The Adoption of Artificial Intelligence in Firms

5. Findings from interviews with firms adopting artificial intelligence

Abstract

This chapter reports the findings of interviews with senior staff responsible for artificial intelligence in firms in the two sectors addressed in the 2022-23 OECD/Boston Consulting Group/INSEAD survey. The interviews aimed to elicit qualitative information to better interpret the quantitative data gathered through the survey, particularly in terms of policy-relevant questions considered novel in the context of other international surveys.

Introduction

Artificial intelligence (AI) could transform industries across the globe, prompting a need for in-depth understanding of its adoption and impact. This chapter presents findings from a series of interviews conducted with senior staff responsible for AI in firms from manufacturing and information and communication technology (ICT) services. These interviews complement and provide qualitative context to the quantitative data gathered through the 2022‑23 OECD/BCG/INSEAD survey, offering insights into a range of policy-relevant questions.

The interview process involved 15 experts holding various high-level positions, including chief information officers, chief technology officers, heads of digital business, directors of digital transformation, a vice president of data science and data engineering, data and machine learning engineers, and heads of research and development (R&D), among others. Some interviewees were drawn from enterprises that participated in the survey, while others were selected from a pool of over 600 candidates. The selection ensured approximately equal representation among countries and the two surveyed sectors.

The interviews focused on three primary themes:

1. Data acquisition: This theme explored how companies acquire data, with particular emphasis on data from research institutes and the public sector.

2. Public services to support AI adoption: The interviews sought to understand how companies interact with public services to develop AI and the key challenges they face in this process.

3. External collaboration for AI development: This area investigated practices for better AI adoption by leveraging the broader ecosystem, including suppliers and academic and research institutions.

The remainder of this chapter is structured to provide a comprehensive account of the findings. The following three sections summarise the salient insights gathered for each of the above topic areas. The chapter presents possible policy implications derived from these insights throughout the sections, summarising them in the conclusion.

Data acquisition

The interviews revealed that while many enterprises acquire data from research institutes and the public sector and recognise the many initiatives taken by governments in this connection, most rely on private data sources. The most common type of data acquired from public sources is generic data, such as demographic information, public company records, labour statistics and weather data. More specific and commercially valuable data sets from public administrations are rare. Private data sources are the preferred choice for most firms, as they offer more specialised and proprietary data that can provide competitive advantage. In addition, there may be more opportunities for giving feedback on data quality to private providers of data than to public providers.

Interviewees indicated that procedural complexities in acquiring public data significantly impede data-driven decision making. These complexities, deeply rooted in administrative processes, often exist to ensure accountability and adherence to established protocols. While such attributes are essential for maintaining data integrity and security, they often come at the expense of efficiency. Multiple layers of approvals, reviews and checks can lead to prolonged waiting periods. One expert observed that permissions can take up to six months or longer. Such delays can render data obsolete when accessed.

Interviewees noted that data from public repositories is often too old for real-time applications. Many businesses must invest time and effort to validate the currency of public data. Many types of data can quickly become outdated due to rapid technological changes and evolving market dynamics, among other conditions. Policy makers need to ensure that data remains relevant and actionable in the face of constant change.

Beyond timeliness, many interviewees expressed concern about data quality. Using publicly sourced datasets is often problematic due to vague terminologies and other shortcomings. The absence of comprehensive documentation can leave users grappling with the data’s true meaning and context. For example, a business might come across a large comma-separated values (CSV) or Excel file from a public source and encounter columns filled with terminology that is not easily understandable, while lacking accompanying documentation for clarification. A common problem the interviewees reported is the data quality itself. For example, it is not uncommon to encounter discrepancies, conflicting information and missing data. Considerable work is often needed on data cleaning and preparation, even for data that require a fee. Some interviewees noted that data obtained from public sources is of lower quality than that obtained from private sources. Overall, policy makers need to ensure that any shortcomings in data quality described here are addressed.

There is also a need for comprehensive documentation and standardised application processing interfaces (APIs). Comprehensive documentation serves as a roadmap for developers, guiding them through the intricacies of the data and helping to integrate it into their systems. Interviewees observed that it is essential that when querying an API, the results align with expectations and that any anomalies or potential quality issues are clearly explained. Moreover, adhering to standard practices ensures compatibility with existing technologies used to develop applications, such as REST API (Representational State Transfer), which is a set of rules and conventions for building and interacting with web services. Such standards streamline the integration process and bolster security and reliability.

A centralised public sector data access platform could streamline the search and retrieval process. A centralised hub could aggregate data and facilitate seamless transitions between databases, enhancing users’ ability to access and link to specific studies or datasets. Interviewees underscored that the absence of a unified platform often complicates data discovery.

Further, interviewees considered that the legal frameworks governing cross-border data flows could be made more compatible. International data sharing can be an intricate process, particularly when navigating diverse data-sharing laws. Different countries have distinct data protection and privacy laws. For instance, the European Union’s General Data Protection Regulation (GDPR) is one of the most stringent frameworks globally. Regulations along the lines of GDPR can facilitate data sharing as they support standardisation and trust. However, other jurisdictions might have more lenient or different standards. This can pose challenges for companies that operate in multiple jurisdictions and need to comply with each region’s specific regulations. To utilise data from different countries, specific protocols must be established to define how the data can be used, whether it can be merged, and under what conditions it can be combined.

Finally, vendor certification is common across industries and could be adapted for data vendors. One interviewee noted that such certification would help to provide assurance and confidence in the data’s authenticity and reliability. Especially for small and medium-sized enterprises (SMEs), checklists of the most important criteria to consider in vendor search and selection would be helpful.

Public services to support AI adoption

Access to information or advice concerning the adoption of AI

The survey data show that 75% of responding enterprises in manufacturing and 69% in ICT utilise public services that offer information and guidance pertaining to AI adoption (see Chapter 3). Most interviewed experts affirm that the insights derived from public sector sources help to make informed decisions and shape business strategies. Companies acknowledge the challenges of staying well-informed about markets beyond their own expertise. Access to information such as economic data, regulatory updates and compliance guidance is considered valuable. Such information is crucial in planning, analytics, market sizing, go-to-market strategies and understanding market dynamics.

Interviewees drew attention to a lack of consolidated information on private-sector AI software or services. Companies frequently receive tailored use case solicitations from vendors. These are often presented in marketing language. One interviewee suggested that governments could help by providing such information but in more neutral ways. One measure could involve providing information on solution providers with a relevant industry-specific track record. A suggestion made by some interviewees was that governments could facilitate decision making for businesses seeking AI solutions – especially SMEs – by developing a preferred vendor list, particularly for companies facing regulatory compliance considerations. Such a list would offer a curated selection of vendors with proven expertise in specific industries. This approach – aimed at reducing firms’ search costs for valuable advice – has been adopted by Singapore’s national research and innovation programme to harness AI’s scientific and economic potential (AI Singapore).

However, other interviewees expressed reservations about this idea, especially as concerns the possible anti-competitive effect of public authorities indicating private-sector contracting preferences. They suggested that governments could adopt an alternative approach of providing guidelines or a framework to aid SMEs in navigating the vendor selection process. By advising them on, for instance, the top ten considerations to bear in mind when choosing an AI vendor, governments could assist firms without favouring specific vendors. This would help firms evaluate AI products or services while preserving their autonomy in making vendor choices.

Interviewees also pointed to challenges in accessing public sector information to facilitate AI development. They highlighted the frequent lack of clear pathways to specific public agencies. The absence of a one-stop interface and streamlined processes and the occasional fragmentation of channels to public services create challenges to identifying the right agency or programme to consult. Interviewees highlighted that policy makers could help by establishing a consolidated platform or resource hub that streamlines access to AI-related information, guidance and advice from public agencies. Especially for SMEs, this would alleviate the burden of navigating fragmented services and help ensure transparency. Guidelines outlining agency roles and expertise, along with mechanisms for companies to communicate their needs, would also enable more targeted and efficient exchanges. Some interviewees emphasised the importance of having a single point of contact to assist in using various public support initiatives. Where this had occurred, having a dedicated contact person had proved highly beneficial.

Some ICT services companies do not consider public agencies an important source of guidance on AI development. These companies often deeply understand AI and possess internal capabilities to effectively drive their AI initiatives. As a result, they tend to rely less on external sources for direction and guidance. They are generally confident in their ability to chart their own course and make informed decisions based on in-house expertise.

Publicly provided or supported training services

The survey data reveal that approximately 58% of enterprises in the sample use public sector training services to help adopt AI (see Chapter 3). Regardless of sector, all interviewees reported challenges in finding AI talent. For instance, in manufacturing, companies seeking to implement AI-driven automation or predictive maintenance often struggle to find skilled professionals with expertise in AI algorithms, machine learning and data analysis. Similarly, in the ICT services sector, companies specialising in AI software development, natural language processing or computer vision encounter difficulties finding qualified professionals with specialised AI knowledge. This scarcity of talent hinders innovation and the deployment of cutting-edge AI solutions.

Those interviewees who expressed a reluctance to use public sector training programmes emphasised the need for more specificity in the training offered. For instance, instead of generic AI training, they found greater value in programmes tailored to industry or business-specific needs. For example, manufacturers may prefer training in AI that focuses on optimising supply chain management, while healthcare companies may seek training on AI applications for medical diagnostics.

Additionally, the experts highlighted the importance of hands-on training oriented towards real-world projects. Programmes that incorporate practical exercises and projects help participants to better apply AI concepts in the workplace. Workshops in which participants use AI tools and datasets related to their industry can be highly effective in boosting readiness to adopt AI.

Various interviewees asserted that public sector providers should collaborate with industry to deliver targeted training. Inviting industry professionals to share their experiences and insights can help develop training materials that provide practical perspectives and reflect best practices that resonate with private companies. The interviewees considered that collaborations between the public and private sectors can contribute to training programmes’ overall efficacy and appeal.

While companies welcome public initiatives to develop human capital, the relevance of these varies depending on companies’ industry and size. Among the experts interviewed, manufacturers more frequently expressed the need for new qualification frameworks. In line with the survey findings, the interviewees reported that some companies face challenges in determining the specific AI skills they need. Often, AI is perceived as a broad, all-encompassing term, overlooking the existence of distinct subfields within it. This contributes to a problem where academic certifications may not sufficiently provide the comprehensive information that employers seek. In this rapidly evolving field, there is a growing need for new qualification frameworks that effectively communicate precise and relevant information regarding candidates’ capabilities and competencies to employers.

The interviewed experts agreed on the need to develop new AI curricula to meet the growing demand for skilled AI professionals. During the interviews, experts provided insights on the content of AI degree programmes and perceived gaps in curricula. Many held that AI degree programmes lack a sufficient focus on industry-specific applications and practical skills. For example, a manufacturing company might require AI professionals with expertise in optimising production processes through AI-driven automation, while a healthcare organisation may seek AI graduates who are well-versed in medical image analysis and diagnosis. The experts highlighted the importance of gaining hands-on experience as a part of degrees in AI. Companies often seek AI professionals who can immediately apply their knowledge to real-world scenarios. Hence, curricula that incorporate practical components, such as internships or industry placements, are highly valued by employers.

Experts also expressed concern about the limited pool of AI graduates possessing specialised skills in subdomains of AI. The rapid expansion of AI has resulted in a scarcity of professionals equipped with the latest expertise in AI systems and applications. To tackle this challenge, certain companies have launched talent development initiatives. These initiatives include sponsoring AI-focused research projects and providing scholarships to students pursuing AI degrees. For instance, a data analytics firm might offer scholarships to students pursuing a master’s degree in AI with a focus on natural language processing, aligning with the company’s core business. Training efforts should also extend beyond individuals in technical roles. They should also include educational and training opportunities for individuals from diverse backgrounds and fields of expertise outside AI.

Publicly provided or supported funding programmes

The survey showed that 42% of enterprises use services provided by the public sector to promote access to finance, including subsidies and credit guarantees (see Chapter 3). It became evident during the interviews that many companies display a high level of awareness regarding the types of available public funding to support their AI initiatives. They demonstrate a clear understanding of the various avenues for financial assistance, such as tax credits, public funding for R&D, subsidies from public investment banks, and financing options guaranteed by the state. Among these forms of support, tax credits were the most frequently utilised.

In the 2022-23 OECD/BCG/INSEAD survey, around 40% of enterprises reported that their utilisation of AI in the past 12 months was constrained due to a lack of external financing (see Chapter 3). Interviewees emphasised that the evaluation methods for securing public funding can be overly narrow. The main issue is that the assessment processes for obtaining public funding, particularly grants, often focus on individual projects rather than a broader range of projects. Assessing projects individually increases the risk of failure for lending programmes overall, especially given the inherent complexities and uncertainties in the emerging field of AI. This funding may also fail to capture the collective impact and transformative potential achievable through a portfolio of AI projects in a company.

Some experts also underscored the importance of streamlining application processes for public funding. Doing so would give reviewers more time to study a project’s merits and, moreover, support SMEs that may otherwise struggle with complex application procedures.

External collaboration for AI development

Engaging with universities and public research institutions to develop AI

The OECD/BCG/INSEAD survey showed that collaboration with universities and public research institutions is widespread. More than half of the responding enterprises collaborate with university faculty members, PhD candidates or postdoctoral students to advance AI development. Approximately 55% of manufacturers and 48% of ICT services providers collaborate with researchers in public research organisations. In addition, around one-third of enterprises form partnerships with undergraduate students to foster AI innovation and research (see Chapter 3).

Collaboration with universities and public research institutes improves access to scientific expertise. Academic research institutions house highly skilled researchers and domain experts. By collaborating with these institutions, firms tap into expertise and opportunities for cutting-edge research that may not be readily available within their own organisations. This access to specialised knowledge helps firms address complex AI challenges more effectively.

The interviews showed that collaborative partnerships can facilitate knowledge exchange and technology transfer between firms and academic institutions. Firms can share their industry insights, practical experience and real-world datasets, enriching academic research. In a complementary way, academic institutions can share their latest research findings, methodologies, and theoretical advances, helping firms to leverage cutting-edge research. These partnerships can also provide access to state-of-the-art research facilities, advanced computing infrastructure and dedicated R&D teams, enabling firms to undertake more ambitious and resource-intensive AI projects.

Collaborating with academic research institutions provides firms with opportunities for talent acquisition and development. By engaging with PhD students, researchers and faculty, firms can identify and recruit talent. Furthermore, these partnerships facilitate internships and joint training programmes that help nurture the next generation of AI professionals in academia and industry.

The interviews suggest that firms often struggle with a lack of clear agreements and frameworks for intellectual property (IP) management and ownership. Collaborations between research institutions and firms, especially those in the ICT services sector, frequently yield intellectual property and the associated IP rights. The interviewees noted that striking a balance between the interests of both parties regarding the ownership, usage and commercialisation of IP can be complex and may give rise to disagreements. Indeed, naturally, firms often focus on commercialisation and return on investment, while academic institutions prioritise scientific discovery, publication and academic recognition. These differing goals and incentives can lead to conflicts in terms of confidentiality and data sharing. One interviewed expert highlighted the potential benefits of developing framework or model non-disclosure agreements to facilitate collaboration between firms and universities.

Challenges around IP tend to be more prominent among ICT services companies. There are several possible reasons for this. First, the ICT services sector is a source of particularly rapid advances and innovations in AI, entailing frequent and significant developments in software, algorithms and digital technologies. These advances can lead to complex and rapidly evolving IP landscapes, making it more challenging to establish agreements and frameworks for IP management. Second, ICT services companies heavily rely on intangible assets such as software, algorithms, patents and copyrights, which are more difficult to protect and manage than tangible assets. The intangible nature of these assets makes it harder to establish ownership, usage rights and commercialisation agreements, potentially leading to disagreements.

Firms and universities often experience difficulties with respect to cultural and organisational differences. Most interviewees drew attention to the complexity of managing the distinct cultures, priorities and operational structures characteristic of corporate and academic environments. These diverse institutions typically have different approaches to decision making and timelines. The interviewed experts emphasise that academic institutions often operate on longer-term research cycles, while firms operate in fast-paced, market-driven environments. This disparity in timelines can pose challenges in terms of project co‑ordination, responsiveness to market demands and the ability to adapt quickly.

Lack of transparency in funding mechanisms and project governance in industry-university collaborations appears to be a roadblock. The interviewed experts emphasised the importance of establishing sustainable partnership models to foster enduring collaborations with academic research institutions. Other obstacles mentioned in some interviews were the lack of transparency in how the funding firms provide is used, how other developments within universities might affect a project (such as a turnover in postdocs) and overall project governance. Interviewees reported uncertainties in terms of funding allocation and accountability. The absence of clear guidelines, transparent processes and well-defined project governance structures can impede the smooth operation of collaborative projects and create a lack of clarity regarding roles and responsibilities, leading to delays, misunderstandings and even conflict.

Collaborative schemes tailored to the needs of SMEs help with the adoption of AI. One interviewee highlighted that centres of AI research predominantly focus on collaborations with medium and large-sized enterprises. At least in some locations, dedicated SME-focused programmes are scarce. Specialised programmes tailored to SMEs’ requirements could unlock a myriad of advantages. SME-specific collaborative schemes would allow smaller businesses to bring their data science problems to the table while gaining access to needed AI research and expertise. Dedicated programmes could help address specific challenges faced by smaller businesses, such as overall resource constraints and more limited access to AI talent.

Interviewees held that financial support for collaborations could help. The interviewed experts expressed concerns about their shouldering most of the financial risk when they sponsor academic research programmes. Given the inherent uncertainties and sometimes high costs associated with AI development (particularly as regards human expertise), the experts generally believed that public financial support would help to mitigate risk. That public financial support could be prioritised for enterprises’ first collaborative experience.

Companies would like less complex processes when applying for public funds to support AI research in collaboration with universities. Simplification and a lower administrative burden would make application more efficient. Interviewees also stressed the importance of enhancing transparency throughout the process. Clear guidelines, well-defined evaluation criteria, practical examples of successful applications, and accessible information about funding opportunities would help companies better understand expectations and requirements. Additionally, interviewees advocated for feedback loops that facilitate communication between funding agencies and applicants. This would help companies learn from previous experiences and increase future success rates.

Conclusion

In conclusion, the interviews with senior staff from enterprises in the manufacturing and ICT services sectors have provided valuable insights that can help inform policy makers in shaping effective strategies for AI adoption and development. The challenges and opportunities identified in data acquisition underscore the importance of streamlining access to public sector data. Policy makers can play a crucial role in establishing centralised platforms, standardised APIs, and cross-border data-sharing frameworks to enhance the quality, timeliness and accessibility of public data. Addressing procedural complexities and ensuring data relevance are essential considerations for policy makers seeking to foster a conducive environment for data-driven decision making in firms.

The findings related to other public services supporting AI adoption emphasise the need for more transparent, industry-specific information and guidance. Policy makers can explore options such as developing guidelines for vendor selection, especially for SMEs. Efforts to streamline application processes for public funding and enhance transparency can help address the financial constraints reported by enterprises, fostering a more supportive ecosystem for AI initiatives. Policy makers should also consider the industry-specific training needs highlighted by interviewees and bear in mind the potential for further development of curricula, as well as the importance of hands-on, work-based training and collaboration between public and private sectors in bridging talent gaps.

Finally, interviewees highlighted the importance of clear agreements, frameworks and funding mechanisms in industry-university partnerships. Policy makers could contribute by facilitating the development of standardised non-disclosure agreements and sustainable partnership models between universities, public research organisations and firms, particularly addressing the needs of SMEs. As regards financial support, simplified application processes and improved transparency in the evaluation criteria could help. Financial incentives could also encourage collaborations between enterprises and academic institutions and might focus on incentivising the first collaboration experience.

In summary, the interviews underscore the multifaceted nature of challenges and opportunities in AI adoption, providing policy makers with a nuanced understanding to help craft informed and effective policies.