In this second article, I build on a common understanding of key GenAI terms as defined in the first article of this series (Part 1 explanation of GenAI) to explore the various applications for GenAI and how it is likely to impact our work and social lives.

Increasing amounts of work activity is subject to automation

Recent McKinsey studies on the future of work estimate that in 2023 some 60% (with the potential to move to 80% by 2030) of existing work task hours could be automated using the technology available at the time of analysis. This high level of automation potential is despite their estimate of some 40% of activities that workers perform requiring at least a medium level of human understanding of natural language. GenAI ability to work with natural language means that many work activities that involve communications, supervision, documentation, and interacting with people in general now have the potential to be automated. Naturally, this has the potential to transform many areas of the economy with estimates that some 10% of the Australian workforce (1.5m people) will need to transition from current roles into new occupations by 2030.

GenAI has the capability to change how work gets done at the activity level

GenAI has the capability to automate, augment, and accelerate how work is performed. It has the ability to perform a range of functions including; classifying, editing, summarising, answering questions, and drafting new content. Each of these actions has the potential to create value by changing how work gets done at the activity level across business functions and workflows. In equal measure, it has the capability to both remove and enhance the role that humans play in undertaking work.

This transition in how work gets done will be reflected in transitioning job descriptions

While there remains some divergence in views to the extent that AI is coming for your job there seems less divergence that at a minimum it is coming to alter your job description. Job descriptions are transitioning from being more ‘people centric’ – people executing processes presented with data powered by technology, to more ‘technology centric’ job descriptions – with technology powered by data executing processes managed by people.

Generic application types that will underpin GenAI value cases for adoption

There are three generic application types that will underpin the value proposition for GenAI adoption that consist of;

  • Analyst – the ability to improve time and quality of searching, navigating, and extracting insights and understanding from complex and unstructured natural language data,
  • Service – the ability to improve customer experience by making digital interactions more personalised, natural, conversational and rewarding,
  • Creative – the ability to generate text, image, code, video and music quickly and multi-modally thereby speeding up creative processes and maximising productivity, and

Practitioners of GenAI will use these base value propositions to Assist (speed and quality) the execution of their work, Create new content that is an extension of the existing knowledge/capability base, and to Explore new concepts by mixing new ideas and concepts that don’t necessarily need to be true (ACE).

GenAI will be integrated into platforms and tools already used in everyday life

In several areas, individual’s choice around adopting GenAI will be taken away from them as capability gets embedded in platforms and tools that already form part of everyday life. This type of adoption is much easier than having to create new habits. We are already seeing key software and platform providers (Microsoft, Google, Apple etc) embed GenAI capabilities into their products and such upgrades have the potential to substantially increase individual productivity. For example, Email systems are providing an option to write the first drafts of messages. Productivity applications will create the first draft of a presentation based on a broad description. Financial software will generate a prose description of the notable features in a financial report. Customer-relationship-management systems are suggesting better ways to interact with customers.

Given complexity and governance early use cases are likely to be inward versus outward facing

While new transformative use cases are beginning to occur across different functions and industries a likely early focus for GenAI applications are work routines internal facing to organisations where the management of risk, ethics, and governance can be matured in more controlled environments. For example, GenAI can revolutionise internal knowledge management systems. Studies suggest that knowledge workers can spend up to a fifth of their time (i.e. one day each work week) searching and gathering internal information. GenAI’s impressive command of natural-language processing can help employees retrieve unstructured stored internal knowledge by formulating queries in the same way they might ask a human a question and engage in continuing dialogue. This could empower teams to quickly access relevant information, enabling them to rapidly make better-informed decisions and develop effective strategies. This knowledge can be equally harnessed for the benefit of customers if leveraged through customer service interfaces.

GenAI applications will be biased towards particular functions

While GenAI is still in its infancy, transformation use cases that offer practical benefits for jobs and workplaces are already emerging. Given the nature of the technology there are a range of functional areas that have the inherent potential to deliver higher business cases relative to other functions and fall in the areas of; customer operations, marketing and sales, software engineering and R&D. It is not surprising that these functions are more prospective than others such as manufacturing and supply chain where the numerical and optimisation applications were a major focus in the early evolution of AI capabilities.

Use cases can be thought of sitting along a spectrum of differential investment and returns

Early use cases are naturally being filtered out based on the level of upfront investment required both financially and in the broad range of risk governance areas associated with the technology. Potential use cases can be thought of along a spectrum with at one end the ability to adopt off-the-shelf GenAI solutions requiring little or no in-house customisation to at the other end of the spectrum requiring building proprietary LLM’s to meet specific applications. Populating mid spectrum are generating proprietary GenAI models off the back of (through API interfaces) public LLM and doing specialist customisation or fine tuning of LLM to meet proprietary needs.

Software engineering is particularly suited for off-the-shelf products

Changing the work of software engineering is probably the most profound off-the-shelf GenAI solutions emerging in the early phases. Treating computer languages as just another language opens up immense possibilities to speed up developers’ code generation. Productivity improvements of 20-45% of current spend has been identified by reducing time on certain activities such as generating initial code drafts, code corrections and refactoring, root cause analysis, and generating new system designs.

Sales and marketing applications are exploding off the back of publicly available models

An area of aggressive introduction of new applications leveraging the foundation models, but requires another step up in investment, is in sales and marketing where text-based communications and personalisation at scale are the driving forces. The technology can create personalised messages tailored to individual customer interests, preferences, and behaviors, as well as do tasks such as producing first drafts of brand advertising, headlines, slogans, social media posts, and product descriptions. In the area of sales, improvements in lead generation and sales conversion are supporting growth in revenues.

Fine-tuning and customisation of foundation models is being adopted in customer service applications

With more investment, that will generally be directed at fine-tuning foundation models, GenAI could revolutionise the entire customer operations function by either automating or augmenting the customer service operations. Model can be fine-tuned to focused customers, segments, and specific questions and answers if needed. This application is gaining traction because of GenAI’s ability to automate interactions with customers using natural language and by enabling less-experienced customer service agents to communicate using techniques similar to those of their highest skilled colleagues.

Creating new foundation models has the potential to supercharge R&D applications

The deepest investments are likely to require the build of a foundation model from scratch if no existing models are available to meet the use case. In general, training a model from scratch costs ten to 20 times more than building software around  existing models with API interfaces. Larger teams (including, for example, PhD-level machine learning experts) and higher compute and storage spending account for the differences in cost. An additional return on investment is required to offset the financial and human capital costs to generate new models. While applications of GenAI are less well recognised in research and development (R&D) the opportunity for significant productivity improvement exists. For example, the life sciences and chemical industries have begun using GenAI foundation models in their R&D for what is known as generative design. Foundation models can generate candidate molecules, accelerating the process of developing new drugs and materials.

There will be differential impact on occupations based on GenAI inherent value case and adoption rates

Given the difference in use case and relative investment required, it is likely that when looking at future demand for jobs, and potential occupational transitions, three distinct occupational groups will emerge. There will be those occupations, such as science and technology, healthcare, and financial/professional services, that are likely to be both resilient and experience growth in demand. There will be other occupations that may maintain their share in the workforce in response to population growth and growing infrastructure demand. The third group will likely be disrupted and subject to declining job demand such as office support, customer services, sales, and programming.

The rate of adoption may not match theoretical GenAI capabilities due to human change management

However, history tells us that there is often a (substantial) lag between the theoretical capability and actual adoption levels – developing capabilities into technical solutions takes time, the cost of implementing solutions may exceed the cost of human labour, and the pace of adoption could be influenced by social or regulatory dynamics. Given the high levels of investment to develop large scale multi-purpose foundation model, it is likely that the GenAI ecosystem will be dominated in the early stages by the large tech giants (Microsoft, Google etc). However, there remains a very vibrant start-up community that is focused on building smaller models that can deliver effective results for some tasks bringing added benefits of enabling model training to be more efficient.

The implications for business share similar fundamentals to earlier technology adoptions

For businesses thinking about GenAI, fundamentals are very similar to any new technology implementation. Activity should broadly be guided by the organisation’s strategy and whether application is targeted at disrupting or augmenting existing practices and whether it is broad or narrow. Identifying suitable use cases to build knowledge and capability (lighthouse use cases) and to test theory with practice usually follows. Identifying the required capability (capital, technical, human resources) to deliver the use cases ensures the practical execution matches desires. Finally, managing risks with suitable governance is a critical precursor prior to broad adoption and scaling.

Some common key success factors for those organisations with proven track record

Businesses that tend to succeed share some common success factors including putting humans not technology at the centre of the transformation, learn to convert efficiently innovative MVP’s into scale applications, and recognise the value in applying technology iteratively at speed to form final solutions. With proper guardrails in place, GenAI can not only unlock novel use cases for businesses but also speed up, scale, or otherwise improve existing ones.

I look forward to you joining me for the final instalment of this GenAI for the layperson series where I focus on the implications of GenAI for society, business and individuals.