AI Talk: Turning observability into business value in the age of AI

Views on news 

Increasingly, CIOs struggle to keep track of what AI systems are doing, who uses them and how they perform. In many cases, CIOs are discovering they have no way to monitor or measure critical factors such as model drift, latency, hallucination rates, performance degradation, shadow AI and output decay. Observability, however, is the link that can lead to higher AI adoption through building trust in probabilistic AI systems.

 

Without that trust, no autonomy – the capability bringing the highest RoI – is possible. Reactive observability is no longer fit for the purposes we want to use AI for. AI can be also leveraged for gaining visibility proactively and identifying workflows that can generate extra revenue.  

 

How observability is instrumental to business outcomes 

While previously there was a focus on customer experience, today, the most important question is how technology is driving business outcomes. Real time business observability such as near real time sales feeds can also open the door for AI deployments. AI personas can do now the monitoring that humans did previously – at scale. This way barriers to productivity and revenue increase can be removed proactively.

 

Observability extends beyond systems, platforms and IT infrastructure into monitoring the wider market context for signals. But with agentic AI, decisions must still get validated by humans before they go live. Cloud services are also taking a preventive approach, where the new metric for measuring system performance is the time between two failures to see how often they happen, as well as how they can be eliminated. Agentic AI systems can then be leveraged to prevent those failures from happening.  

 

The RoI of observability is very industry specific, depending on what business value has been attached to it. It can bring about meaningful changes in the areas of reducing signal noise, faster decision making and innovation. New Relic’s 2026 AI impact report has found that companies using gen AI could halve signal noise and thus manage to accelerate understanding and achieve 25 per cent faster resolution.

 

AI deployments also generate more innovation, creating what’s called an innovation dividend. Observability is also key to agentic AI adoption, as it can monitor hallucinations, drift, token usage or AI’s failure to pull the right document from RAG. Meanwhile, Hiscox uses its agentic AI technology in a proof-of-concept trial to spot incremental usage degradation in AI releases, which humans can’t detect. However, the challenge with probabilistic systems is that the decision it makes for the second time won’t necessary be the same as the first even under similar circumstances.

 

Some mission critical industries such as oil and gas and aviation, where millisecond gaps can be a matter of life and death, already deploy agentic AI to get better observability, as well as to fill those gaps. At this stage of agentic AI, manual deployments and roll-backs are key, which then AI agents can mimic with a human still remaining in the loop. Different stakeholders will have a variety of focuses on either the business or the technical aspect of the company’s operation (risk, governance data, uptime, financial data, improved time to repair, mean time between failures). 

 

By talking to the business, you can determine which metrics matter most – infrastructure, payload, integration metrics, impact on revenue, funnel drop-off. A multi-layered dashboard can provide a drill-down view, offering relevant metrics to every function.  

 

The panel’s advice 

• As long as observability is perceived as an IT cost, executive leadership won’t see its relevance. Rather, this should be a business conversation.  
• Observability tools aren’t just “insurance policies” against outages but also revenue generating tools.   
• Observability must extend to remediation too.  
• Find metrics that evidence how you’re impacting the business.  
• While businesses have all the data, they don’t always give it to the right person for the right purpose.  
• Use agentic AI to look at the data you already have, allowing it to reveal new insights.  

newrelic.com/resources/report/ai-impact-2026?utm_source=external_partners&utm_medium=content&utm_campaign=fy27-q1-emea-ai-talk-webinar