The future of AI for business is human

Earlier this autumn, games designer Jason Allen entered a picture “Theatre d’Opera Spatial” into the Colorado State Fair fine arts competition. It won first prize in the digital arts category.

Nothing unusual there. Except that Mr Allen didn’t really create the artwork. He used Midjourney, an artificial intelligence (AI) text-to-image generator. Did he cheat, or was he simply demonstrating the power that AI has to take over human tasks, even those involving substantial human creativity?

AI is increasingly transforming organisations across all sectors of the economy. It’s used to improve and automate processes, to strengthen customer relationships, to generate predictive analytics, even to support human resource management.

Three-quarters of business executives anticipate that AI will make business processes more efficient, as well as creating new business models. It is no surprise then that, despite economic uncertainty, funding is still flowing into AI start-ups, with annual investment continuously increasing since 2014, and total investment reaching $77.5 billion in 2021.

Getting AI right

However, it is not simple for businesses to use AI effectively. According to Stuart Robertson, Partner at L.E.K. Consulting, a successful AI project requires four things: the right problem, the right data, the right expertise, and the right organisational structure. If any of these isn’t right, the AI project is doomed to failure.

Having appropriate goals (the right problem) means applying AI to problems of high business value, where AI techniques can extract maximum value from data. However, this is an increasingly rare issue for mature organisations, many of which have been exposed to AI-powered services for some years.

At the same time there is a need to ensure appropriate resources (the right expertise, the right data) are in place. AI skills and experience may be in short supply, but they are readily available to organisations who can pay. And most organisations “run on data”, meaning that the issue isn’t usually having the right data assets, but rather deciding on how to best use AI to analyse it.

But even with addressing the right problem, using the right expertise and data, many organisations fail to deliver real value from AI. That’s because putting AI technology and data scientists together in a room is not going to deliver magic solutions to business problems. You need to change the way you are working.

The right organisational structure

As well as investing in AI technology and skills, organisations also need to invest in restructuring the way that they work. Stuart Robertson warns: “In our experience, the most common point of failure is at the organisational configuration level.”

Reconfiguring the organisation so that it works effectively with AI does not have to be complicated. AI should not be regarded as a separate discipline in a specialist knowledge silo. Instead organisational roles and processes need to be adjusted so that people can work alongside the algorithm, using it to support their decisions on a day-to-day basis. AI should be integrated into workflows, rather than being treated as an external support tool that people use from time to time, when they feel like it.

Treated in this way, AI can enhance the way that people work - reducing risk, increasing efficiency and the quality of decisions, and enriching job satisfaction.

Managing the outcomes of AI models

To be seen as successful, any investment in AI needs to demonstrate improvements to business outcomes, effectively (and cost effectively) solving the problem it was set up to address. This requires constant monitoring of the quality of the decisions that the AI system delivers and their effects on the wider organisation.

For example, while the AI algorithm may have been developed using highly structured processes and trained using highly representative data, over time models can still drift as small errors get repeated and amplified, or circumstances change compared to the data it was trained on. A “human in the loop” is needed to retrain, test, tune and retune the model as it is used.

It is at this interface between the AI and the human where some of the most difficult AI management problems are to be found, and their solutions can only be developed by humans.

The future of AI technology

Ever-more powerful computers are improving the effectiveness of AI systems and creating new ways of delivering support to decision makers. Quantum computing, which allows calculations to be undertaken simultaneously rather than sequentially, may become commercially available by the end of the decade, accelerating this process exponentially and enabling complex problems, predictions and simulations to be calculated in seconds rather than years.

AI is only going to get more useful and universally applicable for enterprises. More and more businesses have gone beyond experimenting with it and are now using it in earnest to improve business outcomes.

And they are probably doing better than they think they are. It’s easy to undervalue progress with AI because of the hype around AI in the media.  Any business that has made a start, whether that’s automating a process, building a product recommendation engine, or using AI to analyse media content around their brand, is on their way to success. New and better opportunities will arise as they learn from their experience and as technology continues to evolve.

Today’s AI is very much a controllable management tool. It needs the right techniques of course.  But more than technology and process, it needs people. People and AI systems working closely together represent a powerful future for business.

L.E.K. can help you define how to reengineer your organisation to take full advantage of the potential of AI. For more information please visit www.lek.com