How AI Agents Helps in Azure Cost Optimization

FinOps teams primarily engage with dashboards, reports, and recommendations to oversee their cloud expenditure. However, the future of FinOps promises to integrate AI and reduce the necessity for human input significantly.

Advanced AI agents will monitor cloud costs, propose solutions, and execute those solutions autonomously, enabling human professionals to concentrate on governance and strategic business decisions.

Microsoft has already begun this transformation with its innovations like Azure Copilot, the Optimize agent, and updates to the Operations Center.

AI Agents Enhancing Azure Cost Optimization

Unlike traditional automation that relies on predefined rules or prompts, AI agents function as self-sufficient intelligent entities. They possess the capacity to evaluate data, reason, and complete tasks towards specific goals with minimal human assistance.

When it comes to optimising costs on Azure, these AI agents offer significant support, going beyond merely detecting unnecessary cloud expenses. They can also:

• Investigate cost anomalies and unexpected spikes
• Generate suggestions for cost savings
• Create remediation scripts
• Execute approved actions within defined limits

Below is a concise comparison of AI agents, traditional automation, and AI assistants:

The Shift Away From Conventional Azure Cost Optimisation

Modern businesses often operate complex cloud ecosystems that transform continuously, rendering traditional cost management tools less effective. Enterprises typically manage several subscriptions, extensive resources, hybrid infrastructures, and demanding AI/ML workloads. As these resources fluctuate due to scaling applications and shifting workloads, tracking cloud expenditure becomes more challenging, changing from a fixed amount to a continuously evolving figure.

The complexity intensifies when cost data is decoupled from the systems that explain its variations. For instance, a visitor may notice a sudden spike in compute costs that could indicate:

• a successful workload expansion
• inefficient application deployment
• or an idle resource running longer than anticipated

Understanding why changes occur requires insight from performance metrics, usage patterns, configurations, and ownership context.

This leads to a second concern: while identifying potential cost savings is often straightforward, taking action on those insights is often laborious.

Though Azure Advisor and similar tools can highlight optimisation potentials, the volume of recommendations can overwhelm FinOps teams, leading to fatigue over time. Often, several opportunities are identified, yet engineering teams face bandwidth limitations to investigate and implement these suggestions.

With ownership spanning across infrastructure, engineering, finance, and application teams, even a basic recommendation can take considerable time to reach the right personnel for action.

Furthermore, investigations are typically manual. Teams frequently toggle between Cost Management interfaces, monitoring dashboards, configuration settings, and deployment records to determine the safety of proposed optimisations.

By that time, workloads may have already changed again.

Traditional cost optimisation becomes mired between:

• Continuous cloud operations
• Periodic optimisation processes

AI agents present a solution to bridge this divide by continuously monitoring costs, connecting relevant context, and progressing from insight to action.

Microsoft’s Vision for AI-Enhanced FinOps

Microsoft aims to revolutionise Azure cost management by integrating AI throughout the optimisation process, from identifying problems to strategising solutions.

Azure Copilot plays a pivotal role in this evolution, functioning as an intelligent interface that can assemble specialised agents for tasks such as optimisation, observability, deployments, and troubleshooting.

The Optimization Agent brings specific relevance to FinOps. Rather than merely advising actions like resizing an underutilised VM, this agent can:

• Investigate recommendations comprehensively
• Evaluate available options
• Create CLI or PowerShell scripts for execution

For instance, a FinOps professional could inquire, “What are my leading cost-saving opportunities?” and subsequently ask Copilot to draft a script for a particular recommendation.

This evolution also modifies the role of Azure Advisor. Traditionally, Advisor provided recommendations needing human evaluation and execution. With the integration of the Optimise experience and Copilot, recommendations can initiate an interactive investigation and remediation process.

Moreover, Microsoft’s new Operations Center broadens this concept by integrating cost and carbon optimisation into a wider operational perspective. Rather than requiring teams to treat costs as separate reporting tasks, it highlights potential savings and prioritises actions for further exploration with Copilot.

The key transformation lies in minimising the distance between insight and action rather than merely adding another chatbot to Azure.

Consider a typical scenario where Azure’s spending unexpectedly rises. Currently, a FinOps analyst might detect this variance in Cost Management, consult Advisor for suggestions, discern the relevant subscription or resource, and then engage an engineer to assess what has changed. The engineer must examine Azure Monitor, deployment history, configuration, and workload performance to determine the safety of any recommendations.

An agentic workflow aims to streamline much of this investigative process.

Here’s a quick snapshot of how the FinOps workflow is evolving:

This integration offers two significant advancements in Azure cost optimisation:

• Natural language investigation: FinOps teams can progress beyond manual navigation through dashboards and recommendations. They can ask direct questions, such as, “Why did my Azure spending spike this week?” and request Copilot to delve into essential cost data and usage statistics. The agent can identify pertinent resources, analyse trends, and clarify the causes of changes.
• Automated remediation planning: Once a cost optimisation opportunity is recognised, AI can propose actions and prepare those actions for subsequent implementation. For example, the Optimization Agent can investigate a rightsizing suggestion and generate a CLI or PowerShell script for review prior to execution. This allows engineers to focus on validating and approving recommendations rather than starting from square one.

The Five AI Agents Essential for Every Azure FinOps Team

As AI becomes integral to cloud operations, Azure FinOps is expected to evolve from a system reliant on multiple tools to a dedicated team of specialised AI agents.

Instead of depending on a single assistant, organisations will deploy various agents, each responsible for different stages of the cost optimisation lifecycle.

Here are five agents that we foresee every Azure FinOps team incorporating:

1. Cost Investigation Agent

This agent is dedicated to addressing the question, “What led to our Azure spending changes?”

It will monitor billing and usage data, identifying any abnormal spending patterns. Upon detection, it can explore potential explanations by linking cost fluctuations with resource utilization, deployments, and operational data.

Instead of just signalling a spike, a Cost Investigation Agent elucidates that a surge in compute expenses happened due to unexpected workload scaling, pinpointing the responsible resources.

This agent primarily functions in read-only mode, generating explanations, alerts, or tickets for human teams to consider.

2. Optimization Planning Agent

Once a challenge or potential opportunity is highlighted, the Optimization Planning Agent determines the necessary modifications.

It weighs options such as rightsizing resources, adjusting schedules, switching to different SKUs, or obtaining Azure Reservations and Savings Plans. The agent will assess usage trends and performance necessities before proposing a prioritised optimisation plan.

Moreover, it can create implementation scripts or infrastructure adjustments for engineers to review before any actions are taken.

3. Governance Agent

This agent will help ensure that cost optimisations align with both financial and technical constraints. It will monitor spending in relation to budgets and policies and will flag or enforce established controls as needed.

This agent will necessitate more stringent controls than a read-only investigation agent, obtaining permissions to halt or pause specific actions. Such controls should consist of:

• Scoped permissions
• Approval workflows
• Complete audit trails

4. Engineering Advisor Agent

The Engineering Advisor Agent facilitates cost optimisation early in the development lifecycle. Rather than waiting for the infrastructure to generate costly bills, it can evaluate proposed architectures, infrastructure-as-code, and resource configurations. In doing so, it can flag potential cost issues prior to deployment.

This agent will recommend suitable Azure SKUs, sizing, or architectural strategies, considering performance and reliability needs. This allows engineering teams to access cost expertise during the design and deployment phases.

This approach enables proactive identification and resolution of issues before they escalate into significant costs.

5. Executive FinOps Agent

The Executive FinOps Agent interprets Azure expenditure in terms of business insights. It can analyse aggregated costs, forecast budgets, and provide organisational allocations. You could ask this agent questions like, “Which business units are exceeding their budgets?” or “How does current spending compare to the month-end forecast?”

This agent is designed to prevent overloading executives with complex cost dashboards by offering succinct summaries of relevant trends, highlighting significant variances, and pinpointing financial decisions needing attention.

Due to its analytical role, it will operate in read-only mode.

How AI Agents Will Transform the FinOps Lifecycle

Traditional FinOps processes focus on helping teams comprehend, manage, and optimise cloud spend, often requiring considerable manual effort.

By incorporating AI agents, the entire workflow becomes streamlined. These agents handle investigations, analyses, and planning, allowing your focus to remain on governance and strategic decision-making.

To summarise, here’s how agents will reshape the FinOps lifecycle:

Turbo360: Empowering AI Agents with More Than Just Azure Cost Data

Azure Cost Management provides invaluable insights into cloud expenditures. However, cost data alone doesn’t deliver the comprehensive story. While it can indicate that spending has risen or identify an optimisation opportunity, it fails to clarify why a resource exists or how it supports critical business functions. It also lacks details on resource ownership and the impact of any proposed optimisations.

Your AI agents will require insights into the operational, governance, and business contexts of cloud resources. Otherwise, the recommendations may be technically sound but operationally hazardous.

Turbo360 addresses this gap.

By integrating agentic FinOps, Turbo360 equips AI agents with the information needed to go beyond identifying cost savings, allowing for data-driven optimisation decisions. It offers:

• A unified view: Centralises Azure costs, resources, and governance across various subscriptions and environments.
• Comprehensive operational context: Merges cost, performance, and resource data.
• Business-aware optimisations: Provides visibility into ownership, applications, and organisational context.
• Safe recommendations: Assesses optimisation risks alongside operational impacts.
• Intelligent AI actions: Furnishes AI agents with the context needed to confidently propose and execute optimisations.

Schedule a demo to witness Turbo360 in action. Discover how it seamlessly integrates Azure cost management, governance, and operational intelligence for more effective AI-driven cost optimisation.