The Effectiveness of Life Insurance Sales Force Training: Welcome “Me and AI”

Abstract

After 35 years of a free market in Poland, three life insurance companies have gained a dominant position in the market and developed certain procedural equilibrium in the area of training, allowing their status quo to be maintained. Yet, they do not take into account the opinions of agents and the possibility of using the latest IT developments, including artificial intelligence, which supports increasingly broad areas of activity in organisations with great success. As independent sales force training poses a challenge to any national or multinational company in a constantly changing global economy, the primary focus of this research was to analyse the opinions of the top 438 agents from dominant life insurance companies. A need was emphasised to reconfigure the existing training programmes with the potential for AI involvement to achieve a more effective educational trajectory. The research findings confirmed the necessity to reconstruct training programmes in relation to an agent’s age, education level, and seniority and offered grounds for discussing innovative AI concepts that can be relevant for future academic research in management sciences and improving organisational effectiveness, particularly in life insurance companies or other first-contact personnel-dependent institutions.

1. Introduction

1.1. The Impact of Life Insurance Industry on Global Economy

All economic processes, and thus institutions involved in them, are characterised by a dual social and economic nature. In all countries where economic relations are based on the foundations of a market economy, insurance companies represent a significant component of the supply side of the social security system, with individual life insurance serving as a fundamental pillar. The extant pension system is inadequate in terms of state coverage (Bagchi, 2021), thereby rendering life insurance policies an indispensable social instrument capable of addressing numerous financial challenges confronting European societies. The significance of insurance companies’ role in enhancing national economic performance is reflected in the substantial volume of insurance premiums they collect. In 2023, these premiums reached a total value of PLN 78.9 billion (2.32% of GDP), comprising 29.0% from life and 71.0% from property insurance (cf. www.stat.gov.pl, accessesed on 12 December 2024).

In accordance with the prevailing legal regulations, insurance products must be distributed by intermediaries, whose role in this process is of paramount importance. Insurance intermediation encompasses the performance of remunerated factual or legal actions related to the conclusion of insurance contracts.

Insurance mediation is exclusively conducted by insurance agents or insurance brokers, and it constitutes an economic activity within the meaning of the law. An insurance agent, despite engaging in activities that are characteristic of an employee is, in accordance with the regulations, considered to be an entrepreneur. This is because they are bound by an agency agreement with the insurance company that they represent. They perform activities on the basis of an agency agreement that has been concluded with the insurance company and is entered into the register of insurance agents. The efficiency of life insurance undertakings is contingent upon an optimisation of management processes, wherein the system of managing and the sales volume by the personnel assume a pivotal role, given the distinctive nature of the employer–employee relationship.

The insurance company offers services encompassing the sale of insurance policies, both directly (through its own employees) and through the implementation of insurance intermediation. A subjective approach to insurance intermediation distinguishes entities that are classified as insurance intermediaries.

A broader view of insurance intermediation is provided by the so-called French concept, according to which insurance intermediaries are employees of an insurance company engaged in solicitation activities for the insurer, as well as insurance agents and insurance brokers (2021). The choice of location and distribution channels represents a decision of significant strategic importance within the context of a service organisation.

This particular importance stems from the specific characteristics inherent to the production process. In addition, the term ‘distribution’ encompasses both the place and environment where the service is carried out, as well as the manner in which it is provided. These factors collectively serve to shape the overall image through which the value and benefits associated with purchasing a particular service are assessed. In Poland, insurance companies are required to use the following distribution channels:

-

Direct sales;

-

Agency sales through insurance agents (including multi-agents);

-

Sales through insurance brokers (

Table 1. Distribution channels of life insurance companies [in %], 2014–2022.

Specification	2014	2015	2016	2017	2018	2019	2020	2021	2022
Direct sales	19.56	9.93	14.81	15.28	12.25	14.68	15.26	15.95	16.44
Agency sales	79.80	88.55	84.08	83.60	86.84	84.32	83.62	82.96	82.72
Insurance brokers sales	0.61	0.58	0.57	0.48	0.49	0.52	0.78	0.68	0.46
Other (a.o. banks, post offices, Internet)	0.03	0.94	0.54	0.64	0.32	0.48	0.34	0.41	0.38

Source: author’s own study based on www.knf.gov.pl (accessed on 9 April 2023).

).

An analysis of the share and type of life insurance distribution channels demonstrates that the share of insurance agents’ activity in the total distribution of insurance companies’ products is considerable and remains stable. However, according to Anagol et al. (2016) and Šindelář and Budinský (2018), this activity is also associated with considerable risk for life insurance companies. However, Rogoziński (2016) posits that it is the quality of work performed by frontline employees that determines whether service organisations gain or lose a competitive advantage. This suggests that the HRM department plays a pivotal role in the efficiency of insurance institutions. It is crucial for these organisations to motivate and retain individuals who possess the desired personal qualities and competencies, which can be attained through a well-designed training programme. This is particularly important in the context of life insurance companies, given the unique nature of their operations (Nowotarska-Romaniak, 2018).

Therefore, in the era of globalisation and dynamic technological advancement, organisations have prioritised human resources as a crucial determinant to achieve a competitive advantage and long-term sustainability (Stofkova & Sukalova, 2020). As a result, many institutions have refocused their attention on implementing new technologies to enhance their performance (Schleicher et al., 2018) and aligned their strategies to enhance overall performance based on the application of technical tools such as artificial intelligence (Hilman & Kaliappen, 2015; Mert et al., 2021).

1.2. AI in HRM—Theoretical Approach

AI has gained a reputation as a more strategic tool, linked with organisational values, goals, missions, and visions (Malo, 2011; Aust et al., 2018). Moreover, several papers have addressed AI in HRM from different perspectives, particularly qualitative differences that can be expected in the context of HRM competencies levels after the implementation of AI within an organisation (Sadok et al., 2022), and the opportunities associated with it in HRM (Budhwar et al., 2022).

Yet, even though the abovementioned papers have been issued for the last 10 years, the concept of AI implementation in organisational systems began much earlier. The first discussions of the potential impact of AI on HRM can be traced back to Putman and Van’s (1987). The authors claimed that the widespread adoption of AI would lead to a gradual replacement of human-to-machine human interaction with human–machine interaction in HRM work.

Then, 25 years later, the number of publications relating to various aspects of AI integration in HRM increased noticeably and showed that organisations have moved from experimenting with AI implementation to using AI regularly in such areas of practice as new employee recruiting, training, and development, and career development (Aragón et al., 2013; Wesche & Sonderegger, 2021; Khandelwal et al., 2024). During the recruitment process, generative AI provides new employees with information about the organisation and its employees by answering questions in chats.

AI will also be used to automate formal paperwork, i.e., payroll, legal, and healthcare forms (Yorks et al., 2022). Later, in terms of career development, organisations implement AI to analyse data from employee records, i.e., educational background and previous training, to match employees’ characteristics with job requirements. AI is also used to develop customised training curricula for individual employees, focused on upgrading their skills, based on the results of an analysis of one’s needs (Sowa et al., 2020), which is also emphasised by Graßmann and Schermuly (2020). They claim that there could be significant benefits to AI in coaching, including anonymity and confidentiality, flexibility in accessing coaching at the most convenient time, and significantly lower costs compared to traditional approach coaching. Additionally, AI could assist in creating potential solutions to problems identified, assessing the utility of such solutions and selecting the most feasible ones, as well as monitoring the implementation of recommendations. In consideration of the aforementioned factors, the theoretical domain has thus far been unable to provide a definitive delineation of the preeminence of artificial intelligence over human trainers.

1.3. AI vs. Human Trainers: “Which Way to Go”?

The primary advantage of AI technology is its capacity for hard data computation. The distinctive strength of AI lies in its ability to process large volumes of data and discern latent patterns embedded within both structured and unstructured data (Davenport & Ronanki, 2018; Luo et al., 2019; Puntoni et al., 2020). The foundational AI technologies comprise deep learning-based methodologies such as Natural Language Understanding, Automatic Speech Recognition, Text-to-Speech Synthesis, Voice-operated Switch, and Media Resource Control Protocol. Researchers recognise AI as most suitable for tasks requiring heavy processing of text, speech, image, and video data (Brynjolfsson & Mitchell, 2017; Sundblad, 2018). As AI technologies become increasingly sophisticated, they are able to perform many tasks conventionally carried out by humans, complement human tasks, and even outperform humans (McKinsey Global Institute, 2018). For instance, in the context of outbound sales calls, AI has been shown to comprehend customers’ queries and engage in natural language conversations with them more adeptly than inexperienced workers (Luo et al., 2019). Furthermore, in e-commerce settings, AI has been shown to be capable of effectively handling data-intensive tasks such as machine translation and product recommendations (Brynjolfsson & Mitchell, 2017; Sun et al., 2019).

In the context of on-the-job sales training, the role of the coach is to review agents’ past conversations with customers and subsequently provide feedback that enables them to learn sales skills and improve future performance (Weitz et al., 1986; Kinkead & Riquelme, 2022). This task is characterised by its high data intensity, as coaches must (1) listen to the speech data to identify mistakes in agents’ conversations with customers and (2) it is imperative that specific solutions are provided in order to rectify each of the identified mistakes. It is evident that AI coaches have the capacity to process extensive amounts of speech data with greater efficiency. Consequently, they are able to detect a more extensive range of errors in conversations in comparison to human managers. Furthermore, AI coaches are trained with a vast amount of past call data tagged as best and worst practices to persuade customers, so they can provide more solutions for each mistake identified in speech data processing. It is evident that AI coaches possess distinct advantages over human managers in the provision of feedback to sales agents.

In contrast, human managers’ distinctive strength in their soft interpersonal communication skills (e.g., interpersonal empathy, encouragement, adaptivity, and acknowledgment) are at the heart of the human advantage over machines and is where AI falls short (Brynjolfsson & Mitchell, 2017; Davenport & Ronanki, 2018; Deloitte, 2017; Deming & Silliman, 2024). The effective conveyance of feedback to agents is pivotal for them to learn from the coaching information to improve job performance (Kinkead & Riquelme, 2022; Hall et al., 2015).

The effectiveness of such communications is contingent on the extent to which coaches are able to adapt feedback to the learning capabilities of agents and offer interpersonal support in the form of empathy, acknowledgment, and encouragement (Tversky & Kahneman, 1974). The interpersonal skills of human coaches have been shown to reduce agents’ resistance to coaching feedback (as agents gain more trust in the coaches) and overcome their learning barriers (Atefi et al., 2018; Kinkead & Riquelme, 2022). In summary, human coaches’ distinctive interpersonal skills constitute a relative advantage over AI coaches in communicating feedback to agents.

1.4. AI in Sales Agents Training—Good Practice Examples

According to Luo et al. (2020), based on their research, conducted on an existing sample of 429 active sales agents, the advent of AI has engendered a paradigm shift in the realm of sales training, empowering companies to harness the capabilities of AI coaches to enhance the efficacy and efficiency of agent training. This development has been instrumental in facilitating the optimisation of sales skills and performance. The utilisation of AI coaches in training agents offers distinct advantages over human managers, who may be susceptible to physical fatigue and emotional fluctuations. The consistency and predictability of AI coaches in training tasks ensure a more reliable and accurate training outcome. Moreover, the utilisation of AI coaches has the potential to address a significant challenge in the industry, namely the scarcity of human managers available for the training of inexperienced frontline employees. The utilisation of AI coaches in the domain of sales training offers a number of advantages over human managers, including the following:

-

AI coaches are not subject to physical fatigue or emotional fluctuations, which can be a factor in human managers’ performance;

-

AI coaches are able to perform repetitive sales training tasks in a more consistent, predictable, and accurate manner;

-

AI coaches can address another significant challenge in the industry, namely the limited supply of human managers available to train inexperienced frontline employees;

-

AI coaches can scale up quickly to train thousands of agents simultaneously with minimal marginal costs, making them a cost-effective solution for businesses.

Yet, the authors of the study have highlighted that the impact of AI training on sales can be likened to the U-inverted curve and is contingent on the sales outcomes achieved by agents prior to the commencement of the training process.

Similar findings were reported by Alghizzawi et al. (2025) on a sample of 178 pharmaceutical representatives. Authors claim the implementation of artificial intelligence (AI) applications has the potential to substitute for arduous processes, thereby liberating sales personnel to focus on high-value activities. Predictive analysis has the potential to enhance consumer segmentation, optimise targeting, and serve as a decision-making support tool. The utilisation of AI in personalising customer experiences is also recommended. AI systems analyse data and present specific suggestions according to user behaviour and interests. As a consequence, the personalisation of services and communications is conducive to enhancing the client experience. Yet, according to the authors’ opinion, the following are imperative:

-

Ensure that strategies underpinned by AI adhere to established ethical standards;

-

Maintain client data security while upholding confidence;

-

Emphasise that the primary function of AI should be to complement human activities, rather than to fully substitute for them;

-

Equip sales personnel with the necessary knowledge and skills to utilise AI effectively; as a result, businesses can leverage the potential of AI-driven strategies in conjunction with training;

-

Ascertain which organisational evangelists can encourage the use of AI.

Respectively, businesses that effectively integrate AI with learning will gain a competitive advantage. In an industry characterised by direct contact with clients, the ability to adapt swiftly to these changes is paramount.

1.5. AI in Insurance Sector—State-of-the-Art

AI is already used in selected areas in the insurance sector (Eling et al., 2021), i.e., fraud detection (Wang & Xu, 2017), virtual chatbots (Weizenbaum, 1966; Riikkinen et al., 2018), or calculating insurance risk (Vassiljeva et al., 2017). These applications of AI determine insurance business processes. As these trajectories of AI innovation start merging, the impact will be greater and, possibly, transformative. AI, in the largo sense, refers to a technology that has an “ability” to learn, whether or not it is supervised (Ghahramani, 2015; LeCun et al., 2015), making AI more accessible to organisations. There is a spectrum of capabilities of AI as well as an abundance of applications. As being more capable, AI is implemented across the value chain, more data are required, and more processes are changed. Some organisations cannot make this alignment as they do not possess the necessary data or the will to change their procedures. This suggests that some organisations can be propelled to a new business model by AI while others need to implement a new one to utilise AI.

Along with the insurers and AI, a third factor in the diffusion of AI in this sector is the consumer. Purchasing insurance and making claims are part of a relationship between the consumer and the insurer. Most contemporary applications of AI are used to increase the capabilities and knowledge of the insurance company (and not the customer).

As the information asymmetry becomes deeper, the insurer can provide lower-quality services and, as a consequence, lower prices. In the context of the provision of services, it is important to note that the cost implications of the services provided may have a significant impact on their availability. Indeed, it is conceivable that more expensive and higher-quality services could be withdrawn, and they will be unable to compete on price. It is evident that the insurer has the capacity to withhold further information pertaining to the decision-making process, thereby increasing moral hazard. The abovementioned catalogue of examples of AI implementation is described in

Table 2. The 20 cases of AI in insurance.

No	Case	Position in Insurance	Examples of AI Applications
1	Lloyd’s of London	Large incumbent, B2B, multichannel	Data entry quality control: AI analyses contracts to ensure regulatory compliance.
2	Willis Towers Watson	Large incumbent, B2B, multichannel	When processing a claim or evaluating a business customer, AI can identify the right expert and bring in their expertise at the right point.
3	AVIVA	Large incumbent, B2B, multichannel	A virtual assistant can be asked questions with natural language and reply.
4	Compare themarket.com	Incumbent, tech-focused, B2C, company offering a platform	Digital voice: full or semi-automatic interaction with the consumer using natural language processing and big data is supported by AI.
5	Bupa	Incumbent, health insurance, and healthcare provider	AI is used for illness and disability claim prediction.
6	Confused.com	Online, tech-focused, B2C, low cost, focused on car insurance	AI supporter virtual assistants (chatbots) support the customer with sentiment analysis and automation of time-consuming and repetitive tasks.
7	Cuvva	Small start-up, online, B2C, offers innovative services	AI supports underwriting, claims processing, and fraud detection.
8	Axa	Large incumbent, B2B, multichannel	AI supports underwriting, customisation, claims processing, and fraud detection.
9	Zurich Insurance	Large incumbent, B2B, multichannel	Fraud detection and reduction. AI also supports underwriting and claims processing.
10	Tokio Marine	Large incumbent, B2B, multichannel	Claim documentation, AI ‘reads’ hand-written documents.
11	Manulife	Large incumbent, B2B, multichannel	AI can underwrite insurance independently.
12	Allstate	Large incumbent, B2B, multichannel	AI supports underwriting, customisation with a virtual assistant, claims processing, and fraud detection.
13	Geico	Large incumbent, technology focused, B2C, specialised in car insurance	Recruitment, AI carries out interviews and matches the applicant’s skills to the right roles.
14	Progressive	Large incumbent, technology focused	AI is used to improve efficiency across a complete process so that humans are not needed.
15	Lemonade Insurance	Online, no physical stores, relatively new, technology focused, B2C, offers innovative services	Their virtual assistant is called ‘AI Jim’. ‘AI Jim’ can interact with the consumer when selling a policy, switching the claims process from another insurer. AI supports behavioural analysis.
16	Fri:day	Small start-up, online, technology focused, B2C, offers innovative services	The Guidewire system used is achieved by AI that is trained specifically to identify patterns that indicate fraud in insurance.
17	Ping An Insurance	Large incumbent, B2C, multichannel, offers innovative services	Innovative services utilise AI to be proactive and shape behaviour, like improving health.
18	Alibaba	Tech and e-commerce giant, with large user base entering the insurance sector with innovative services	Alibaba uses AI heavily in healthcare and this is being integrated with new insurance products.
19	Tencent	Tech and e-commerce giant, large user base entering insurance with innovative services	AI is used across this technology giant’s e-commerce and social media for analysis, facial recognition, natural language processing, fraud detection, and security.
20	Tesla	Tech company influencing insurance sector	There are new opportunities and challenges for insurance created by AI in self-driving cars.

Source: Zarifis et al. (2019).

.

The analysis of the data presented in the table above suggests that no insurance companies (including life insurance) have attempted to utilise artificial intelligence in the training process of insurance personnel. This finding indicates a clear organisational and, consequently, research gap. As the jobs of a pharmaceutical representative and a life insurance agent are similar in nature, in the context of a dynamic market, life insurance organisations that embrace the abovementioned findings can navigate change while achieving long-term development. The rationale underpinning this declaration has been empirically validated by the findings of scientific research conducted in the aforementioned sectors of the global economy. Hence, it appears appropriate to propose an approach aimed at addressing this gap and providing recommendations for future research and organisational effectiveness.

2. Methodology

2.1. Trainings of Sales Force as a Crucial Element of Institutional Effectiveness—Research Questions

It is a common practice among companies, including those in the insurance sector, to invest considerable time and financial resources in training their sales force with the aim of increasing sales revenue, company productivity, and profitability (A. Attia et al., 2014), job satisfaction, and organisational commitment (Arnette & Pettijohn, 2012). The 2024 Training Industry Report (Freifeld, 2024) estimated investment in training to be USD 98 billion. Approximately 7% of all training investment is allocated to real estate/insurance with the objective of updating the evolving roles of sales professionals. When the potential loss of sales revenue is taken into account, the total cost of sales training is even higher (A. Attia et al., 2014). In one of the most significant studies in this field, A. M. Attia et al. (2021) identified several key dynamics in sales training. These include systematic presentation and communication of ‘good selling practices’ to sales professionals. The primary objective of sales training is to enhance sales performance. The performance of a salesperson can be defined as the actions they undertake that contribute to the achievement of organisational goals. An individual’s skill level is a significant predictor of sales performance, and it is closely related to their ability to perform sales tasks (Koponen et al., 2019). The literature indicates that training improves salespeople’s knowledge base and skill levels, which, in turn, leads to enhanced performance. For example, Itani et al. (2019) found that the most critical factors contributing to a salesperson’s failure can be addressed through training. Taking into consideration the aforementioned circumstances, the first research question was established, as follows:

RQ1: What is the current state of training and the direction of its development in the leading life insurance companies in Poland, in the agents’ opinion: do they depend on education, gender, age, and length of cooperation with the life insurance company?

In order to answer this question precisely, the following specific questions were introduced and asked to insurance agents—trainees at the dominant life insurance companies in Poland (see Appendix B. RQ1).

From the aforementioned perspectives, training facilitates enhanced learning and enables salespeople to attain more acceptable performance levels in a shorter timeframe than it would have been possible through direct experience alone (Brixiová et al., 2020). These points, which are partially consistent with this effect, were reported by Bharadwaj and Shipley (2020) in an exploratory study based on data collected from salespeople from three companies. To illustrate, following a series of outcomes, one sales manager identified sales training as a pivotal element in enhancing the performance of the sales force, and the participation of the parties involved in the process described was considered crucial (Mahmood et al., 2024). This approach formed the foundation for establishing the second research question, as follows:

RQ2: What are the agents’ needs/expectations for training: are they dependent on education, gender, age, and length of cooperation with the life insurance company?

To answer the aforementioned research question (RQ2), the following specific questions for agents were formulated (see Appendix C. RQ2).

In this context, since each insurance company is usually a separate organisation with its own strategy and procedures, it was decided to increase the degree of precision and statistically verify the existence of differences between the various institutions participating in the study. The correctness of this approach is justified by the author’s previous research (Janowski, 2007) and confirmed by the conclusions of Adeoye and Adong (2023) and Limanowski et al. (2024). As a result of the above, the third main research question was established, as follows:

RQ3: Are there differences in the answers to particular questions between agents working for the leading insurance companies surveyed?

In order to answer these questions, dedicated statistical tools were introduced, in line with the established methodology used in management sciences, to verify the answers obtained from insurance agents.

2.2. Sample and Data Collection

As mentioned above, the main and crucial distribution channel for life insurance companies in Poland is agency sales performed by self-employed entrepreneurs. I decided to conduct research in this professional group in dominant (over 85% of the Polish market share) life insurance companies in Poland (i.e., Allianz, Nationale Nederlanden, and PZU). The inductive method was applied in the research, as this tool is particularly useful and adequate when the conceptual base cannot determine identifiable dimensions in a simple manner (Williamson et al., 1982). It requires an expert approach to an analysis of the sample content, is based on a post hoc factor analysis (Anderson & Gerbing, 1991; Kerlinger, 1986), and also asserts a correct categorisation of factors (Ford et al., 1986). Additionally, a comparative analysis of the subject literature increased the level of research result validation (Eisenhardt & Graebner, 2007). The case study was constructed through the use of an iteration process, which was based on a consonance of theoretical assumptions and empirical evidence (Dubois & Araujo, 2007; Dubois & Gadde, 2002). The introduction of a case study in the context of theory development enhances inductive research through the creation of an adequate theory, which facilitates scientific progress and is verifiable (Gibbert & Ruigrok, 2010). The purpose of the study was to analyse the current state of trainings in dominant (based on their market share) life insurance companies in Poland (indicating the existing differences in approaches in each) and to compare them with the cooperating agents’ needs/expectations to recommend the further development trajectory in order to improve the effectiveness of training in the future and the efficiency of the insurance organisation determined by it.

The research was conducted from 1 March 2024 to 30 August 2024. The questionnaires were carried out by the author directly with the participants, which made it possible to maintain the highest standards of reliability of the results. The content of the agent’s questionnaire was established as a result of using the Delphi method of combined efforts by academics from the European University of Medical and Social Sciences and the chief training managers from the three most effective life insurance companies in Poland. This part of the research focused on increasing the research sample’s representativeness level. We examined 12 individuals (4 from each company, who agreed to the interview first) as the initial research. The inaccuracies and errors identified were then corrected, and the main research proceeded. Finally, the research sample consisted of 438 top (according to the sales results from the preceding three years) agents (35.62% men and 64.38 women, among whom 19.18% had secondary education, 31.51% had secondary technical education, 37.67% had higher education, and 11.64% had higher technical education), obtained on a voluntary basis, with the sample size determined by the confidence interval for the population. The final version of the research questionnaire is attached in Appendix A.

2.3. Statistical Approach

The statistical analysis employed the χ² (maximum likelihood), which is considered to be the most reliable method, the contingency ratio (C Pearson’s), and the p level to test relations. The participants evaluated the contemporary training curriculum within three dominant life insurance companies, the competencies of the trainers, and the potential sales outcomes after completing the course. They also identified future directions for change to enhance sales success. In order to increase the precision of statistical calculations, it was decided to create rank intervals. With regard to the variables of age and work period, a contingency ratio of Pearson’s C was established in the following ranges: very weak (for C ≤ 0.2999), weak (for C: 0.3000–0.3999), moderate (for C: 0.4000–0.4999), strong (for C: 0.5000–0.5999), and very strong (for C ≥ 0.60).

All the calculations and their results were provided in Statistica 12.5, SPSS 30.0.0, and Microsoft Excel. The sample size was calculated based on the confidence interval of the sample for the fraction (1), as follows:

$$P\left(\widehat{p}-z_{\alpha /2} \sqrt{{\displaystyle \frac{\widehat{p} \left(1-\widehat{p}\right)}{n}}}\le \mathfrak{n}\le \widehat{p}+z_{\alpha /2}\sqrt{{\displaystyle \frac{\widehat{p} \left(1-\widehat{p}\right)}{n}}}\right)=1-\alpha$$

(1)

where

n—sample size; P_i—unknown population proportion; $\widehat{\mathrm{p}}$—population proportion in a sample study; 1 − α—confidence coefficient, probability that the interval will cover an unknown population proportion; ${\mathrm{z}}_{\mathsf{\alpha }/2}$—quantile of the N (0;1) distribution.

The selection of this model was made based on a significantly large general population exceeding 1 million. The draw was conducted without returning, even though the model allows it. An option to make an alternative choice could be a model featuring a limited population draw. However, with such a vast general population, both approaches result in almost identical trial population sizes, differing by less than one person.

Assuming that the admissible estimation error of the population proportion is not to exceed the set value of d (2):

$$z_{\alpha /2}\sqrt{{\displaystyle \frac{\widehat{p}\left(1-\widehat{p}\right)}{n}}}\le d$$

(2)

and assuming $\widehat{p}$ = 0.5, the minimum sample size was the following (3):

$$n\ge {\displaystyle \frac{{\left(z_{\alpha /2}\right)}^2}{4d^2}}$$

(3)

The confidence ratio in the study was 1 − α = 0.95 and d = 0.05. Therefore, n $\ge$ 385, and the obtained number of n = 438, fulfils the assumption of the maximum error in the fraction estimation, which is 4.6% with 95% probability. The research participants were agents from dominant life insurance companies in Poland (based on their market shares), including the following:

-

A total of 53.42% from Allianz;

-

A total of 16.44% from Nationale Nederlanden;

-

A total of 30.14% from PZU.

The agents’ opinions on the current state of training curricula and their needs were evaluated through a Likert scale ranging from 1 (highest) to 5 (lowest) for questions 1–11 and 13–15. In the context of others, two or three answer options were available, namely “yes”, “no”, or “I do not know.” The outcomes are presented in the Section 3.

3. Results

Relations between the agents’ age, gender, length of cooperation, and education, and their opinions in the context of the curricula of training programmes, including in the research questions, were the subject of statistical tests (

Table 3. Statistics for RQ1.

Answers			Age				Gender				Education				Seniority
Main	Aux.	No	C	p-Value	Str. *	Ver **.	C	p-Value	Str.	Ver.	C	p-Value	Str.	Ver.	C	p-Value	Str.	Ver.
RQ1	rq1.1	1	0.5412	0.00030	strong	+	0.1975	0.22032	none	-	0.4296	0.00057	moderate	+	0.5792	0.00002	strong	+
	rq1.2	2	0.6015	0.00141	very strong	+	0.4019	0.00178	moderate	+	0.4300	0.11574	none	-	0.6220	0.00027	very strong	+
	rq1.3	3	0.5821	0.00000	strong	+	0.1736	0.30053	none	-	0.3895	0.01407	weak	+	0.5464	0.00001	strong	+
	rq1.4	4	0.4887	0.00214	moderate	+	0.2724	0.05697	very weak	+	0.4566	0.00053	moderate	+	0.5062	0.00063	strong	+
	rq1.5	7	0.5975	0.00004	strong	+	0.1869	0.45528	none	-	0.4970	0.00056	moderate	+	0.5575	0.00023	strong	+
	rq1.6	8	0.5664	0.00003	strong	+	0.2217	0.30074	none	-	0.6643	0.00000	very strong	+	0.5599	0.00014	strong	+
	rq1.7	9	0.5273	0.00124	strong	+	0.2381	0.12671	none	-	0.6087	0.00000	very strong	+	0.4876	0.03874	moderate	+
	rq1.8	13	0.5080	0.00233	strong	+	0.4408	0.00017	moderate	+	0.5502	0.00007	strong	+	0.5252	0.00334	strong	+
	rq1.9	14	0.5675	0.03210	strong	+	0.4172	0.00152	moderate	+	0.5622	0.00027	strong	+	0.5556	0.01339	strong	+
	rq1.10	16	0.4912	0.00460	moderate	+	0.0920	0.74038	none	-	0.4980	0.00005	moderate	+	0.5128	0.00047	strong	+
	rq1.11	18	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a

* Strength of relation, ** verification. Source: author’s own research.

).

An analysis of the data in the table above, describing the agents’ responses in the context of the state of training in life insurance companies, shows that the answers given by the agents are significantly highly determined by age; a particularly strong relationship is observed in the case of rq1.2 and a strong relationship in rq1.1, rq1.3, and rq1.5 to rq1.9; these together account for 72.72% of all the responses that could be statistically verified. In the case of the length of seniority of the agent’s cooperation with the insurance company, the correlations are even stronger, with as many as 81.81% of the responses having a relationship with it (compared to age, a positive change occurred in rq1.10). Education is another element that determines the answers provided by the life insurance agents, with a particular emphasis on the answers to questions rq1.6–rq1.7. However, the percentage of very strong and strong dependencies decreased to 36.36%. The factor with the lowest impact on the agents’ responses was their gender—only in 36.36% of the cases is there a statistical dependence and, at most, to a moderate degree. In the case of rq1.11, statistical tests ruled out the existence of a relationship. The above allows, subject to a selective approach, a holistic answer to RQ1. The agents’ responses to questions in the RQ2 area were similar in nature compared to RQ1 (

Table 4. Statistics for RQ2.

Answers			Age				Gender				Education				Seniority
Main	Aux.	No	C	p-Value	Str. *	Ver **	C	p-Value	Str.	Ver.	C	p-Value	Str.	Ver.	C	p-Value	Str.	Ver.
RQ2	rq2.1	5	0.4212	0.30521	None	-	0.4653	0.00008	moderate	+	0.5552	0.00039	strong	+	0.5429	0.00233	strong	+
	rq2.2	6	0.5259	0.00705	strong	+	0.4056	0.00241	moderate	+	0.5061	0.00646	strong	+	0.5575	0.00019	strong	+
	rq2.3	10	0.6807	0.00000	very strong	+	0.3145	0.02090	weak	+	0.4638	0.00082	moderate	+	0.6635	0.00000	very strong	+
	rq2.4	11	0.6401	0.00000	very strong	+	0.3543	0.00677	weak	+	0.4320	0.01945	moderate	+	0.6216	0.00000	very strong	+
	rq2.5	12	0.3975	0.02023	weak	+	0.2439	0.03804	very weak	+	0.3174	0.12160	none	-	0.3713	0.06705	none	-
	rq2.6	15	0.5988	0.00007	strong	+	0.2274	0.40193	none	-	0.5381	0.00018	strong	+	0.6312	0.00002	very strong	+
	rq2.7	17	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a
	rq2.8	19	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a	n/a
	rq2.9	20	0.6846	0.00000	very strong	+	0.1655	0.31503	none	-	0.2879	0.25109	none	-	0.5110	0.01140	strong	+

* Strength of relation, ** verification. Source: own research.

).

The responses given by the agents, included in Table 3, indicate the existence, in principle, of a linear similarity with those contained in the RQ1 area, with strong and very strong correlations decreasing slightly to 55.6%, 66.7%, and 33.3%, respectively; however, it should be noted that the share of very strong correlations increased, compared to the RQ1 area, by three times. In the case of gender, the percentage share increased to 55.6%; however, the strength of the dependence decreased. In this area, statistical tests did not confirm the existence of any dependencies in rq2.7 and rq2.8, which, however, does not invalidate the comprehensive response to RQ2 (including the results cited above in this section). The information in the referenced tables is reflected in the differences between the responses given by the agents from the life insurance companies surveyed with the largest shares in the Polish insurance market (

Table 5. The statistics for RQ3.

Answers and Statistics
Main	Aux.	No	C	p-Value	Strength	Verification
RQ3	rq1.1	1	0.2218	0.38920	none	-
	rq1.2	2	0.3420	0.14458	none	-
	rq1.3	3	0.1986	0.55126	none	-
	rq1.4	4	0.3506	0.03033	weak	+
	rq1.5	7	0.4339	0.00540	moderate	+
	rq1.6	8	0.4041	0.00980	moderate	+
	rq1.7	9	0.4636	0.00013	moderate	+
	rq1.8	13	0.1843	0.86441	none	-
	rq1.9	14	0.2153	0.78958	none	-
	rq1.10	16	0.2746	0.08369	none	-
	rq1.11	18	n/a	n/a	n/a	n/a
	rq2.1	5	0.5541	0.00001	strong	+
	rq2.2	6	0.4152	0.00760	moderate	+
	rq2.3	10	0.3498	0.00880	weak	+
	rq2.4	11	0.3432	0.07817	none	-
	rq2.5	12	0.2727	0.08605	none	-
	rq2.6	15	0.5188	0.00022	strong	+
	rq2.7	17	n/a	n/a	n/a	n/a
	rq2.8	19	n/a	n/a	n/a	n/a
	rq2.9	20	0.2400	0.14760	none	-

Source: author’s own study.

).

The information in Table 5 indicates the existence of differences between insurance companies; however, strong differences are found only in rq2.1 and rq2.6. In 12 questions (60%), there are no statistically significant differences between the answers from agents from different insurance companies; in 4 (20%), the difference is moderate; and, in 2 (10%), the difference is weak, so this confirms, along with the abovementioned determinants, that the answer to RQ3 is positive (according to the ceteris paribus rule); however, the author would like to stipulate at this point that further research is needed in this area.

4. Discussion

The relationship between the agent and the insurance company is situated within an interdisciplinary field that encompasses entrepreneurial theory, human resources, systems, and behavioural economics. Furthermore, the agent is not an employee but rather a cooperating entrepreneur (B2B). It is also noteworthy that the insurance company has a limited ability to influence the quality of the agent’s work in non-product areas, while the agent’s work is a significant factor in nearly 83% of the cases (Table 1). It cannot be argued that the training process for sales personnel requires the utmost care, particularly in terms of technique and content (A. M. Attia et al., 2021). This is particularly true for those agents (Bowen et al., 2023; Alavi et al., 2024) who represent public trust organisations. The implications of these conditions are reflected in the findings of the research. It seems reasonable to suggest that, after more than three decades of operational activity, insurance companies have reached a point of market maturity and need a (technological) breakthrough to lift this equilibrium to a higher level. As the dominant life insurance companies in Poland do not differ fundamentally in the context of training strategies for their sales personnel (Table 5), it can be reasonably asserted that the abovementioned state is characteristic of each of the insurance companies surveyed. Furthermore, no life insurance company has implemented a training programme that differentiates based on an agent’s gender, education, age, or seniority; however, based on the research results, in training curricula, such distinctions may have a strategic role in effective sales. However, it requires the implementation of the latest technologies with high computing power.

Although such power is currently available, as can be seen from the contents of Table 2, insurance companies use AI in areas that support core operational activities (back office), although theoretical works already exist to confirm the effectiveness of AI implementation in the human resource management process. For example, according to He and Burger-Helmchen (2024), AI can significantly support knowledge management strategies, also by automating the entire process. Moreover, the aforementioned authors further argue that with advances in machine learning, an ability to learn and contribute to an organisation’s knowledge base will no longer be an exclusive domain of humans, although machine learning differs significantly from human learning. The former is based on data processing (Martins & Galegale, 2023), while the latter is a social activity (March & Simon, 2023). In this context, organisations should therefore synchronise human and machine learning to achieve the highest level of training. The potential of artificial intelligence in the context of reducing the time required by humans to create knowledge will help to speed up the process. As Table 2 shows, artificial intelligence can effectively replace humans in some aspects of explicit knowledge transformation and act as an effective knowledge repository. However, AI is not currently able to do so in the area of tacit knowledge processing or in making decisions and taking actions based on a human’s own experience, although it brings efficiency and precision to these processes, enabling individuals, groups, and organisations to share knowledge effectively with others, particularly when applied on a global scale (Lesjak & Natek, 2022).

However, every new phenomenon in the world of science is accompanied by a significant number of publications questioning the positive qualities of technological achievements. In the case of AI, one can find opinions highlighting the ethical concerns that accompany the use of AI in human resource management departments in organisations, where computer program algorithms determine the well-being and development of human individuals (Tanaka & Tambe, 2022). An example of such ethical concerns is provided by Graßmann and Schermuly (2020) in their discussion of the use of artificial intelligence in coaching. The authors argue that while AI can select developmental goals for the trainee, it cannot be trusted to make ethical decisions and select goals that are beneficial to the client while not being harmful to other individuals or the environment. In addition to ethical ambiguities, researchers highlighted the potentially harmful effects of AI on human learning and development. For example, Ardichvili (2022) discusses how AI-based automation can lead to deskilling and a loss of skill development for expertise. Examples from accounting also show that cognitive-based automation provides employees with fewer opportunities to learn by performing increasingly complex tasks (Arnold et al., 2013). Furthermore, the automation of cognitive work leads to a breakdown of the task into smaller and easier-to-implement parts (Jarrahi et al., 2023). This leads to a loss of control over the overall work process, as dividing and delegating a large part of the work to machines makes it difficult to view the work process. This, in turn, leads to disenfranchisement and a reduced potential to find meaning in and through work. Furthermore, Anthony (2021) demonstrated that knowledge workers who utilise artificial intelligence tend to place excessive trust in analytical technologies, often lacking a comprehensive understanding of their operational mechanisms. Poquet and De Laat (2021) posit that artificial intelligence has the potential to constrain human cognition and social interaction. They further posit that humans possess a limited comprehension of the manner in which artificial intelligence will transform formal, informal, and incidental learning processes. Furthermore, there is a significant risk of cognitive overload as a result of an accelerated delivery of new content (Poquet & De Laat, 2021). The accelerated implementation of artificial intelligence has heightened concerns regarding its potential adverse impacts on human cognition. Integration of chatbots into various contexts, beyond mere information provision, poses a significant risk of overconfidence in responses to complex enquiries, thereby diminishing opportunities for reflection, learning, and further cognitive development. In summary, the extant literature on the use of AI in human resource departments suggests that widespread implementation of AI can bring both significant benefits and challenges (Jaiswal et al., 2021; Singh et al., 2024; Złotowski et al., 2017; Mirbabaie et al., 2021). The cited authors present a variety of scenarios and forecasts regarding the impact of artificial intelligence on the aforementioned aspects. However, the studies do not provide any definitive conclusions or recommendations, with the exception of one, which suggests the necessity of integrating human resources and machine “thinking” to enhance the capabilities of the former. This hypothesis is also supported by the findings of the present study. The implementation of the aforementioned recommendation appears to be a logical and economically viable strategy for increasing agent sales. The life insurance industry is currently experiencing a phase that Liberman et al. (2001) foreshadowed, namely, a state of stagnation that has undoubtedly affected life insurance companies in Poland. It appears that a solution to this problem lies in the implementation of artificial intelligence into a training process (Jagdip et al., 2019; Guenzi & Habel, 2020; Raisch & Krakowski, 2020). Until recently, due to the level of technological development (the computing power of computers, very limited machine learning), this solution could only remain in the realm of postulation. As Malla et al. (2021) observed, a reconfiguration of the strategy is now required in order to achieve a higher level of organisational development and sales effectiveness (cf. also Loureiro et al., 2024). Additionally, an earlier study conducted by Luo et al. (2020) in a typical sales agency environment corroborates the veracity of the aforementioned thesis and the findings of this article. The authors similarly advocate for a selective and tailored approach to the agent population (cf. also Atefi et al., 2018). A combination of artificial intelligence and a human trainer in one team was demonstrated to exhibit the highest effectiveness. This hybrid approach demonstrated superior training effectiveness both for the lowest- and highest-ranked agents, outperforming each individual trainer. The results of this study, based on a representative sample of insurance agents, corroborate the hypothesis proposed by Luo et al. (2020). Additionally, the results justify an introduction of a lean management trajectory into the specific context of individual life insurance sales, particularly in relation to the agent’s education, seniority, and age, which is only possible with artificial intelligence involvement. This should, however, be correlated with the training specificities of each insurance company separately. The implications of these findings extend beyond the academic realm, offering valuable insights for organisational practice.

5. Conclusions

The training of salespeople necessitates a considerable investment of resources on the part of sales organisations. This is required in order to enable salespeople to deliver personalised presentations based on their perception of the uniqueness of each sales situation and to implement marketing concepts at the customer level. The current research at the most prominent insurance companies indicates that to enhance the return on these investments, salespeople could potentially benefit from a reconstruction of training programmes in the non-product area. It is therefore recommended that sales training programmes prioritise the delivery of the agents’ desired training content at optimal times, taking into account their age and seniority as primary considerations. Given the considerable number of life insurance agents, it is evident that the aforementioned recommendation cannot be fully realised through solely human endeavours. The current limitations in the computational capacity of human capital, coupled with the initial challenges in estimating the time and resources required for “just in time” training implementation and evaluation, make this approach impractical. It would thus be advisable to make use of the latest developments in the field of artificial intelligence, at the very least, in order to manage training logistics. Nevertheless, as relationship-based sales of individual life insurance policies remain effectively resistant to the online channel, the training process cannot be deprived of the participation of trainers. This is because, as Pelau et al. (2021) and Smith et al. (2024) demonstrated, artificial intelligence is currently unable to generate either empathy or other emotions in a manner equal to human transmission. Given the experience of the corporations listed in Table 2, it is justifiable to claim that AI will be very successful in managing the non-product training base of life insurers on a ’just in time’ basis, especially as there is access to agents’ appointment calendars and the products they sell. AI may also suggest the agent attend non-product training if there are problems in meeting the sales requirements. However, because selling is based on emotions, the involvement of a human coach will be essential in this context, although AI can indicate and suggest the trainer who has the best achievements in the area of the agent’s shortcomings (according to other agents’ evaluation). Therefore, in light of the findings presented in this study, it appears that adopting the “Me and AI” approach, as previously mentioned, could prove to be a cost-effective strategy for optimising the efficiency of the agent’s training process. The high effectiveness of such a solution has already been confirmed in the studies by Butt et al. (2021), where AI was a human avatar in games, and Erengin et al. (2025) claimed that AI-generated personal and psychological added value for customers. Finally, Osborne and Bailey (2025) found that AI significantly aids human decision-making. Therefore, it seems justifiable to assume that the concept of “Me and AI” should contribute significantly to improving the efficiency of the agent’s training processes in life insurance companies, although the author cautions against excessive optimism in this area, as each approach has its limitations.

6. Limitations and Future Research Directions

As previously stated, an agency contract represents a collaboration between two companies, with all the inherent consequences. Insurance agents constitute a distinct professional group, and the standard training tools designed for employees are not applicable in this context, primarily due to the lack of opportunities to exercise more comprehensive control over the work of agents. Consequently, agents are primarily interested in training initiatives that are designed to enhance their sales capabilities, either directly or indirectly. Strictly back-office activities are of secondary importance to them. This is particularly salient given that AI is predominantly employed in the domain of after-sales service within life insurance companies. The author would also like to emphasise that, to the best of his knowledge (as of 8 March 2025), none of the life insurance companies in Europe have initiated real activities aimed at enriching the training process of agents with the use of artificial intelligence. Therefore, the conclusions and recommendations in this area must, for the time being, remain purely theoretical. In that context, although this research represents a pioneering study in Central Europe, it would be premature to generalise the findings to non-insurance contexts without further investigation in other sectors of the economy. Additionally, as agents were self-reporting, results may be impacted by social desirability bias or selection bias. Nevertheless, it is possible that the conclusions may prove useful for training managers in organisations offering other types of services to retail customers.

At the organisational level, the author posits that CEOs responsible for training departments in life insurance companies should establish a working team comprising top agents, sales managers, and IT specialists. They need to endeavour to construct software, with due consideration for the recommendations described in this study. The subsequent step would be to introduce the project to a selected group of agents and, in the case of positive results, implement the tool in the life insurance company’s management strategy and attempt to transfer good practices to other sales organisations.