How does a navy that strives for operational advantage to fight and win, demonstrate its ethical prowess and pivot around employee experience?
The best way for the Royal Navy (RN) to pivot around employee experience and demonstrate ethical prowess, is to leverage recent transformations in Artificial Intelligence / Machine Learning (AI/ML) and imminent RN investment in the PODS ecosystem and Augmented Reality (AR). The Maritime Operating Concept (MarOpC) already contains a framework to explain how they could deploy. MarOpC’s modular ‘Sense / Decide / Effect / Enable’ construct translates neatly into how AI/ML, PODs and AR could drastically improve ethical prowess and employee experience. AI/ML can Sense employee experiences and Decide which interventions to make; while a combination of AI/ML, AR and physical Effects, deployable via PODs, will collectively Enable a pivot around improved employee experience. (1)
Figure 1. The MarOpC Capability Framework is as relevant to improving employee experience as it is to weapons and sensors.
Many employers structure their workforce interventions around Maslow’s ‘Hierarchy of Needs’ theory, further refined by the psychologist Frederick Herzberg. Herzberg highlighted those workplace factors that cause job dissatisfaction through their absence (‘Maintenance Factors’, e.g. salary and working conditions) and those that cause job satisfaction through their presence (‘Motivation Factors’, e.g. recognition, opportunities for personal growth).
The RN should therefore deploy cutting-edge technology to Sense where Maintenance and Motivation Factors are lacking. Once detected, the same technology can improve them at minimal relative cost, promoting the Factors that sailors and marines care most about. Some Maintenance Factors are either not under RN control (e.g. salaries), or already have suitable reporting policies in place (e.g. physical safety reporting); application of technology will clearly be less effective here.
Sensing Employee Experience. AI/ML can capture and analyse the current state of employees more quickly, cheaply and comprehensively than current methods, providing the RN a better understanding of which Maintenance Factors to improve and which Motivation Factors to provide.
While the RN relies upon the Divisional System or confidential methods to surface Maintenance Factors, such as Psychological Safety issues, this feedback is time-late or siloed, reducing both its effectiveness and the motivation for employees to engage. Even when the RN routinely collects feedback, it has limited ability to impact employee experience. After large-scale exercises, only the senior 2 or 3 leaders are interviewed; large numbers of RN/RM personnel are only proactively interviewed during the Armed Forces Continuous Attitude Survey. Due to cost and manual processing, AFCAS is only annual, requires 9 months to analyse, and involves just a portion of the RN/RM. This results in the loss of meaningful data. For example, by inviting all to take part, even the failure to complete AFCAS would become a useful statistic, which could reduce expensive voluntary outflow. (2)
However, a new form of Machine Learning is increasing the speed and breadth of feedback, while reducing its siloed nature and cost. ‘Transformers’ have enabled the creation of Large Language Models (LLMs), a form of Generative AI. Trained on a huge corpus of text, and deploying neural network architecture, LLMs demonstrate a behaviour called ‘emergence’, whereby they can undertake many tasks, despite not having been trained to do so. GPT-4, the largest LLM, matches or beats humans at a wide range of analytical tasks and improves daily.
Widespread deployment of an LLM-powered feedback tool means employee issues, including cultural and ethical concerns, would be elevated faster, allowing corrective actions to also be identified and taken quickly. Fast improvements will incentivise individuals to proactively offer feedback more often, resulting in a virtuous spiral. All RN personnel could anonymously update their feedback in an unstructured form (e.g. through the MyNavy app, with feedback stored for automatic upload when back in range of WiFi). Those needing to interrogate the RN’s collective thoughts would query the LLM while defining a response format, e.g. ‘What was the overall experience of female ratings during CSG21? Provide the top 3 recommendations for improvement. Focus on Motivation Factors, respond with a JSP101 Point Brief.’ LLMs will dramatically cut the time between workforce and ethical risks being identified and mitigated, especially at a time when HQ personnel numbers are reducing.
Decide the best way to improve employee wellbeing. LLMs can be tailored by use-case and trained on proprietary data; despite their newness, such LLMs are already being deployed across medical and commercial settings for HR analytics, resulting in the fastest uptake of a new technology in history. (3) Their ability to understand the needs of (and recommend interventions for) collective employees is resulting in faster, accurate recommendations about which workforce intervention to deploy at any given time and budget.
Affecting Employee Experience. LLMs can affect employee experience directly, not just Sense it. LLMs are already being deployed to reduce the bureaucracy associated with low-level tasks, such as reports or orders writing. This empowers individuals to spend less time on Maintenance tasks and more on Motivational ones, or simply achieve a better work-life balance. (4) The MyNavy App already demonstrates the effectiveness of this approach on low-level tasks. AI and AR, potentially deployed via PODS could disproportionately improve this.
LLMs will soon be available as psychological or career counselors. For some use-cases, a human-in-the-loop is required should the LLM flag up concerns, albeit the human should achieve a higher workload and support more sailors and marines. The recent deployment of a social worker on QNLZ could have been more effective had she been teamed with an LLM that allowed her to understand needs, explore interventions and write reports faster. (5)
LLMs can also improve on taskbook training by acting as an SQEP instructor and tailoring teaching styles by student, or oversee exams. A basic example is at Figure 2; an untailored GPT-4 LLM is teaching about a gas turbine and will respond appropriately and instantly to answers. (6) Each sailor could have their own LLM tutor, tailored to their learning styles and professional needs. A Training LLM will empower individuals to indulge their personal learning at a time of their choosing, a key Motivational Factor.
Figure 2. GPT-4 LLM acting as an ME instructor, using only untailored, open-source information.
When combined with Augmented Reality, LLMs will supercharge the ability to learn, especially at sea. The RN is already purchasing AR systems for shoreside teaching. AR Classroom PODS modules will let shoreside instructors, or LLMs, train seagoers, and allow them to access PODS-based teaching materials, mitigating poor connectivity. At relatively low cost, this reduces the expensive skillfade, or demotivation that would otherwise occur. With onboard real-estate reserved for PODS, such classrooms will prove dividends for employee experience and overall OC, especially if systems were made available for personal growth pursuits, innovation projects, or even off-watch recreation.
PODS spaces will be reserved for warfighting or humanitarian capabilities. But when a unit is outside an operational area, or has spare PODS spaces, or can quickly swap out PODS while forward-deployed, then the options for improving employee experience are considerable. (7) A CTG’s RFA would host logistics, plus, via bespoke PODS, the Taskgroup’s main AR training complex, an additive manufacturing system and drone engineering workshops. While employee PODS will primarily deliver Motivation Factors, Maintenance Factors could also be improved. Improved gym facilities, or fresher food via hydroponic farms could meet Maintenance needs; provided, of course, that the RN’s improved employee feedback systems indicated that these were problems worth solving. (8)
Figure 3. Examples of AR Classroom and Hydroponic PODS. The top image is synthetic and was created by an LLM.
But the RN and industry have focused on using AI/ML and PODS to improve data, sensors and weapons, rather than employee experience. Improving employee experience with these technologies is already proven in the commercial sector and wider Government, has a high ROI and, unlike AI-enabled weapons and sensors, does not require significant legal, ethical or technical approvals. It can deploy now and deliver an improved experience and ethical monitoring. Fortuitously, LLMs have emerged just as the RN is deploying both AR and PODS, 2 forms of flexible hardware that can pivot as required. With LLMs Sensing employee needs faster, the RN has a once-in-a-generation opportunity to pivot around employee experience, becoming a peer to, or better than, the most ethical organisations.
Footnotes
‘Maritime Operating Concept: The Maritime Force Contribution to the Integrated Operating Concept’, Ministry of Defence (UK), 02/02/2023, 20220629-Maritime_Operating_Concept__OFFICIAL___Publication.pdf (publishing.service.gov.uk) (checked: 02/04/2023).
For example, Meta conducts 2 firmwide surveys per year. Employees who fail to fill out both surveys are 2.6x more likely to leave within 6 months and can be proactively contacted before they submit their notice. ‘Employee Surveys Are Still One of the Best Ways to Measure Engagement’, Judd et al., HBR, 14/03/2018, Employee Surveys Are Still One of the Best Ways to Measure Engagement (hbr.org), (checked: 16/04/2023).
GPT-4 is one of a number of LLMs that powers OpenAI’s ChatGPT service. ChatGPT reached 1 million users in 5 days. It took Meta 10 months, and Twitter 2 years, to achieve the same market penetration. Buchholz, J, statista, 24/01/2023, Chart: ChatGPT Sprints to One Million Users | Statista, (checked: 16/04/2023).
PwC is the largest organisational investment to date. PwC US, 26/04/2023, PwC US makes $1 billion investment to expand and scale AI capabilities, (checked: 29/04/2023).
The first survey assessing LLMs in clinical psychology settings was published last month. Greco, C. et al, Pattern Recognition Letters, 03/2023, Transformer-based language models for mental health issues: A survey - ScienceDirect, (checked: 29/04/2023).
Clearly an LLM tailored on OEM or RN material would provide more precise answers.
T31 has spaces for 6 PODS, T26 has space for 10, and CVS and RFAs considerably more.
MSC Cruises deployed hydroponics at sea in March 2023. Note that this could also provide a Motivation Factor for RN caterers, who could take responsibility for, and experiment with, their ingredients. AgriTech Tomorrow, 04/03/2023, Hydroponic Farming at Sea – MSC Cruise lines and Babylon Micro-Farms partnership create a world first. | AgriTechTomorrow, (checked: 16/04/2023).