Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

How can firms prepare for AI’s growth in 2025 and beyond?

AI can be transformational, but we must use it in the right places and fuel it in the right way to get the results and value we need.

How can firms prepare for AI's growth in 2025 and beyond?

[insta_photos / Adobe stock]

Profile photo of Darren Anstee

Chief Technology Officer for Security · NETSCOUT

Whether it’s overcoming human resource constraints, improving services and products, or gaining greater operational security and resilience, every customer I meet asks about how artificial intelligence (AI) can be used to solve their challenges. And, if you were at the Mobile World Congress in Barcelona this year, AI was the solution to everything.

We all know that isn’t the case, but there can be massive value when AI is used in the right way to help our organizations move forward. This means using AI to help solve the highest priority challenges an organization is facing, and with the right data to get reliable outcomes. Both of these are pretty obvious but merit further discussion.

AI FOR HIGH-PRIORITY ORGANIZATIONAL CHALLENGES

So, what do I mean by using AI to solve today’s highest-priority challenges? I work in the field of cybersecurity, so I am exposed to it predominantly in that area, and in adjacent areas such as observability and risk management. In security, I see growing value coming from augmented working, where AI is used to both increase operational efficiency in security processes and accelerate responses, by automatically collating and enriching event data and lending an expert (system) hand to more junior operations staff. These types of AI assistance have been around for a while, but became more readily available in 2024 as they’ve been introduced to more platforms.

AI assistants are evolving fast, from being fairly rigid in their capability to providing natural language interfaces that allow much more freeform user interaction. These new implementations offer easier access to data and are also starting to understand and react to user intent—the goal being to further multiply the capability of security personnel in a more extensible way by helping organizations improve their overall security posture without requiring additional expensive and hard-to-find security personnel.

Here, AI is being used to help solve a real problem that numerous organizations have in cybersecurity: too many events, complex processes involving disparate systems, and not enough skilled people. This may sound extremely obvious, but focusing on AI investment like this is key to delivering demonstrable value.

I highlight this because the days of funding technologies in the hope of future value are over—we have to be able to demonstrate value as we learn, understand, and deploy new technologies, and that means taking our own operational realities into account when it comes to things like AI.

For example, in security, I see unsupervised AI being used to find new anomalous behaviors within infrastructures. These anomalies can be interesting, and we can identify things that would otherwise have been missed, but the key question we must ask is: “Are these new events actionable for our organization?”

If the new events simply sit in a huge queue which can’t be triaged due to operational resource constraints, or if the events can’t be explained (where you can’t tell if they are real without an intensive background investigation that chews through already-stretched operational resources), then the platform generating them won’t deliver demonstrable value in the short term. That might be okay, but that depends on the organization, and this should be a key consideration when we look at the investments we want to make around AI.

SELECTING THE ‘RIGHT’ DATASET

Identifying the right problems for AI isn’t the only challenge though; data is the food that AI consumes. As we all learn more about AI, we appreciate how selecting the right dataset is essential for getting a good result, in the same way that having a good diet supports optimum health. The better the quality of what we consume, the better something can be for us, and the same is true for AI. If we use aggregated, low-fidelity data to drive AI, regardless of how much of it we use, the outcome will likely not be reliable and valuable to our organization.

Again, this sounds obvious, but every organization now uses a range of different infrastructure technologies, and they have business process dependencies that are spread across those technologies. Whether the goal is automated contextual enrichment, anomaly detection, or predictive analytics, having a consistent picture of what is going on—without overly abstracting away the details or having blind spots—is key if we want to feed this data into either an AI-enabled product or our own in-house-developed AI analytics stack.

Before we invest in AI, we should make sure we have the instrumentation and telemetry we need to feed it. If we don’t, we again risk not being able to deliver real value.

AI can help us overcome many challenges within our organizations. As a tool, it can be transformational. However, it isn’t magic. We must use it in the right places, and fuel it in the right way, to get the results and value we need.