AI Trends: What to Expect in the Year Ahead

Interest and investment in artificial intelligence (AI) are at an all-time high: ChatGPT was the #1 Wikipedia page and one of the top Google search terms in 2023, and global spending on AI could exceed $500 billion in the next few years.

AI is an incredibly complex and fast-moving area. In the next few years, AI will continue transforming many industries, including healthcare and field service. It is a game-changer for strategic planning, scheduling, data management, and many other business functions.

First things first, a definition:

Artificial intelligence (AI) is a category of technologies that enable machines to perform functions similar to human cognition. AI technologies are combined into systems and trained on massive amounts of data to recognize patterns in written, visual, and/or numerical data. With those patterns—and lots of fine-tuning—AI models can analyze data, make recommendations, replicate dialogue with natural language processing, and even generate new images and videos.

Some AI models are polished, end-user applications, while others are underlying components that support other technologies. There are broad AI models that can perform a wide variety of functions, and there are specialized models that are narrowly tailored to specific tasks.

Training an AI model requires huge amount of data. Google trained its large language model, LaMDA, on 1.56 trillion words, 2.81 trillion tokens (strings of characters), and 137 billion parameters.

AI terms to know

Several AI terms are household names: technologies like generative AI, providers like OpenAI, and publicly available platforms like ChatGPT and DALL·E. Key terms are listed below:

Natural language processing

Natural language processing is the ability for a machine to understand human speech. It is a collection of not only black-and-white rules, like grammar and spelling, but also nuances and contextual elements of language. Natural language processing (NLP) is a key component of many AI technologies, especially large language models.

Large language models

Large language models (LLMs) are foundation models trained on a huge quantity of text to identify patterns in conversation. With enough data, training, and fine-tuning, LLMs can predict and generate more accurate, realistic dialogue at scale.

Machine learning

Machine learning is a specific type of AI that helps systems learn from data autonomously. Algorithms teach the machine how to interpret data and how to improve the system going forward, compared to other AI systems that depend on training and fine-tuning instead.

Generative AI

Generative AI refers to AI tools that can generate content based on an underlying deep learning (foundation) model. ChatGPT is a prime example of the massive potential in generative AI, which could be worth trillions of dollars in yearly global productivity. In a survey of 1,000 workers, Generative AI was the most commonly used AI technology, and most respondents believe AI will improve their efficiency (82%) and ability to focus on higher value work (81%).

Emerging AI technologies to watch

There are many emerging AI technologies that will make waves in the next two years. Here are two AI use cases likely to grow in 2024 and 2025:

Retrieval-augmented generation (RAG)

Retrieval-augmented generation (RAG) is a type of LLM that pulls information from external knowledge sources to improve the quality of its responses. RAG is used in applications like customer service chatbots, which can provide more accurate responses by accessing up-to-date information from various databases. Retrieving this data from external sources, rather than hosting locally, helps reduce the size and increase the speed of the LLM. It also ensures the model has the most current data and information can be traced back to its original source(s).

Multi-modal AI systems

Multi-modal AI systems can process a wide variety of input and output formats—not only text or only images. These inputs and outputs could be video, audio, speech, numerical datasets, images, or text. As multi-modal systems improve and become more energy-efficient, it could make a large impact in fields that require analyzing many unique variables, like healthcare. OpenAI’s GPT-4 already supports text, images, and speech—and GPT-5 is expected to be “fully multimodal with speech, image, code, and video support.” Google’s Gemini is also rapidly adding multi-modal functionality and will continue to do so in the coming months.

Honorable Mentions

Digital twins – create virtual models of real-world buildings, systems, and cities and run scenarios based on an AI model
Decision intelligence – weigh decision factors using causal AI
Edge AI – Internet of Things (IoT) components that include AI
Synthetic data – use artificial data like simulations or samples to train AI models instead of personally identifiable data

5 AI trends for 2024 and 2025

What’s next for AI on an enterprise scale? Here are a few trends to watch in 2024 and 2025.

1. Less hype, more integration

As AI becomes a more common part of everyday life, in a strange way, it will become less visible. Going forward, there is likely to be less hype about AI as a concept, while common tools become “invisibly smarter” due to AI-based features.

We will continue to see everyday tools—for business and personal use—integrate AI features into their interface. You can already see this trend at work for two major tech companies: Copilot (for Bing and Microsoft Edge) and Bard (for Google search and Google Chrome) are conversational AI tools that allow users to search the web and refine results using natural language patterns. AI features in tools like Microsoft Office 365 help users write emails, analyze data, and schedule meetings more efficiently, without the users even realizing they are using AI.

2. AI models improve quality, consistency, and reliability

Early experiences of AI “hallucinations” can erode trust—but as time goes on, AI models get more reliable. More data from user interactions and more time spent testing and fine-tuning the model will net better, more consistent results. Companies like OpenAI and DeepMind are continuously improving their models to reduce instances of AI hallucinations, making these tools more reliable for business applications.

The rapid advancement from large providers is good news for reliability in the larger market. Google announced its large language model PaLM (Pathways Language Model) in April 2022, its successor PaLM 2 in April 2023, and then Gemini in December 2023. Only a few short years after the PaLM launch, Gemini 1.5 Pro outperformed its predecessor in 87% of Google’s benchmark tests.

The general public will see this improvement through the use of tools like ChatGPT, DALL·E, and Midjourney. More reliable models will reduce instances of inaccurate content and “coherent nonsense. But these models still have major challenges to overcome for enterprise use. When it comes to commercially available generative AI like ChatGPT, data science leaders were most concerned about security, reliability, and IP protection in a 2023 survey.

3. The focus on responsible AI increases

Responsible AI is an umbrella term for the ethical, legal, and business principles involved in working with AI. It involves how to mitigate risk and bias, how to promote fairness and accountability, and how to ensure compliance, safety, and privacy.

The challenges of responsible AI are well-documented. Human decision-making is flawed due to cognitive biases, limited datasets, preconceived notions, and much more—and AI can magnify both the good and the bad. New advancements in image generation and video generation raise new questions about AI and ethics, intellectual property, and copyrighted material.

“In 2024, trust in AI will be critical. It’s not just about data security and compliance, but an approach that is based on responsible AI and the data practices and policies that ensure trust is built in, not bolted on.”

Expert AI

There is momentum toward responsible AI legislation and standard-setting in multiple countries, including the Artificial Intelligence Act from the European Commission and the Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence in the U.S. Major tech firms like IBM and Microsoft are leading initiatives to develop ethical AI guidelines and tools that help ensure AI systems are fair, transparent, and accountable.

But it’s a complex area with lots of regional differences, and Gartner estimates the peak of responsible AI is still at least five years away.

In the near future, explainability will become an important part of building, customizing, or buying an AI solution. Companies will seek foundation models that can explain the reasoning behind its insights and conclusions, which helps ensure the model follows ethically and legally sound principles.

“Enterprises in 2024 can differentiate through bespoke model development, rather than building wrappers around repackaged services from ‘Big AI.’”

IBM

4. Enterprise customization becomes the standard

In the near future, look for more companies to develop bespoke AI models. With an in-house data pipeline and custom model, an enterprise can build, customize, deploy, and fine-tune a tool based on their proprietary data and business needs.

This will enable companies to evolve from individual or small team use cases to large, enterprise-level use cases.

Highly specialized fields will benefit the most from this development. Healthcare, law, advanced sciences, international compliance, and other highly specialized fields have unique vocabulary and concepts that are not part of the dataset used to train “off the shelf” AI tools. In these cases, customizing or adapting a model with this specialized knowledge is the best option. For example, pharmaceutical companies are developing custom AI models to accelerate drug discovery, tailored to their specific research data and objectives.

5. AI enhances decision-making in healthcare

Clinical decision support is the use of advanced AI technology to support patient care. Skilled healthcare providers can use AI tools to recognize patterns in medical images and vital signs. The system helps manage the complicated, interconnected factors that contribute to diagnosis and treatment: patient demographics, symptom presentation, health history, and much more.

Medical imaging is an early success story of how AI can augment human capabilities. AI-supported clinical decision tools outperform a traditional metric, the Modified Early Warning Score (MEWS), used in hospitals to measure the risk of imminent clinical deterioration. Startups like PathAI are using AI to improve the accuracy of pathology diagnostics, providing critical support for medical professionals in making informed decisions.

Implementing AI in healthcare is particularly complex, but results to-date are promising. A review published in 2024 stated:

“AI tools have been applied in various aspects of healthcare decision-making. The use of AI can improve the quality, efficiency, and effectiveness of healthcare services by providing accurate, timely, and personalized information to support decision-making.”

Going forward, healthcare organizations will pursue new and improved ways to use AI to empower clinicians and overcome barriers to implementation.

Everyday uses for AI in business in 2024

Heading in 2025, business leaders will find additional practical ways to use AI in day-to-day operations. Generally speaking, the business case for AI fits into three categories: operations, data, and staff.

Operations

Use AI to streamline and improve business operations. For example, AI-driven inventory management systems help retailers like Walmart reduce overstock and stockouts by accurately predicting product demand.

Smart scheduling – automatically schedule the right resources for the right work
Internal communications – share job details and schedule updates
Customer communications – answer common questions, route requests
Work orders – automatically create, then autofill known info
Route optimization – provide the best route between work sites
Marketing – generate personalized marketing copy at scale
Troubleshooting – access technical data in a user-friendly interface

Data

Use AI to improve data quality and access to key metrics for informed decision-making. For example, financial institutions use AI for fraud detection, analyzing transaction patterns in real-time to identify and prevent fraudulent activities.

Data analysis – identify trends and outliers
Data reporting – generate KPI reports and dashboards
Transparency – share job details based on enterprise permissions
Trend analysis – isolate specific variables and build trends across overlapping data points
Compliance – meet reporting requirements for compliance activities

Staff

Use AI to improve employee coaching, training, and professional development. For example, AI-based learning platforms like Coursera and LinkedIn Learning offer personalized training programs that adapt to the individual learning pace and style of employees.

Recruiting – identify prospects and perform basic vetting
Training – offer personalized training and coaching to staff
Onboarding – equip workers with technical guidance and best practices
Communication – improve emails, job notes, and other written tasks
Continuing education – track and renew staff credentials and licenses

Prepare for the future of AI

We don’t know exactly what the future holds for AI. Some experts predict bigger models that can handle more data, while others predict a shift toward smaller, more specialized AI models that require less energy.

In the face of the unknown, organizations can improve their resilience by creating a thoughtful AI strategy. Businesses should start by conducting an AI readiness assessment to understand their current capabilities and identify areas for improvement. Investing in employee training on AI tools and ethics is essential for successful implementation. Invest in company software and tools with AI functions that are useful, reliable, and always improving.

Skedulo offers automated scheduling and intelligent optimization in a user-friendly platform. On average, Skedulo customers who use the AI-powered optimization engine complete 64% more jobs and spend 17% less time traveling between jobs.

Read more about how AI can transform workforce scheduling.

AI Trends: What to Expect in 2024 and 2025