Generated by Midjourney v5 prompt "Corgi using chat gpt"
*Pithy title generated by GPT 4
Summary
This report is a follow on of the analysis of Technology & Innovation Management capabilities within Cochlear for Large Language Models, and specifically Generative Pre-Trained (GPT) Models.
We perform a Technology Assessment focused on mitigations for employees, customers, business users, suppliers, the environment, and communities.
This analysis is based on the PESTEL (political, economic, social, technological, environmental, and legal) model, but omits political considerations as out-of-scope.
We identify the core business stakeholder functional groups affected and recommend specific mitigation strategies to address any impacts and risks through a Solution Design process.
Technology Description
Large Language Models (LLM) are artificial intelligence neural network models that use machine learning algorithms to process and generate natural language text. Trained on vast amounts of structured text data, unstructured video, audio, and online content, and can generate human-like responses to text-based inputs, such as questions or statements. (Xu et al., 2022)
LLMs are increasingly used in various applications, language translation, chatbots, and virtual assistants, to enhance natural language processing capabilities and improve user experiences. These models display emergent cognitive facsimiles and are increasingly finding new applications across traditional human white-collar functions.
Technology Forecast
LLM technology, such as GPTs, will continue to advance and become more sophisticated in the future. There will be improvements in language processing capabilities, such as a better understanding of context and nuance in text-based inputs, and a greater ability to generate more human-like responses.
These technologies are also expected to become more integrated into various applications and devices, such as smart home assistants, virtual reality and augmented reality systems, and mobile devices. Additionally, LLMs will play a growing role in industries such as healthcare, finance, and education, to enhance natural language interactions and automate tasks.
The future of LLM technology is likely to be characterized by continued innovation and increasing integration into various aspects of our daily lives.
Problem Definition
LLMs will have impacts on existing workforces related to the potential displacement of human workers as these systems become increasingly capable of performing tasks traditionally done by humans, particularly in customer service and related fields.
The adoption of LLMs in these areas has the potential to streamline operations, reduce costs, and improve customer satisfaction, but it may also lead to job losses or reduced job security for human workers. While LLMs can provide fast and accurate responses to customer inquiries, they may lack the empathy and understanding that human workers can provide. This could lead to a decrease in customer satisfaction and loyalty, particularly if customers feel that they are not being heard or understood.
These impacts on existing workforces will necessitate the need for retraining and upskilling of human workers to adapt to the changing nature of work. This could require significant investment in training and development programs to ensure that workers have the necessary skills and knowledge to remain competitive.
Societal and Business Context Descriptions & Forecast
The wider societal context is largely positive, with the term artificial intelligence first coined by John McCarthy in 1956 when he held the first academic conference on the subject. The journey to understand if machines can think began much before that. In the seminal work As We May Think (Bush, Vannevar. 1945) they proposed a system that amplifies people’s own knowledge and understanding, which was based on the model of artificial neurons by McCulloch and Walter pits in 1943.
Five years later Alan Turing wrote a paper on the notion of machines being able to simulate human beings and the ability to do intelligent things, such as play Chess. (Smith & McGuire, 2006). As AI technology develops, the boundary line of whether human-level intelligence has been surpassed by artificial intelligence or not keeps being pushed back. Currently, AI far surpasses human ability in many narrow-focus fields, and activities.
The social business context of these technologies will be both positive, in the creation of additional value and force amplification of existing employees, and negative in both potential employment redundancies, and upskilling requirements.
Business Function Stakeholders
Based on the technology analysis performed previously, it was identified that the following functions would potentially provide the most value with LLM technologies.
In addition to these previously identified stakeholders, these are the other stakeholder groups that will be affected by the adoption of this technology.
Design Solutions
Technology assessment and Stakeholders’ Impact mitigation
Some of the possible implications this technology will have for customer services and the different implications for various stakeholders involved are easy to predict, however, there may be several unknowable future impacts - “Known unknowns” and “Unknown unknowns”. (NASA. 1981) (Rumsfeld. 2002).
Employees: LLM can automate many repetitive and time-consuming tasks, freeing up employees to focus on more complex and meaningful work. Effectively this technology will act as a force multiplier for staff, allowing each person to add more value and to be able to focus on where best to add that value.
Employee training and upskilling can also be significantly accelerated using GPT technology.
This technology can provide customer service staff with valuable data and metrics to monitor and evaluate the quality of customer service. Customer service staff will still need to use this information in analytics to identify opportunities for improvement and to take action to improve the quality of customer service.
Some staff may be concerned about the potential loss of their jobs or a reduction in their responsibilities. Wherever possible roles should be rounded out and given more scope to apply respective skills towards the best possible outcomes for customers and employees.
To mitigate this Cochlear should communicate clearly with our employees about the benefits of LLM and provide adequate training and support to help employees adapt to the new technology and enhance their skills.
Should there be any need for staff reduction in certain domains, up-skilling and cross-organisational employment opportunities should be explored, or alternately generous redundancy allocations made.
Business: In this context, Business stakeholders would be the corporate and technology functions of Cochlear.
For the corporate function, there would be a significant upside in both cost reduction and customer benefit translating to better user experiences across both Recipients and professional medical customers. There would be a need to perform a more detailed impact analysis for the low-level interaction points such as account management processes, analytics, feedback channels and escalation pathways, to validate that due diligence has been performed for both corporate and medical regulatory reasons.
For the technology function, there would be a significant addition of complexity in both the physical layer (servers, cloud instances, software) and for technical staff and knowledge uplift. The cost and skills of maintaining the LLM infrastructure would require large initial upfront investments in technologies, training and outsourcing support.
Customers: The Customer stakeholder group will be divided into Recipients (including Candidates) and Professionals, such as Clinicians, Clinics, Audiologists and Hospitals.
The implications for Recipients are mostly positive. Improved response times and accuracy of customer service, leading to better customer satisfaction. Better personalized service by analysing customer data and understanding preferences and needs.
For Professional customers, the impacts will be in supporting both the speed of training and the level of indirect support that they offer Recipients. There would need to be a larger component of training, and change management for this group.
Customer service staff will still need to manage the more complex aspects of account management, such as addressing customer concerns or handling escalated issues for Recipients.
Technical staff will need to manage the more complex aspects of technical issues and medical concerns of escalated issues for Professionals.
Total Quality Management (TQM) should be applied throughout the process of adopting this technology, and to make sure that any negative impacts or “Unknown unknowns” are addressed as soon as they are identified.
A supplemental increase in marketing and audience awareness activities would be beneficial to both promote and mitigate any effects.
Suppliers: Suppliers such as clinics and hospitals will be affected by the introduction of LLM technology in customer service. There will be process and training changes needed to adapt their communication and collaboration methods with the organization. Overall the Supplier group may be the least impacted of all stakeholders, as this proposal has little to no direct changes in either the relationship, nature or processes to this group.
Environment: LLM technology is unlikely to have any direct impact on the environment, although it may indirectly contribute to reducing the carbon footprint by enabling remote work and reducing the need for physical travel.
Unfortunately, this is offset by the energy costs of running the technology, such as data centres and information technology infrastructure. To train a model such as GPT 3 has the equivalent energy and CO2 impacts as a return flight to Australia, at a cost of approximately $5 million USD. (Patterson, 2022). Carbon offset credits could be purchased.
Communities: The introduction of LLM technology in customer service can have positive impacts on communities by creating job opportunities in technology-related fields and supporting local economic growth. However, there may be some concerns about the potential impact of technology on job losses and the replacement of human workers with machines. A marketing campaign to expand awareness and capitalise on public perception in a positive manner should effectively mitigate any negative sentiment.
Business Impact and mitigation matrix, ranked by magnitude of the anticipated impact
Mitigation strategies
Change management program
A structured approach to planning, implementing and managing change within an organization. It involves assessing the impact of the change, developing a plan for implementing the change, and managing resistance to the change. The goal of a change management program is to ensure that changes are implemented smoothly, with minimal disruption to operations and with maximum buy-in from stakeholders.
Training programs
Educating employees about the use and capabilities of this technology. Would include information about how LLMs work, how they can be used to improve business processes, and best practices for incorporating LLMs into existing workflows. The training program includes hands-on exercises or simulations to help employees become comfortable in real-world scenarios. The goal of the training program would be to ensure that employees have the knowledge and skills necessary to effectively use LLMs to improve business outcomes.
Total Quality Management (TQM)
Total quality management (TQM) consists of organization-wide efforts to install and create a permanent climate where employees continuously improve their ability to provide on-demand products and services that customers will find of value. This methodology uses frequent user feedback loops to close out any quality issues.(Ciampa. 1992)
Parallel operations
Changed processes are run in parallel with existing processes for a period until the number of issues reported becomes zero or are easily and quickly mitigated. This carries increased costs for its duration, however, provides safety for mission-critical demands.
External support services
Consultancies and vendors provide the required technical skills and support to bridge any requirements that are not in-house, or that will require additional support, to the business.
Impact assessments
Specific impact assessments to attempt to flush out any “Unknown Unknowns” through low-level process analysis, time studies, and feedback mechanisms.
Policy and Legal considerations
This proposal assumes that all legal liability considerations are provisioned in the vendor solution requirements and that any policies would be directed by external expert consultancies mentioned previously in this report. As such these are outside of the scope of this document.
Conclusion
The benefits of the implementation of LLMs within Cochlear could have a meaningful impact on customer satisfaction, reducing costs and further advancing market leadership ahead of any challengers.
These benefits are weighed against several risks across almost all stakeholder groups, internal and external to the business, which can be effectively mitigated with an ethical approach to using this technology to augment existing functions and staff.
This technology will develop and improve over time, which will allow early adopters to more easily adapt to what is quite foreseeably a paradigm shift in technology, and possibly even society.
DISCLAIMER: At the time of writing I am an employee of Cochlear LTD. All information used in this, and related, articles are based on publicly available sources. At no point was privileged or internal information used in any way. This was done as a part of a master's degree assignment.
*Generated by Midjourney v4 prompt "Corgi using chat gpt"
References
3 Areas Where AI Will Boost Your Competitive Advantage. (2021). Harvard Business Review. https://hbr.org/2021/12/3-areas-where-ai-will-boost-your-competitive-advantage Bush, Vannevar. (1945, July). As We May Think. The Atlantic Monthly.
Ciampa, D. (1992). Total quality : a user’s guide for implementation https://archive.org/details/totalqualityuser00ciam
Cochlear LTD. (2022). Cochlear Annual Report 2022. Corporation Listings. https://www.listcorp.com/asx/coh/cochlear-limited/news/2022-annual-report-2749627.html
Çetindamar, Phaal, R., & Probert, D. (2016). Technology management activities and tools (2nd ed.). Macmillan Education.
Chen, C.-J., Huang, Y.-F. and Lin, B.-W. (2012) ‘How Firms Innovate through R&D Internationalization? An S-curve Hypothesis’, Research Policy, 41(9), 1544–1554.
"NASA Program Management and Procurement Procedures and Practices” (1981)
Patterson, D. (2022, February 15). Good News About the Carbon Footprint of Machine Learning Training. https://ai.googleblog.com/2022/02/good-news-about-carbon-footprint-of.html
Smith, C., & McGuire, B. (2006, December). The History of Artificial Intelligence. Washington University. https://courses.cs.washington.edu/courses/csep590/06au/projects/history-ai.pdf
Tamkin, A. (2021, February 4). Understanding the Capabilities, Limitations, and Societal Impact of Large Language Models. arXiv.org. https://arxiv.org/abs/2102.02503
Xu, P., Zhu, X., & Clifton, D. A. (2022). Multimodal Learning with Transformers: A Survey. ArXiv (Cornell University). https://doi.org/10.48550/arxiv.2206.06488
Comments