Generative AI & human-like automation

Amidst all the hype and headlines generated following the launch of ChatGPT in late 2022, it would have been easy to think that AI was a brand-new technology. It is anything but. In fact, the idea of Artificial Intelligence goes back as far as the mid-1950s, says Alan Lavery of KPMG.

The increasing application of complex mathematical processes and ever improving technology capabilities has seen AI move through the realms of data science, big data and, since 2017, into an area of transformer models and neural network architecture, which has given rise to modern Large and Small Language Models (LLMs, SLMs).

The recent hype and surge in excitement over the past 18 months has been largely due to the opening up of the technology to consumer markets through the release of the ChatGPT application. This has opened the doors to AI consumption across entire organisations and not just specialised data teams.

The level of the hype generated has been quite extraordinary. We saw a mix of emotional response to the technology with many people becoming over-excited, yet others were driven to a state of caution and concern by their fears about how it could affect them personally as well as wider society.

Both feelings are valid, to a certain extent at least. AI’s potential is mind-blowing, and it will undoubtedly have transformative effects on how we live our lives and on the way we conduct business. That same power can and will also have negative impacts.

Already we have seen the Screen Actors Guild (SAG) in the US negotiate a clause in their new contract to prevent AI being used to create on-screen ‘characters’ based on data collected from real actors. Similarly, song and script writers have expressed legitimate fears in relation to the use of generative AI to effectively learn from and replicate their past work to create new compositions at an extremely rapid pace.

These things don’t have to be perfect all the time to be a real threat to humans, just occasionally is enough given the volume of material it can churn out in short time periods.

The speed of movement at present means that those who do not consider and adopt AI in their business risk being surpassed by those who do."

Alan Lavery

The pressure to adopt AI at pace

That power is also creating enormous pressure on business leaders to adopt the technology as quickly as possible. Again, this pressure is understandable. The speed of movement at present means that those who do not consider and adopt AI in their business risk being surpassed by those who do.

This is not necessarily always the case, however, and AI adoption should be approached with an open mind but with clear eyes and a degree of caution and control. The starting point is to look inside your business, at the challenges you face, the problems you are trying to solve, and the changes you need to make.

Quite often, when you drill down into some challenges you might find that they are best addressed through traditional data means (data engineering, management, quality control and governance) and don’t necessarily require AI adoption immediately to solve them. Good data engineering and management still forms the basis of trusted reporting, analytics and insights derived through AI.

In this light, business owners need to ask what AI is and how they can make it work for them. They need to measure its impact and establish an informed business case for it in the same way as they would for any other investment.

Measurable positive impact is key

They must also ensure they get value from the change they are implementing. Change for its own sake is a waste of time and resources, and change driven by pressure to use a new technology is worse than pointless.

AI adoption must have measurable positive impacts. Always front of mind must be whether it is tackling the challenge or not.

Some use cases are clear. For instance, in the retail sector, it can help to analyse shifting consumer habits and trends to enable better inventory/stock management and enhancing demand forecasting with positive effects on costs, supply chain performance and sustainability.

The development of LLMs and SLMs has given life to the massively untapped potential held in unstructured data: files, documents, audio, video, and imagery.

Learning from, accessing, querying, retrieving, and generating content based on this data is now delivering previously unobtainable and impactful insights straight to the hands of the user, at a pace that has never been felt before.

They can examine vast volumes of policy documents, contracts, regulations, operating manuals, and other pieces of data and find previously undetected anomalies, areas of risk, trends, and omissions, and in turn generate recommended actions and responses.

We must continue to view AI as an early-stage technology with enormous power that needs to be harnessed constructively.

Alan Lavery

Controlling without curtailing innovation

More specialised LLMs are now targeted as specific sectors such as legal, research, and code development, further driving enhancements to these skilled professionals. Their deployment always contributing to enhanced insights, efficiency, and new areas of business value.

Like most things in life, everything comes with risks, of course. Screen actors and songwriters aren’t the only people exposed to its potentially negative impacts. Combining AI with facial recognition technology and feeding it CCTV and drone footage to analyse has clear implications on civil liberties, for example. This is before we go anywhere near the topic of intellectual property.

There is a need for tight regulation and the EU AI Act and other pieces of legislation around the world will bring control without curtailing innovation, helping businesses, their workforce and the customers feel the benefits without the loss of trust in the system.

In the meantime, we must continue to view AI as an early-stage technology with enormous power that needs to be harnessed constructively. While LLMs can engage with humans in a transactional way, making suggestions for actions they might take or questions to ask, it still hasn’t quite progressed to the stage of autonomous problem solving or general intelligence, for now.

Human oversight remains crucial

Augmentation and enhancement is therefore its principal role at present; helping people and organisations get to the answer faster. It can surface and enhance better insights quicker and help users get to the answer in seconds rather than days. It can also carry out laborious tasks without ever getting bored or tired, helping us to be more effective. Ultimately, it’s about making humans and organisations better.

Training is an important consideration, and that’s where humans have to remain in control. Humans must form part of the adoption process throughout. If we are going to trust AI to make important decisions, there needs to be human oversight to ensure that it is making ethical choices and that the decisions are not damaging. AI will never replace human knowledge, connection, empathy and institution in the workplace.

But the technology is continuing to advance rapidly, and we need to be open to both its current and potential capabilities. By putting the correct governance and controls in place and beginning with low-risk test applications and building from there, organisations can adopt AI safely and obtain real benefits from it.

We must accept that, just as the internet and the smartphone changed our lives over time, AI will do the same and lead to the creation of entirely new business models that we cannot even conceive of today, e.g. AI to AI competitive procurement. The organisations ready for those changes and who adopt AI in a controlled and safe way will prosper, those that do not may find their viability threatened in a relatively short period of time.

At KPMG, we are working with organisations to help them think through what AI means for them, develop strategies for its adoption, put governance and controls in place, scale solution sensibly, and ensure business leaders get real value from their investment. Whatever the goal is - growth, expansion into new markets or territories, new customer acquisition, mergers and acquisitions, we help them decide if AI can help, and if it can, how to use it in the right way.

Get in touch

At KPMG we understand the pressure business leaders are under to get it right on tech and AI.

To find out more about how KPMG perspectives and fresh thinking can help your business please contact Alan Lavery of our AI team. We’d be delighted to hear from you.