The market is booming for artificial intelligence (AI). Inflationary pressure and interest rates driving the cost of capital are among the current economic realities growing the appetite to mine our businesses for efficiencies. 

Generative AI (GenAI) has provided the impetus for businesses to realise the power of AI to transform entire families of tasks and job functions. With our proprietary GenAI Impact Index, we looked at the tasks that underpin 8,000+ job families, assessed the extent that elements, repetitive tasks within those roles could be augmented, accelerated or innovated with emerging AI capabilities. This approach is not new. It started with Henry Ford’s production process and has expanded beyond factories to offices. Recent thought leadership on the future of work estimates between 83 and 300 million jobs will be affected over the next five to 10 years, as reported by the World Economic Forum and Goldman Sachs.

But AI cannot do everything, as most AI systems have a narrow focus and cannot perform multiple tasks. For example, a chatbot can tell you how to cook a steak, but it can’t cook it for you. Multi-tasking will eventually be solved, but leading experts believe AI won’t replace tasks requiring some innate ‘human-ness’, as prominent computer scientist Stuart Russell says in “Living with AI: AI in the economy”.

AI may also require human intervention, since it won’t be 100% accurate. In a recent notable example, GenAI “hallucinations” wiped $100 billion from Google’s market value when it shared inaccurate information in a promotional video. This provides solid evidence that, although some industries and jobs will be more transformed than others – as we have seen with the writers’ strike at Universal – machines will not fully replace humans. The need for "humans in the loop" is evident. Humans will be enhanced with AI, leading to the emergence of new roles and responsibilities to ensure human oversight and value in automated tasks.

In this article, we address those questions centred on the future of work.  We outline the dominant use cases that we expect to be filled by GenAI, and why “no-regret” ones focus on Contact Centres and Customer Service, the Software Development Lifecycle, Content Creation (Marketing and Back office) and Document Summarisation.

1. Contact Centres and Customer Service

“Just Google it”. One of the most transformative parts of the World Wide Web is the instant access to unfathomable amounts of information in a huge number of languages, accessed through our refined way of using search engines, adjusting keywords to find the answers that we need. 

Now, what about: “Just ask ChatGPT”? Whereas Google simply fetches results, ChatGPT answers your query directly. They both have access to the same data from the World Wide Web, but ChatGPT understands the intent of your questions and provides answers in a conversational style that is typically easier to understand and gets you closer to providing the answer in the format you need. It is ‘generative’ – in that it generates your answer whereas with Google you still need to construct an answer from the search results. 

For many first-line support queries, this Q&A interaction is all that is required by the customer. This automation already exists to a degree, but GenAI catalyses a leap forward in customer engagement and specifically Chatbot performance. Customer experiences are hugely improved by removing wait times, responding in your preferred language, in your preferred conversational style, and providing as much information as you would like. Its access to larger data stores, including your history, moves it from responding from a library of answers to delivering the best answer for you – and if Meta’s chatbots with personas move forward, even as ‘someone’ you can relate to.

During the learning process, if the chatbot can’t answer your question, then you will be passed to the human second-line support while the learning process continues, enriching the effectiveness of the chatbot.

Numerous studies, articles and case studies highlight the potential of GenAI to make Contact Centres more effective. Studies claim up to 30% cost reduction in Call Centre Operations. Gartner claims that by 2026, conversational AI deployments within Contact Centres will reduce agent labour costs by $80 billion, with one in 10 agent interactions automated. 

Octopus Energy is already using GenAI to reply to email based communication with customers. Its CEO said “Gen AI models are doing the work of 250 people by answering customer emails and had received an 80% satisfaction rate, higher than the 65% achieved by workers.” 

We think that chatbots will be adopted quickly, easily, and more widely because of their early adoption in Contact Centres. Therefore, they form one of our “no regret” use cases. However, one question that surfaces in their deployment is who should oversee them? As chatbots can extend across customer engagement and experience touchpoints to ingest data, derive richer insights, feed those insights back into touchpoints, and prompt actions to various functions, how should the existing governance model change to manage this holistic view?

2. Software Development Lifecycle

AI can’t do it all and neither can humans. But this is nothing new. Humans have been working with technology since the invention of the first tool. AI is a natural next step in this relationship. For some time, we have been seeing our clients take advantage of machines’ ability to sift through large volumes of data with decisions made by people to deliver business value. In a June 2022 article, our European Head of AI & Data, Amelia Green, highlighted the $1 billion investment by Microsoft in OpenAI (before they launched ChatGPT) that was being used to build a “virtual companion to help engineers to write code”. This became GitHub Copilot.

The software development community has embraced GenAI as an additional tool. It is one of the most collaborative communities, where 96% of all code contains open-source code, further supported by dedicated forums. Content and code are shared for free and posted online, which is a key enabler for the training of GenAI models. ChatGPT and Copilot are used as virtual companions to help developers write code. Developers go back and forth between the suggestions from GenAI and the code that they write, drawing on all the expertise available online. As a result of the collaboration with GenAI, code development time can be reduced by up to 40%, and significantly aid the creation of documentation and translation of code into other languages. 

GenAI isn’t just used for software development and documentation in the Software Development Lifecycle (SDLC). It is also used in app building, testing, deployment, and monitoring. Some apps have seen 95-97% reduction in time to launch (from 60 days to 2-3 days). Testing under certain conditions has improved by 90%, with 3x faster tests and 40% reduction in bugs during production. Deployment can also be faster, with test time cycle times reduced by up to 98%. Monitoring is more robust, too, with examples showing a reduction of 75% to the mean-time-to-recovery.

We believe that because of the way that the software development community has embraced this technology, GenAI will continue to be adopted quickly and impactfully, making this another of the “no-regret” use cases.

3. Content Creation (Marketing and Back office)

GenAI is good at repetitive tasks; it can process vast quantities of information far faster than human capabilities and it can make sense of complex datasets quickly. Currently, we use humans to perform many such tasks in back-office functions such as Marketing or Human Resources (HR). For example, GenAI was famously used to create a new advertisement for Coca Cola in record time, and its potential to create marketing content tailored to potential customers based on social media activity, combined with purchasing patterns, is being explored. This could be applied to each individual customer automatically, albeit curated by marketing personnel. 

In the meantime, GenAI is used to develop winning product descriptions, optimise SEO, and accelerate AB testing of the most engaging content. We can reimagine other evolutions in back-office functions; imagine matching job seekers to job vacancies based on their detailed experience and published online portfolio together with a detailed job description based on Project Management notes, instead of a one-page CV and a one-page job description.

We think that GenAI’s capacity to generate highly customisable content at scale aligns exceptionally well with the nature of tasks in back-office functions – making it our third “no-regret” use case.

4. Document Summarisation

We highlighted that SDLC and some back-office functions use GenAI for document summarisation and documentation. For example, the UK-based law firm Allen & Overy has partnered with Harvey AI, which assists “lawyers with research, drafting, analysis, and communication”.

HarveyAI helps streamline tasks and reduces response times, which research has found is a key factor for client satisfaction. GenAI is able to quickly read large volumes of text, quickly summarise them and communicate the key points. This is the type of task that was performed by paralegals or business analysts. Using GenAI to perform these tasks allows humans to make decisions in a much more efficient and more informed manner.

We think that document summarisation and documentation are ideal use cases for GenAI, and they form the fourth “no regret” use case.


The strongest initial adoption of new technologies is seen when it is an upgrade to the existing solution. For example, chatbots are already used in first-line support. GenAI adoption there is fast because there is already a degree of comfort established, with a system set up to make the integration of GenAI straightforward. Gartner expects that, by 2024, 40% of enterprise applications with have embedded chatbots. We can draw parallels with the advent of electricity, which quickly replaced steam as a power source in factories.

However, electricity took more than 40 years to become commonplace in factories and homes. This evolution was only possible once changes to the wider system had occurred – devices had to be redesigned so that they could be powered by electricity. The same is true of GenAI. Systems, processes, and operational models will need to be redesigned so that they can be powered by GenAI. 

Successful adoption of GenAI in business comes with common challenges:

  • Balancing Cost and Value: Determining the right models to deliver value, while factoring in their associated costs, e.g. compute and storage costs.
  • People: Ensuring that the company's culture, organisation strategy and support systems align with GenAI adoption, including employee training. 
  • Information: Collecting, managing, and preparing training data tailored for GenAI use.
  • Operations: Having accessible secure, scalable IT environments that empower efficient use of GenAI outputs. 
  • Managing Risk: Dealing with new risks like cyber, privacy, ethical, and societal associated with GenAI use. We are entering a renewed 'digital arms race' of development and benefit vs risk and regulation. GenAI carries with it new risks, such as hallucinations and data privacy concerns.
  • Regulation: Abiding by emerging regulations as governments and industries address responsible use of GenAI. Companies will need to abide by new legislation. Currently, legislation is being written and task forces are being assembled. The UK recently pledged an initial £100million to establish a Foundation Model Taskforce that will act as a global standard bearer for AI safety. 

Many companies can enhance their technology with GenAI in specific areas, such as chatbots. Some will already have systems in place to use GenAI more broadly, such as software development. However, true widespread adoption and reaping the full benefits will only happen when we have the right systems in place.

And herein lies the next challenge: defining and implementing the essential capabilities for effective risk-managed AI, where the breadth of capabilities will vary based on the use case and organisational maturity.