How does artificial intelligence work?
Here at Gartner, we define artificial intelligence (AI) as applying advanced analysis and logic-based techniques, including machine learning (ML), to interpret events, support and automate decisions and to take actions. This definition is consistent with the current and emerging state of AI technologies and capabilities, and it acknowledges that AI now generally involves probabilistic analysis (combining probability and logic to assign a value to uncertainty).
Other organizations and individuals may use different definitions. There is no single, universally accepted descriptor for artificial intelligence as there is such a wide range of ways in which AI can support, augment and automate human activities, and learn and act independently (see “What is machine learning?”).
To capture the opportunity of AI as an organization, however, you will need to articulate and agree on a generally accepted definition focused on what you want AI to accomplish. (See “What is enterprise AI strategy?”).
Leave room for differences of opinion, but make sure that business, IT and data and analytics leaders don’t fundamentally disagree about what AI means to the organization or you will be unable to design a strategy that captures the benefits.
Note that AI technology vendors also are likely to have their own definitions of the term. Ask them to explain how their offerings meet your expectations for how AI will deliver value.
What are machine learning and deep learning?
Machine learning is a critical technique that enables AI to solve problems. Despite common misperceptions (and misnomers in popular culture), machines do not learn. They store and compute — admittedly in increasingly complex ways.
Machine learning is a purely analytical discipline. It applies mathematical models to data to extract knowledge and find patterns that humans would likely miss. ML also recommends actions, but it does not direct systems to take action without human intervention.
More specifically, machine learning creates an algorithm or statistical formula (referred to as a “model”) that converts a series of data points into a single result. ML algorithms “learn” through “training,” in which they identify patterns and correlations in data and use them to provide new insights and predictions without being explicitly programmed to do so.
Deep learning, a variant of machine learning algorithms, uses multiple layers of algorithms to solve problems by extracting knowledge from raw data and transforming it at every level. Deep learning can outperform traditional ML (or shallow learning techniques) by working with complex and often high-dimensional data, such as images, speech and text. Still, either rule-based systems or traditional ML can effectively solve many AI problems.
In most organizations, deep learning solutions are not yet a significant part of the product roadmap (rule-based systems or traditional ML can effectively enable most AI use cases today), but their use is quickly increasing alongside advancements in data processing and breakthroughs in computational techniques.
Using ML, including deep learning, to make predictions enables an AI-driven process to automate the selection of the most favorable result, which eliminates the need for a human decision maker.
What are some other key AI technology terms?
- Natural language processing (NLP) enables an intuitive form of communication between humans and intelligent systems using human languages. NLP drives modern interactive voice response (IVR) systems by processing language to improve communication. Chatbots are the most common application of NLP in business.
- Advanced virtual assistants, sometimes called conversational AI agents, are powered by conversational user interfaces, NLP, and semantic and deep learning techniques. Progressing beyond chatbots, advanced virtual assistants listen to and observe behaviors, build and maintain data models, and predict and recommend actions to assist people with and automate tasks that were previously only possible for humans to accomplish.
- Computer vision (CV) is a process that can capture, process and analyze real-world images to allow machines to extract meaningful, contextual information from the physical world. CV techniques have technology and infrastructure requirements that differ from traditional ML approaches. CV, which is becoming more accurate at identifying organic objects, underpins the development of applications such as self-driving cars, autonomous drones and automated retail stock checks.
- Edge AI refers to the AI techniques embedded at the touchpoint where physical devices meet the digital world — for example, where a sensor on a factory floor connects to the internet and can autonomously send data to place a service request. Edge AI or AI at the edge underpins the Internet of Things (IoT).
- The Internet of Things (IoT) comprises the network of physical objects (things) that contain embedded technology to sense or interact with their internal workings and the external environment. This doesn’t include general-purpose devices, such as smartphones. Examples of IoT in action range from smart plugs to driverless vehicles. IoT relies on a wide range of IT endpoints and gateways to function and data to drive the AI, especially for real-time responses (e.g., for autonomous vehicles).
- Generative AI learns about artifacts from data and generates innovative new creations that are similar to but don’t repeat the original. Generative AI has the potential to create new forms of creative content, such as video, and accelerate R&D cycles in fields ranging from medicine to product development.
- Synthetic data is artificially generated through machine learning. It mirrors the statistical properties of real data, but it doesn’t use that data’s identifying properties (e.g., names and personal details). AI needs huge amounts of data to generate usable outcomes, and synthetic data will be an important source of large datasets that can model outlying scenarios while protecting sensitive and personal data.
How is generative AI different from other forms of AI?
Generative AI refers to AI techniques that take artifacts from data and use them to generate novel material that retains a likeness to the originals but doesn’t repeat them. It can produce media (such as text, images, audio, code and video), learning methods, processes, applications, synthetic data and models of physical objects.
Today, generative AI most commonly creates content in response to natural language requests — that is to say, it doesn’t require knowledge of or entering code — but we’re likely to see an increase in multimedia inputs and outputs over time. Generative AI grabbed mainstream attention in late 2022 with OpenAI’s release of ChatGPT, a chat-product innovation driven by the Generative Pre-trained Transformer (GPT) large language model (LLM). Transformer models, such as DALL·E 2 and BERT, are similarly generating novel images and other content.
These transformer models and their human-like outputs are fueling great interest and investments in AI. Like other emerging technologies, however, generative AI comes with both hype and risk. After appearing on our list of 2022 of Top Strategic Technology Trends, we currently place it within the Peak of Inflated Expectations on the Gartner Hype Cycle™ for Artificial Intelligence, and predict that it will reach the Plateau of Productivity in two to five years.
IT and data analysis leaders can use AI techniques to solve a wide array of business problems and can generate significant returns on investment; however, the question for most organizations is how to use artificial intelligence to create or accelerate the growth of digital business.
The main opportunities of artificial intelligence lie in its ability to:
- Reveal better ways of doing things through advanced probabilistic analysis of outcomes
- Interact directly with systems that take actions, enabling the removal of human-intensive calculations and integration steps
Gartner research consistently shows that CIOs see an enormous opportunity in the benefits of AI but still struggle to capture those advantages in practice. Nevertheless, AI will ultimately reshape how work is done as the technology replaces some tasks typically performed by employees and changes how day-to-day decisions are made. Use cases mainly fall into three categories: automating and optimizing, generating insight and creating human-like engagement (e.g., chatbots and virtual assistants). (See “What are examples of artificial intelligence applications in business?”).
For now, however, AI hype can be rife, making it difficult for some organizations to set the right expectations regarding business outcomes. Untamed hype gives rise to projects that have no chance of success. When that happens, business leaders with unrealistic expectations blame the technology and science for its inability to create the transformations for which they hoped.
Make sure to establish an enterprise strategy for AI to identify use cases and metrics of success from the outset. Common ways of measuring benefits include risk reduction, speed of process, improved sales, increased customer satisfaction, and reduced labor needs or costs. Many business cases rely on a combination of tangible and intangible benefits. (See “What is enterprise AI strategy?”).
What are examples of artificial intelligence applications in business?
As an emerging technology, AI’s full impact and benefits have yet to be realized. AI innovation is one of a multitude of forces disrupting existing markets and enabling new digital business initiatives, for example. But AI is also being applied across industries, organizations and functions in a range of ways. A few examples from business operations are:
- Machine learning as a backbone for human-like communications. ML drives common AI applications like chatbots, autonomous vehicles and smart robots.
- Deep learning techniques provide biometric solutions using facial recognition, voice recognition and neural networks that hyper-personalize content based on data mining and pattern recognition across huge datasets.
- AI in the IT operations/service desk. Virtual support agents (VSAs) provide IT support in an IT service management (ITSM) scenario alongside the IT service desk. AI can also be useful for ticket routing, pulling information from knowledge management sources and as an ITSM tool to provide answers to common questions.
- AI in supply chain management. Use cases include predictive maintenance, risk management, procurement, order fulfillment, supply chain planning and promotion management. AI can also be useful for decision-making automation because it is orders of magnitude more consistent and faster than humans at specific tasks.
- AI in sales and sales enablement. Identify new leads and opportunities based on similar existing customers, nurture prospects by establishing relationships through intelligent activity tracking and messaging, and use guided selling to improve sales execution and increase sales revenue.
- AI in marketing can help with real-time personalization, content and media optimization and campaign orchestration to augment, streamline and automate marketing processes and tasks otherwise constrained by human costs and capability. The most compelling value proposition is AI’s ability to uncover new customer insights and accelerate marketers’ ability to deploy them at scale.
- AI in customer service can predict what customers will ask for and proactively deflect inbound inquiries. Virtual customer assistants (VCAs) with speech recognition, sentiment analysis, automated/augmented quality assurance and other technologies provide customers with 24/7 self- and assisted-service options across channels.
- AI in human resources. Use cases include recruitment (matching talent supply and demand or predicting recruitment success) and skills (using NLP to establish consistent skill and job ontologies for next-generation search and matching). HR is also leveraging recommendation engines for learning content, mentors, career paths and adaptive learning.
- AI in finance. The best candidates for near-term AI enablement are dynamic processes that require judgment and involve unstructured, volatile and high-velocity data. Examples include complying with new accounting standards, reviewing expense reports and processing vendor invoices.
- AI in sourcing, procurement and vendor management (SPVM). Basic ML technologies are being deployed for spend classification and contract analytics, but more sophisticated use cases are emerging in areas such as risk management, candidate matching (within contingent workforce management), sourcing automation, virtual purchasing assistance and voice recognition.
- AI in legal. Common applications include contracts (assembly, negotiation, due diligence, risk scoring and life cycle management), e-discovery (document classification, data extraction and text analysis), and spend (invoice classification).
What is enterprise AI strategy?
For a business to capture the benefits of AI, executive leaders should establish an enterprisewide AI strategy that identifies use cases, quantifies benefits and risks, aligns business and technology teams and changes organizational competencies to support AI adoption.
To ensure you derive value from AI, choose initiatives strategically, focusing on what your organization is trying to accomplish and the business problems you’re working to solve. For AI to really take off, you’ll need to employ AI as part of your existing application family — and that includes having data from every area of the business to power the features it offers.
Organizations at the earlier stages of AI maturity are more likely to pursue use cases around cost control before advancing to key elements of the value proposition, such as customer experience. Gartner research shows that as maturity grows, AI is applied more broadly and more impact is realized.
Key elements of enterprise AI strategy are:
- AI vision.Tie AI goals to enterprise ambitions. For example, articulate how AI will enable digital transformation objectives. Outline approaches and focus areas designed to encourage and enable organizationwide fluency and adoption of AI. Get specific about success metrics.
- AI risks. Assess your exposure to and mitigation plans for different key areas of risk, including regulatory (e.g., privacy laws), reputational (e.g., AI bias) and organizational (e.g., lack of competencies or infrastructure).
- AI strategic action plan. Identify the impact on business models, processes, people and skills, and take a portfolio approach to AI opportunity. Assign accountability for AI strategy development and execution. Interdisciplinary teams and data literacy will be key to success.
- AI adoption. Spell out the use cases (human-like engagement, process optimization, generating insight, etc.), and use value maps and decision frameworks to prioritize adoption.
- Pursue buy-in to the AI program. Evangelize the initiative’s launch and subsequent successes to peers, and give other C-suite leaders the ability to tell the AI team’s stories.
What is the future of artificial intelligence and AI technologies?
The AI discipline is evolving rapidly through new techniques, dedicated infrastructures and hardware. Over the next five years, Gartner expects organizations to adopt cutting-edge techniques for smarter and more reliable, responsible and environmentally sustainable artificial intelligence applications.
The trajectory of AI now more closely follows that of technologies that have preceded it. For companies and governments, AI is becoming more:
- Familiar: IT tools and skills are now AI-friendly
- Scalable: AI is cheaper, and success is more achievable than ever
- Useful: IT and business leaders more frequently consider AI as a way to improve applications
Going forward, organizations will continue to pursue AI to enhance their decision-making processes. Savvy ones that adopt these methods quickly will drive more competitive differentiation and become more agile and more responsive to ecosystem changes.
Executing AI strategies remains a challenge for infrastructure and operations teams. Starting on-premises means investing in infrastructure and architecture that can be difficult to predict, staff and fund. That makes cloud options appealing, but as the need for AI grows and the required investment increases, the cloud may become harder to afford (and commitment to cloud providers more concerning). That’s why the emergence of strategies that balance investment in cloud function with investments in infrastructure are so attractive (so-called cloud/on-premises hybrid strategies).
Among Gartner strategic planning assumptions for AI are that by 2025:
- 50% of enterprises will have devised AI orchestration platforms to operationalize AI, up from fewer than 10% in 2020
- AI will be the top category driving infrastructure decisions, due to the maturation of the AI market, resulting in a tenfold growth in computing requirements.
- 10% of governments will use a synthetic population with realistic behavior patterns to train AI, while avoiding privacy and security concerns.
Can businesses trust artificial intelligence?
Most business organizations do not know or understand the inner workings of artificial intelligence, which creates potential for concerns about fairness, security and privacy. But AI cannot thrive if the business does not trust AI techniques, so organizations need checks and balances to assess and respond to threats and damage and to ensure integrity is embedded into AI.
Gartner refers to its AI risk management framework as “MOST” because it has the following three pillars:
- Model Operations, to support the reliability, predictability and accuracy of the AI
- Security, to prevent hackers and malicious insiders from manipulating AI inputs, applications and outcomes
- Trustworthiness, to support AI fairness, ethics, societal well-being and “responsible AI” overall
As AI goes mainstream in an enterprise, threats will inevitably follow and result in serious organizational risks. Organizations must evaluate the threats proactively. In doing so, they can increase stakeholder trust in AI.
Indeed, Gartner expects that by 2025, regulations will necessitate a focus on AI ethics, transparency and privacy, which will stimulate — instead of stifle — trust, growth and better functioning of AI around the world.