Artificial Intelligence (AI) is everywhere these days. It’s writing emails, helping with customer service, spotting typos, and even making playlists. But how did AI get here?
AI has come a long way and has undergone a remarkable transformation — from a tool that analyzes data to AI that can now create content and manage complex goals. AI has been around for a long time, and it impacts our world in many ways. Here’s a closer look at how AI has evolved from traditional systems to generative models and the emerging frontier of agentic AI. We also examine how each stage transforms enterprise operations and what businesses need to prepare for next.
The early days: Traditional AI as a predictive workhorse
While AI may seem like a recent phenomenon, its roots trace back to the 1950s. In its early form, AI and machine learning focused on pattern recognition and predictions. Traditional AI — sometimes referred to as predictive AI — relies on clearly defined rules and statistical models to analyze large datasets and infer likely outcomes. Its scope is narrow, requiring human input to define what the AI should do. It is useful for automating certain manual tasks.
Most people have used traditional AI without even realizing it. It’s found in search engines, auto-correct in your documents, virtual assistants, and tools that filter spam or suggest calendar times. Traditional AI is embedded everywhere, working behind the scenes to help us save time and make more informed decisions. Enterprises also rely on traditional AI for business applications and work tools.
The big shift: Generative AI as a creative engine
Generative AI is relatively new and came to the forefront in 2023 when tools like ChatGPT became more mainstream. It is used in schools and businesses alike. However, data scientists have been working on generative AI since about 2010.
Instead of analyzing data and recognizing patterns, generative AI can create things — new content such as text, images, videos, even code. You give it a prompt, and it generates something new. The other thing about generative AI is that it appears very human-like. It's really leaning into an AI technology called natural language processing, which helps it understand how we talk and write, so conversations with AI feel surprisingly natural.
What’s up now? Agentic AI focused on goal-driven execution
Now we’re entering a new chapter — agentic AI. Instead of giving AI one prompt at a time, as in generative AI, you can give agentic AI a high-level goal, and it will figure out the steps to achieve this goal across systems, tools, and processes. For example, imagine you are a retailer running an extensive network, and your goal is to have your point of sale applications running 24/7. Agentic AI will look at your installed network, explain how to configure devices, and determine if you need to implement specific policies. Agentic AI can ensure your entire payment system stays connected, preventing lost sales. That’s powerful!
Looking beyond Agentic AI
Agentic AI isn’t the final stage. Experts anticipate the arrival of Artificial General Intelligence (AGI) — systems with open-ended reasoning and the ability to learn and adapt to any task. While we’re not there yet, AI is a rapidly evolving technology.
How does AI impact enterprises?
Overall, AI has had a major impact on how enterprises run their operations. We can offload more tasks to AI than we could in the past. Traditional AI helped automate manual repetitive tasks like analyzing large amounts of data, logs, or flagging issues. It’s not efficient for humans to do that, but very efficient for machines.
We use generative AI for creative tasks, from producing images, videos, and texts to coding software. Agentic AI holds the promise of being able to oversee complex workflows or systems, like a manufacturing supply chain, for example, with more autonomy.
Powering AI with resilient and secure connectivity
As AI becomes more capable, enterprises must ensure their networks are ready. Resilient and secure connectivity is the backbone of AI. Here’s why:
- AI needs continuous access to data for training and inference. AI models require large volumes of data, often streamed from sensors, edge devices, or cloud platforms. This requires real-time processing for AI applications such as predictive maintenance, connected vehicles, video analytics, and more. Connectivity must be stable, high-bandwidth, and with low latency.
- AI is vulnerable without protection. AI systems often process sensitive data and widen attack surfaces across cloud, edge, and devices, creating multiple entry points for cyber attacks. Connectivity must be encrypted, authenticated, and built on zero trust.
- AI operates across networks. IoT devices, cameras, and industrial sensors use local edge AI models that need to sync with central systems and networks. Agentic AI will operate across systems and processes, requiring secure APIs and infrastructure.
Leading AI in networking
At Ericsson, we are an industry leader for AI in networking, and our approach to AI has two aspects. First, we recognize that AI is the most transformative technology since the Internet, which benefits our customers. With many AI applications and their increasing power, enterprises can reduce costs, increase efficiency, and accomplish more with less. We can provide the reliable 5G connectivity to ensure that the AI applications organizations rely on, whether in a factory or a retail store, perform as promised.
Second, we use AI technology in our Ericsson NetCloud management and orchestration platform to simplify the network's design, deployment, operation, and troubleshooting. Here are specific ways we are applying AI.
Meet ANA, our generative AI-based virtual expert
ANA is our virtual expert within NetCloud, which leverages generative AI to assist our customers in managing Ericsson’s Wireless WAN, SD-WAN, and security solutions. NetCloud is a comprehensive cloud-managed networking solution that provides secure, scalable, and simplified connectivity across branch sites, vehicles, and IoT deployments. Hosted internally to protect against data leaks, ANA is trained on the entire knowledge base of technical documents, reference architectures, and release notes about Ericsson Cradlepoint products and technologies.
Because ANA is built on generative AI, ANA can read, understand, and infer to generate new content to answer customers’ queries directly. For instance, ANA can troubleshoot connectivity issues by providing the relevant steps to resolve the problem instead of linking to a document. This helps customers increase productivity and simplify operations.
Detecting network anomalies with advanced AIOps capabilities
Another example of how we have embedded AI into our products is our NetCloud AIOps, which enhances the network's resilience and quality of experience. With AIOps, it flags performance anomalies before these issues impact services. By providing a baseline of traffic to detect latency and jitters specific to the customer’s environment, AIOps offers fault detection, isolation, root cause analysis, and recommendations for resolving the issues. Currently available as part of our SD-WAN and security solutions, NetCloud AIOps will also extend to our WWAN solutions as well.
Keeping up with the pace of AI innovation
Looking ahead, we are committed to keeping pace with AI innovation. One thing is certain: the future of AI is helping us unlock ways of working better, faster, more innovatively, and with greater agility.