Can AI truly transform businesses? Yes, if adopted thoughtfully

The buzz around Artificial Intelligence (AI) is inescapable. While there are no widespread examples of substantial impact yet—outside specific areas like software development—many believe AI could be transformative for businesses. McKinsey & Company, for instance, estimates that AI could add up to $4.4 trillion to annual global output by 2030.

While the possibilities prompt excitement, sceptics argue that they have seen such hype before, that an earlier wave of digital transformations was powered by lofty promises that did not convert to measurable business value. McKinsey research found that about 70% of transformation projects failed to meet their goals—not because of the digital technology employed, but because of a lack of user-centricity, slower-than-expected execution and poor tech adoption strategies—despite having the right intent and execution model. Businesses can avoid repeating the same mistakes with AI by following three major learnings from successful digital transformations.

Also Read: India must forge its own AI path amid a foundational tug of war

First, find truly game-changing projects: AI has many useful but non-transformational use cases, such as analysts generating first-draft reports, lawyers drafting contracts or employees conversationally querying a large base of information and/or data. While these improve individual productivity, they rarely make a transformative impact on the performance of the entire organization. Yet, these remain the most visible and common AI efforts today.

A robust roadmap for AI adoption could be invaluable to help distinguish projects that create meaningful business impact from those that deliver only localized productivity gains. Both have value, certainly, but the transformational ones need to be highlighted and replicated.

For example, in customer sales and relationship management, AI could be truly transformative. Take ‘intelligent agents’ for business-lead management or customer engagement that interact with customers via chat, voice or email, to answer queries and push tailored nudges; early pilots have seen double or triple the engagement and conversion rates, with 10-20% or more in incremental revenue impact.

Such AI solutions work best where a human is not required and machine learning can significantly outperform an average human put to the same task.

Also Read: Artificial intelligence tools raise productivity. They can also act as great equalizers

There are also ‘autonomous agents’ that work well for procedural tasks such as simple customer-service requests where machine learning might not add value; here, AI can cut costs by 40% or more, even with high AI-tech operating costs.

Then there are ‘copilots’ that are best suited for business adoption where human involvement is required and machine learning can meaningfully enhance human performance, such as assisting business-to-business salespeople with real-time support to handle queries, customize proposals, draft contracts or follow up with customers. Early copilot adoption tests have shown effectiveness gains of 10-30%.

Second, get the set-up right: This is best done with a seasoned leader, a ring-fenced team of domain and technology experts and a ‘garage-like’ operating model that fosters co-creation. Businesses need to be able to build quality data pools, use agile development for fast iterations, partner with third-party providers and run parallel pilots if needed. They should enable rigorous testing and model training and robust impact measurement.

Just as important is an adaptive budgeting approach. Funding should be ramped up swiftly for any initiative if early success is seen and kept flat otherwise. This would work far better than the typical approach of an annual budget for each project.

Also Read: Nitin Pai: India can shape how artificial intelligence diffuses through society

Third, plan for systematic adoption: ‘AI copilots,’ the usage of which depends on large groups of employees changing their task behaviour, require thoughtfully crafted adoption plans and committed leaders to champion their use. The same applies to ‘autonomous’ or ‘intelligent AI agents.’

Often overlooked is the importance of good content and audio-visual adoption guides in driving organization-wide AI adoption. In the example of ‘intelligent AI agents’ for lead management or customer engagement cited above, the message that gets conveyed to employees, ‘how to’ videos and adoption incentives make a big difference. This approach fosters ownership and obtains the buy-in of people who will eventually use it.

Transparent change stories, clear success metrics and supportive learning environments where even trainers get trained are crucial in expediting adoption. Where end-users include consumers, vendors or front-line workers, a continuous AI-usage support system may be needed to encourage timely adoption by each individual.

Applying these three learnings—from project selection and set-up to adoption—could help companies significantly boost their chances of AI adoption making a transformative impact. This could be the difference between gaining or retaining market leadership or falling behind in a world where AI-native competitors are on the rise. As businesses consider their path with artificial intelligence, an even more fundamental question needs to be addressed: Should they try and retro-fit AI into their core, or should they build new clean-slate ‘AI-native’ units from scratch?

The authors are, respectively, a partner and a product expert at McKinsey & Company’s Mumbai office.