HALO is not a chatbot, or a dashboard feature. It is a persistent, configurable, auditable AI workforce that extends the capability of your FinOps team into the hours, the workflows, and the complexity that no human team can fully cover on their own.
There is a version of FinOps that most organisations know well. It looks like this: a practitioner logs into the cost management platform, reviews the latest spend data, works through a list of recommendations, flags a few things for the engineering team, writes a summary for the weekly finance review, and closes the laptop.
Good work. Important work. But work that competes with every other demand on that practitioner’s time – and work that stops happening the moment the laptop closes.
HALO changes that.
HALO is Opactiv’s agentic AI layer – a system of intelligent agents that work continuously across your connected cloud, AI, and cost data, completing complex multi-step analysis and operational tasks autonomously, surfacing findings when they matter, and pausing for human judgement precisely when a decision is too consequential to make without you.
HALO agents are purposeful. Each one is built around a specific operational goal, drawing on the full breadth of data flowing through the Opactiv platform – cloud infrastructure costs across AWS, Azure, and GCP, AI API consumption, datacenter total cost of ownership, Microsoft 365 licence utilisation, and multi-entity spend hierarchies.
The agents that ship with the platform out of the box cover the use cases that matter most to FinOps teams, immediately, without any configuration required.
The Cost Monitor works continuously across all connected accounts, analysing spend patterns, identifying week-over-week anomalies, and surfacing the cost drivers that are having the greatest impact on your budget. When your cloud bill has moved unexpectedly, Cost Monitor has already found out why before you open the platform.
The Rightsizing Advisor reviews how your compute and AI resources are actually being used versus how they are provisioned. It identifies the gap, quantifies the savings available from downsizing, and presents specific, evidence-backed recommendations – not generic observations, but actionable guidance tied to the resources in your environment.
The Idle Resource Hunter goes looking for the resources that are costing money without delivering value: stopped instances that are still incurring charges, volumes sitting detached from any compute, IP addresses reserved but never assigned, data stores that have gone dormant. It finds them, documents them, and – crucially – brings them to your attention in a way that preserves your ability to make the right call on each one.
The Policy Watchdog keeps a continuous eye on your budget policies and quota rules, alerting you when pools are approaching their limits before they breach them. No more end-of-month surprises. No more reactive conversations with finance about why a cost centre went over budget.
The Security Posture Advisor broadens the lens beyond pure cost, reviewing your connected cloud accounts for the misconfigurations that create both security risk and unnecessary expense: open ports, publicly accessible storage buckets, accounts without MFA enforced. Security and cost efficiency are more connected than they might appear, and this agent treats them as the unified concern they are.
The Weekly Spend Report does exactly what every FinOps team should do every week but rarely has the capacity to do consistently – it produces a structured, comprehensive summary of spend across your entire pool hierarchy, complete with trend analysis, anomaly callouts, and prioritised recommendations. Every Monday morning, the report is there. No matter what else happened last week.
Every resource shares common attributes – owner, pool, cloud account, region, tags, first seen, last active, enabling cross-cutting analytics.
The agents that ship with HALO are a starting point, not a ceiling.
Every organisation has workflows that are specific to its structure, its cloud footprint, its reporting requirements, and its team’s way of working. HALO’s agent library is built on three tiers that give you the flexibility to work with standard agents as-is, adapt them to your needs, or build something entirely bespoke.
Global agents – the out-of-the-box set described above – are maintained by Opactiv and available to every user in every organisation immediately. They require no setup and deliver value from day one.
Organisation agents are created by your platform administrators and shared across your organisation. When your team has a specific investigation workflow, a custom reporting format, or a governance check that reflects your organisation’s particular policies and pool structure, an organisation agent captures that workflow and makes it available to everyone who needs it. Any global agent can be duplicated as a starting point, then customised with your own instructions, parameters, and configuration.
Personal agents are private to you. The bespoke analysis you run every fortnight, the specific data pull that feeds your board presentation, the investigation sequence you have refined over months of FinOps practice – personal agents turn your individual expertise into a repeatable, automatable workflow.
Launching an agent takes a few seconds. From the agent library, you select the agent you want, and a clean configuration panel appears – surfacing any parameters that are relevant to that particular agent. Date ranges, numeric thresholds, account scope, boolean switches. You set what you need to set, choose between running immediately or scheduling on a recurring basis, and confirm.
That’s it. The agent begins working.
Running immediately is exactly that – the agent starts executing within seconds and appears in your live execution view as it works through its task.
Scheduling creates a persistent recurring run: daily, weekly, or monthly, at the time you choose. The scheduled agent appears in your execution view with its next run time clearly shown, and it runs reliably on that schedule without any further action from you. The Cost Monitor can run every morning. The Weekly Spend Report can run every Monday. The Policy Watchdog checks every evening. Your FinOps practice gains a consistency and completeness that is difficult to achieve through manual effort alone.
HALO’s Agents in Motion view gives you real-time visibility into everything that is currently running, pending, or waiting for your input.
Running agents show you exactly how long they have been executing and what their current status is.
Pending agents show when they are next scheduled to run. And when an agent has paused to ask for your guidance – which we will come to in a moment – the platform makes sure you cannot miss it.
A prominent alert banner at the top of the screen tells you how many agents are currently waiting for your input. This is not a notification you might scroll past. It is a clear, persistent signal that your attention is needed before something important can proceed.
From this view, you can stop a running agent if circumstances have changed, or unschedule a recurring agent without affecting runs that are already in progress.
HALO is built on a conviction that sits at the heart of how Opactiv approaches AI: autonomy should serve human judgement, not replace it.
Agents can do extraordinary things with data – analysing patterns across thousands of resources, identifying anomalies that would take a human analyst hours to find, synthesising findings across multiple accounts and cost dimensions into coherent, actionable output. But there are moments in complex workflows where the right answer depends on organisational context, business priorities, or risk tolerance that only a human can provide.
HALO’s human-in-the-loop capability is built for exactly those moments.
When an agent is configured to pause at decision points – either because its task is consequential by nature, or because the person who built it made a deliberate choice to keep humans in the loop – it reaches a checkpoint, surfaces everything it has found and concluded up to that point, and asks you a specific question before proceeding.
Consider the Idle Resource Hunter working through a large and complex environment. It discovers 62 potentially idle resources across multiple accounts, calculates the estimated monthly savings from acting on all of them, and presents this finding clearly. Before it does anything further, it pauses. It asks you how you want to proceed – which resource types to prioritise, which accounts to include in the full analysis, which resources you have already reviewed and want to exclude. You respond with the guidance it needs. It resumes from exactly where it stopped, with your direction shaping what happens next.
The agent does not start over. It does not lose its context. It continues with the full understanding of everything it already knows, augmented by what you have just told it. This is the difference between an AI tool that asks you to do its thinking for it and an AI partner that thinks alongside you.
One of the most powerful aspects of HALO is what happens after an agent finishes.
Every completed run – whether it ended naturally, was stopped, or completed after a human-in-the-loop exchange, retains its full session history. The agent remembers every step it took, every tool it used, every intermediate finding it produced on the way to its final output.
This means that a completed agent run is not a static report. It is a living, searchable data source.
Open the conversation view for any completed run, and you are back in dialogue with an agent that has full awareness of the work it just did. Ask it to drill into a specific finding. Ask it which of the anomalies it identified are in your production accounts. Ask it to generate a Jira-ready description of the three highest-priority recommendations, formatted for your team’s ticketing workflow. Ask it to compare this week’s findings against what it found last week.
It answers with full context, because its context is intact.
This transforms HALO from an automation layer into a genuine analytical partner – one that completes the work, retains the knowledge, and remains available for the follow-up questions that always emerge once you start looking at the findings.
HALO is not a black box. Every action taken by every agent is logged with a time-stamped audit trail. The agent audit log shows every tool call, every decision point, every intermediate result, and every human interaction that took place during an agent execution. For organisations with compliance obligations, internal governance requirements, or simply a preference for understanding what their AI systems are actually doing, this transparency is not optional – it is fundamental.
Agent executions carry clear approval records where human-in-the-loop confirmations took place, documenting what the agent proposed, what the human confirmed, and when. Write operations that require human authorisation are clearly distinguished from read-only analysis. The record of what HALO did, and who approved it, is always available.
The underlying AI infrastructure that Opactiv utilises manages tool access, session security, and the control logic that governs what agents can and cannot do. This is the layer that ensures agents operate within the boundaries set for them.
FinOps is, at its best, a continuous practice – not a monthly report, not a quarterly review, but an ongoing discipline of visibility, accountability, and optimisation that keeps pace with an environment that is changing every day.
HALO makes that continuous practice achievable.
It means your Cost Monitor is running analysis while your team is in meetings. Your Policy Watchdog is checking limits while your engineers are deploying. Your Weekly Spend Report is writing itself while your FinOps practitioner is focused on the strategic conversation that needs their expertise, not the data collection that does not.
It means the insights that used to require hours of manual analysis are ready when you need them. The recommendations that used to sit in a backlog are surfaced and prioritised automatically. The questions that used to require a specialist to find the right data can be asked in plain language and answered in seconds.
And when something truly matters – when a decision needs your judgement, your organisational context, your understanding of what the business actually needs, HALO stops, tells you clearly, and waits for you.
Because the best AI does not try to replace the humans who know their business. It works alongside them, handles what it can, and makes sure the things that need human wisdom reach the humans who have it.
That is HALO.
