and scalability, but the underlying work and process remain the same.

On the other hand, when enterprises use AI to transform, entire value propositions are reimagined, the customer experience changes, processes are redesigned end-to-end, and the remaining work becomes fundamentally different than before.

So AI-powered transformation is as much about designing a new business model, training employees, and integrating them into upstream and downstream processes as it is about neural networks and model management. It is important to note that AI in the enterprise is 20% about technology and 80% about people, processes and data.

Create a base of data

We are moving from a data-poor world to a data-rich world. We are increasingly integrating telemetry and digital devices into our corporate environments that open up new data sources that were previously unavailable.

With AI, data that was traditionally in unstructured formats can now be easily extracted, converted, and used productively. As a result, data now available to support business operations and decision-making is unlike anything ever before.

Building a foundation of data is critical to reaping its benefits. The management of data, not only in terms of the core data infrastructure, but also in terms of quality, security, permitted purpose and granular access is essential.

Focus on digital ethics

With the growing footprint of ambient intelligence comes the associated risk of security breaches, model anomalies, unintentional bias, and unethical use. As AI use cases expand and proliferate and massive amounts of data are collected and managed centrally, this opens up the potential for security breaches.

Model drifts happen when AI models – as they tune themselves with new data – eventually drift towards lower accuracy results. If not purposefully designed, bias can often be unintentionally introduced into AI systems. The use of AI should be monitored to ensure it is used ethically.

Digital ethics must be pre-incorporated into the design and architecture of the system. Adding it as an afterthought is not a comprehensive approach and leaves too much room for harmful exposure. Redesigning for ethics can end up being a costly and wasteful exercise.

In the long run, companies that build and succeed with industrialized AI systems will not get there by chance, but by focusing on building digital ethics and governance into their platforms from the start. Many organizations will likely have a Chief Ethics Officer or board-level ethics subcommittees in the near future.

Change management and culture are key to success

With AI, we are driving the pivot of the business, not just to increase efficiency or reduce costs.

The technology of AI itself is not difficult to implement. What poses a challenge is the significant integration, contextualization, governance and adoption required for success. Best-in-class AI projects in manufacturing require a thoughtful process to reshape the business, seamless integration into upstream and downstream processes, a fundamental change in the way we work, and the adoption of user technology. This requires a company culture of change, learning and agility.

Ultimately, culture will separate winners from losers when deploying AI.

Using AI benefits everyone

Industrialization and automation have changed the way we work and live. The opportunity with AI is to go beyond the limitations of predefined and already known rules-based automation. As we do, AI will disrupt entire businesses and new business models will emerge. AI will become critical to delivering sustainable business and sustainable benefits.

By following these five best practices, enterprises can begin their journey to take full advantage of the promise of AI.

Sanjay Srivastava is chief strategist at Genpact.