Agentic AI is a seriously hot topic right now. With its autonomous reasoning and problem-solving skills, it promises a huge step forward for the way we work. By intelligently automating processes and workflows, it offers to boost productivity, cut costs, and allow employees to focus on more innovative tasks. But what exactly is agentic AI, how does it work, and what challenges are enterprises facing that make agentic AI so relevant?
Enterprises Have More Data Than Humans Can Cope With
Many enterprises are sinking under the weight of the data that they, their customers, and their partners generate. In 2019, it was found that the average company was collecting data from 400 different sources, and more than a fifth of organizations were using 1,000 or more data sources. That figure is likely to be even higher today.
Some of this data is already being used in the day-to-day running of the business, but it takes up significant human time to process because it’s unstructured. Think of documents like invoices, for example, which can have any number of different formats.
Other information – like social media postings showing how customers feel about products, or operational data like network logs – is just too prolific for humans to be able to fully harness.
Enterprise Operations Are Incredibly Complex
Enterprise processes and workflows can be long and complicated, and often involve retrieving data and completing actions across a number of different systems – some estimates suggest that a large enterprise uses an average of 190 applications.
Imagine, for example, a broadband provider’s customer support ticketing system.
A customer might use email or live chat to contact their provider with a problem, and a ticket is created in response. That ticket is routed to the right team to deal with the problem, such as the finance department for billing issues, or a technical team if the customer has a problem with their service.
The progress of that ticket, including updates, actions, and communications, is logged. Once the problem is resolved, the ticket is closed, the customer receives a notification, and they may be invited to share feedback.
These multistep processes can span different departments and employees, and a whole variety of systems, services, and applications.
Historically, human-level reasoning has been needed to manage this level of complexity and unpredictability. But with agentic AI, this is changing.
What Is Agentic AI?
Agentic AI is a form of artificial intelligence that can autonomously work out the best way to achieve specific goals. It can make decisions, perform actions, and change its strategies in response to new situations and fresh information.
Each software system is known as an AI agent. These agents use machine learning (ML), reinforcement learning, and complex algorithms to carry out tasks, even ones with multiple steps, with minimal or no human intervention.
Agentic AI has the capacity to go beyond rigid rules and single-step actions. Instead, it can break down overall goals and workflows into smaller tasks and independently make decisions about the best way to achieve them. This makes agentic AI different from GenAI, which is designed to generate outputs like images, text, and programming code, and can’t reason independently.
Because they’re adaptable, AI agents can make adjustments on the fly, making processes faster and more efficient. They can also switch between different tools and systems to achieve the end goal.
Half of North American organizations are already actively integrating agentic AI into applications and workflows. At the moment, this is generally limited to the use of AI agents within individual enterprise apps to improve the efficiency of everyday tasks.
As time goes on, however, we’ll see more complex examples with multiple agents working together across different business functions, applications, and systems.
Agentic AI has potential applications across almost the entire enterprise, such as finance, HR, IT, procurement, and supply chain management. Let’s take a look at some specific use cases.
Customer Service
Today’s chatbots and automated voice services can only answer a limited set of questions. Agentic AI, on the other hand, can switch between systems to find out what’s happening, communicate information, and troubleshoot problems.
Imagine that a customer gets in touch because their delivery hasn’t arrived. Agentic AI first checks that the order number and customer details are correct and verifies that the package has been sent. It then connects to the delivery company’s systems via an API to check the delivery status and establish how long the delay is. If the delay is too long, it arranges a replacement delivery and sends an update to the customer.
All of this takes place without human intervention unless it’s specifically needed, in which case the AI agent will flag the problem for an employee to review.
Software Development
GenAI is already being used to augment software development. It’s comparatively limited in scope, though, and requires human oversight to make sure there are no coding errors or problems with best practice, compliance, or integration.
Agentic AI, however, has the potential even for non-technical users to define and achieve outcomes like “build a database that combines customer account information with the details of their orders across different sub-brands.”
Agentic AI would then break this down into subtasks, create, debug, test, and deploy the software, and suggest broader improvements to supporting infrastructure and platforms.
Process and Workflow Automation
Automating workflows and processes in areas like research, finance, HR, and administration is a major focus for agentic AI. Around two-thirds of organizations already using AI agents report seeing increased productivity, and 57% say they’re making cost savings.
Agentic AI, for example, could take the notes from a meeting and identify that a business trip to a conference needs to be booked. It would then check routes, flights, and hotel prices near the venue, book them all, and create a schedule for the people participating. Each traveler would receive relevant reminders and updates on travel disruption, and costs would be allocated to the correct budgets within the organization’s financial systems.
Agentic AI Holds Great Promise – But There Are Potential Pitfalls
By autonomously finding the best ways to achieve specific outcomes, agentic AI has the potential to make fundamental changes to the way businesses operate – but it could also amplify risk and create operational and governance issues.
In our next blog on this topic, we’ll look at how the rise of Managed Intelligence Providers could help enterprises to avoid the hazards and make sure agentic AI deployments deliver what they promise.