The Adoption of Artificial Intelligence in Firms

Annex A. Comparisons among recent AI surveys

Table A.1. Selected features of recent national and supranational surveys of AI in firms

Survey (A)	Technologies, functionalities, activities (B)	Sectors and sample size (C)	Questions similar to those in the OECD/BCG/INSEAD survey
Canada “2017 Survey of Innovation and Business Strategy”	Internet of Things AI Geomatics or geospatial technologies Nanotechnology Biotechnology Blockchain	Multiple sectors, including utilities, manufacturing, mining, oil and gas. 13 252 firms, each with at least 20 employees and revenues of CAN 250 000 or more	Questions on obstacles to innovation which are of indirect relevance to AI diffusion barriers: - uncertainty and risk - skills - regulatory or government competition policy - internal funding - external financing
Germany “Survey on the Use of Artificial Intelligence, 2020”	Speech recognition Image or object recognition Pattern recognition Algorithmic decision making Automation of machines or vehicles	Unknown at the time of writing	Sourcing strategies Are the AI applications in your company based on software developed in-house, on open-source software, commercial software packages, or individual solutions for your company from external software providers? Barriers to adoption Please tell me in each case whether this aspect is a major problem, a minor problem or not a problem for your company when using AI: - difficulty finding appropriate use cases for AI - proof of the added value of AI over alternative methods - lack of employee skills in AI methods - integration of AI in existing systems, such as IT systems or machines, etc.
Germany “Artificial Intelligence and Industrial Innovation: Evidence From Firm-Level Data (ZEW)”	Language/text understanding Image/pattern recognition Machine learning Knowledge/expert systems	Mining Manufacturing Utilities Service sectors (wholesale, transportation, information and communication, banks and insurance, professional and technical services, business support services). Responses from 8 821 firms	Sourcing strategies, i.e.: - mainly developed in-house - mainly developed by others - both in-house and by others No related question on adoption barriers
Sweden “Artificial intelligence in Sweden (Statistics Sweden, 2020[1])”	Do firms use AI to: - develop or increase knowledge of customers or users -develop a new product or service -improve an existing product or service -develop or improve internal processes -other usage	Manufacturing Energy and recycling Construction Wholesale and retail trade; repair of motor vehicles and motorcycles Transportation and storage Sample size: 3 831 firms	Obstacles to the use of AI, i.e.: - knowledge of existing technologies and applications - employees´ skills, training or experience - compatibility with existing software or hardware - data (e.g. quality issues, lack of data) - services or equipment costs - data security or data integrity
United Kingdom “Understanding the UK AI Labour Market: 2020”	Robotics – training robots to interact with the world in generalisable and predictable ways Computer vision – gaining high-level understanding from digital images or video Natural language processing Collaborative systems – autonomous systems that can work collaboratively with other systems and with humans Bio-inspired computing models – this includes evolutionary algorithms and algorithmic game theory Bio-inspired hardware – new forms of AI-enhanced hardware, e.g. using neuromorphic computing techniques Edge intelligence – combining AI with edge computing, as in the Internet of Things and “smart home” devices Classification – assigning a class or label to a previously unseen input, e.g. to identify spam emails Predictive machine learning – estimating the value of a discrete variable based on historical data Regression for machine learning – estimating the value of a continuous variable based on historical data	118 AI firms, including firms whose core business was developing AI-led products or services and others in wider sectors developing or using AI tools, technologies or techniques to improve their products, services or internal processes.	What AI is used for: - to predict - to automate Policy priorities and support relevant to data: - providing funding (e.g. loans, grants, tax benefits) - the data-related regulatory framework - access to data science talent
United States “Advanced Technologies Adoption and Use by US Firms: Evidence from the Annual Business Survey 2020 (US Census Bureau)”	Augmented reality Machine learning Machine vision Natural language processing Cloud computing Robotics Radiofrequency identification (RFID) Automated vehicles	All private, non-farm sectors, including agriculture, manufacturing, finance and healthcare 583 000 firms (i.e. responses)	Technologies used (see Column B) No related questions on AI sourcing strategies or adoption barriers
United States “Survey of Manufacturers, 2019, Information Technology and Innovation Foundation”	AI Digital modelling/prototyping (CAD, CAE, CAM) Cloud computing Industrial robotics Computer numerical control (CNC) machining 3-D printing Internet of Things Big data analytics Augmented reality/Virtual reality (AR/VR) Digital twins	Manufacturing 60 responses (enterprises with annual turnover between USD 500 million and USD 10 billion)	No related questions on sourcing strategies or adoption barriers
European Union “European Enterprise Survey on the Use of Technologies based on Artificial Intelligence 2020”	Natural language processing Anomaly detection Computer vision Sentiment analysis Machine learning Recommendation and personalisation engines Process optimisation Process automation Autonomous machines Creative and experimentation activities	Wide variety of NACE sectors: from Sector A (Agriculture, Hunting and Forestry) to Sector Q (Extraterritorial Organisations and Bodies) Aims to achieve responses from 9 640 firms from across the EU27	A number of the processes listed in Column B (particularly process optimisation, process automation and autonomous machines) AI sourcing strategies Various obstacles to adoption, i.e.: External obstacles: - public or external funding - liability for damage caused by AI - the need for new laws or regulations - access to high-quality private data - access to or availability of public data Internal obstacles: - hiring staff with the right skills - the cost of adoption - the cost of adapting operational processes - skills of existing staff - understanding of algorithms - IT infrastructure - internal data
European Union “Eurostat, Community Survey on ICT Usage and E-commerce in Enterprises 2021”	AI Text mining Machine learning Computer vision Speech recognition Natural language processing Deep learning Robotic process automation Autonomous robots Self-driving vehicles Autonomous drones	NACE Rev. 2 Sections C to N, excluding Section K, but including manufacturing The survey population consists of enterprises with ten or more employees. Out of around 1.5 million EU enterprises with at least ten persons employed, a sample of almost 142 000 were surveyed (survey 2020). Of the 1.5 million enterprises, approximately 83% were small enterprises (10‑49 persons employed), 14% medium (50‑249) and 3% large (250 or more).	Sourcing strategies, i.e.: - developed by own employees - commercial software or systems modified by own employees - open-source software or systems modified by own employees - ready-to-use commercial software or systems - external providers contracted to develop or modify Barriers to adoption, i.e.: - cost - expertise in the enterprise - incompatibility with existing equipment, software or systems - difficulties with availability or quality of necessary data - concerns regarding data protection and privacy - lack of clarity about legal consequences (e.g. liability in case of damage caused by the use of AI) - useful or not for the enterprise
European Union + European Investment Bank “Artificial Intelligence, Blockchain and the Future of Europe: How Disruptive Technologies Create Opportunities for a Green and Digital Economy 2021”	AI Blockchain	100 SMEs located in the 27 member states using AI and blockchain	This study aimed to identify and address general market failures and access-to-finance barriers faced by SMEs (although survey questions are not included in the overall report).

Source: OECD desk research.

References

[1] Statistics Sweden (2020), Artificiell intelligens i Sverige (Artificial intelligence in Sweden), Statistics Sweden, https://www.scb.se/contentassets/4d9059ef459e407ba1aa71683fcbd807/nv0116_2019a01_br_xftbr2001.pdf.