Generative AI

By Ram Gopal

The AI of the beholder: Managers need to examine where AI can improve workflow processes.

Generative AI’s impact on productivity could add trillions of dollars in value to the global economy, according to consultants McKinsey.

Generative AI can learn, reason, make decisions, and automate many creative and intellectual tasks – providing the potential to supercharge productivity in the workplace.

As a starting point, companies need to identify where AI can help. This involves tasks within specific company functions, so the process needs to begin from the bottom up, rather than top down.

Every manager and decision-maker should analyse their team’s workflow to identify tasks that AI can help with. Tasks that AI can do alone should be automated, although this is a nascent area due to errors, so supervision is still often required – but is likely to expand rapidly over the coming years.

A step up from this, there are tasks where managers can learn to work seamlessly with AI, integrating it into their workflow and passing work back and forth (termed ‘centaur tasks’).

Then there are the tasks that can be delegated under tight oversight, including checking for errors and giving feedback. Both centaur and delegated tasks are likely to expand as AI becomes more flexible and can tackle more parts of the workflow.

Beyond this remain the ‘just me’ tasks that employees either don’t want to give up, where the AI is not capable, or it is too hard to get the AI to do a good job.

From the company’s perspective, managers then need to be empowered, resourced, and enabled in their use of generative AI. Firms need to be more proactive in this, rather than defensive – those that wait and see will lose out to competitors that jump in.

This includes establishing guardrails, best practices, and a sandbox for experimentation. Companies should plan for bespoke Large Language Models (LLMs) – machine-learning AI that uses neural networks trained through the input of data. For the best results the LLM should be tailored to the firm’s field, as has been done with LegalBERT, BioGPT and others, and this may require significant investment in IT and data acquisition.

LLMs can respond to questions, summarise, edit, translate, program, assign labels, extract facts, infer relationships, and reason about texts using learned knowledge. The larger and deeper LLMs enable the emergence of greater understanding and a higher level of response – in a similar way to humans as they accumulate knowledge growing up.

In addition to saving managers’ time and enhancing the way they do their job, AI can identify savings and streamline processes and procedures, as well as monitor performance and personnel.

More specifically, managers are likely to find that the help available from generative AI falls into one of the following six categories:

1. Helping with research 

AI can be used in most areas of research and analysis. For example, it can be used to look up and summarise patents, and search for potential customers or investors.

This can be extended to the provision of cost estimates to develop and produce new products, as well as the risks involved in their commercialisation. AI can also carry out research and analysis on competitors or potential acquisitions, including market positioning, customer satisfaction, service offerings and risk.

Based on this, it can go on to analyse and compare the positioning of companies in relation to current market trends and demands, and how they may be able to enhance customer satisfaction and optimise service offerings.

Managers partnering with the AI need to watch out for accuracy and originality of responses, or potential bias that may have been written into its system or through the information it digests.

2. Brainstorming ideas 

AI is very good at idea generation, including in the areas of planning and strategy, which can be hugely beneficial.

It can be creative, coming up with ideas for product logos, names and design. It can be employed in artistic design as well as text generation, helping with the donkey work in the process.

Concerns over low quality in this area are fast receding, with analysis now unable to distinguish human and AI created content – indeed, there is growing evidence to suggest AI creativity is of a high quality. The worry about losing the human touch appears to be fast receding, but this is a concern that users should always be aware of.

3. Supporting content creation 

Once ideas have been settled on, AI can adapt them to fit with brand identity or other requirements and can help develop marketing content.

Human feedback can be provided to the AI to help re-work particular aspects of this process. Managers must be wary of the possibility that AI may share sensitive or confidential information and adjust for this.

They are excellent in creating customer content, including writing about specific customer needs, and this applies to every industry.

4. Writing and understanding computer code 

AI can program in whichever language is required given the right LLM, which can enhance operational flexibility and cuts time and resources from programming tasks.

Online it can interact seamlessly with other computers and AI, as well as being able to conduct full analysis of websites and their performance, along with the quality and extent of their SEO, security and visitor interaction.

5. Automating interactions with customers 

Among the more immediate applications of generative AI automation is an improvement in the quality of automated interactions with customers – which should provide enhanced customer service, replacing clunky old chatbots.

With AI’s ability to ingest all the information relevant to that company and customer, the field should see rapid improvement.

However, proprietary AI may be required here, with considerable security to ensure customer safety. But it should provide more meaningful, friendly and more useful interactions with consumers, across a wide range of sectors.

6. Translating languages 

AI is now able to translate and interpret in whichever language is required to a high degree of accuracy and fluency.

Machine translation is a rapidly developing field, with many apps and websites available, such as Google Translate. There is also Neural Machine Translation (NMT), which sees a machine translation service pairing with an artificial neural network to provide better outcomes than standard translations.

For companies exporting or looking into new international markets this could be a valuable service and any translation can be cross-referenced with another translation app.

Looking to the future, possible developments include open source LLMs, as well as highly personalised AI put together for the needs of individual managers.

Moreover, AI enables everyone to be a manager and hugely increases productivity, which could have huge implications for organisational structure and strategy in the longer term.

Higher productivity could lead to lay-offs or re-direction of work to new areas. Organisational structure could change to one where individuals predominantly use AI independently – a leveller that may alter the dichotomy of management and workers, changing the traditional hierarchical nature of companies. Or the structure may involve a central AI monitoring every move being made, making control more centralised than ever.

AI itself will no doubt eventually have a suggestion of its own.

Learn more about WBS Digital Innovation & Entrepreneurship Executive Education courses here: https://www.wbs.ac.uk/courses/executive-education/?studytheme=Digital+Innovation+and+Entrepreneurship

About the Author

Ram D. Gopal

Ram Gopal is the Information Systems Society’s Distinguished Fellow and a Professor of Information Systems and Management at Warwick Business School. Learn more about WBS’ Digital Innovation & Entrepreneurship Executive Education courses here: Executive Education | Courses | Warwick Business School (wbs.ac.uk)

LEAVE A REPLY

Please enter your comment!
Please enter your name here