The future of FinOps isn’t more dashboards. It’s a loop – where AI acts, the human governs, and the estate never stops improving.
Cloud waste is not a new problem. Organisations have been losing money to idle instances, orphaned storage, and over-provisioned databases since the first AWS bill landed on a CFO’s desk. What is new – and what Opactiv HALO solves for the first time – is that you can close the loop between identifying waste and eliminating it, continuously, without routing every decision through a human approval queue.
HALO stands for Hybrid AI Loop Optimisation. It is the engine inside the Opactiv platform, and it works on a different premise to every FinOps tool that came before it. The industry has been stuck on the same model, despite the FinOps Foundation’s State of FinOps 2025 listing more than 100 tools claiming to solve the problem.
The Problem With Recommendations
Every major FinOps platform – every percentage-of-spend platform, runs on the same model: ingest billing data, surface recommendations, wait for a human to act. It is an open loop. The tool identifies that an EC2 instance has been running at 4% CPU utilisation for six months. It surfaces a rightsizing recommendation. Then it waits.
According to the State of FinOps 2024 and 2025 reports, workload optimisation and waste reduction are the single highest priority for FinOps teams. The reason it remains the highest priority is that teams are not making progress. Identifying waste and eliminating waste are two completely different disciplines, and most platforms only do the first.
ProsperOps and nOps close this loop for a narrow slice of the problem, AWS Reserved Instances and Savings Plans. CAST AI closes it for Kubernetes workloads. But neither covers the full estate, and both charge a percentage of savings delivered, which creates a perverse incentive: the more you save, the more you pay, and disputes about what constitutes a ‘delivered saving’ are a routine part of the relationship.
What HALO Does Differently
Hybrid AI Loop Optimisation is built on a simple premise: the AI should act, and the human should govern. Not the other way around.
In the HALO model, Opactiv’s AI continuously monitors every managed resource across the estate – compute, storage, databases, SaaS, pipelines, and network. It classifies each resource as Unoptimised (actionable waste present, autonomous management active) or Optimised (healthy, platform monitoring). The classification updates as conditions change – a resource that was Optimised six months ago can become Unoptimised as usage patterns shift, and HALO will respond without a ticket being raised or a sprint being planned.
The ‘hybrid’ in HALO is not a compromise, it is a deliberate design decision. Some decisions should not be autonomous. Decommissioning a production database. Changing a commitment strategy ahead of a major product launch. Responding to a compliance freeze. For these, HALO surfaces the decision to a person with full context: what the AI has observed, what it recommends, what the risk profile looks like, and what will happen if no action is taken. The person makes the call. The AI executes it.
Where the line sits between ‘HALO acts’ and ‘a person decides’ is not fixed by Opactiv. It is set by the customer in their governance parameters – and it moves over time. A new customer might start with elimination of clearly dead resources only and require human approval for anything else. Six months in, with the platform’s record in front of them, that same customer might delegate full rightsizing autonomy to HALO across non-production. Trust compounds. The line moves. As much AI as the customer is comfortable with – no more, no less.
This is the loop.
OBSERVE > CLASSIFY > ACT > MONITOR
It never stops. And because it never stops, savings compound rather than plateau. A resource rightsized in January creates headroom that allows better commitment planning in March. Better commitment coverage in March reduces waste flags in June. The estate improves continuously, and the platform’s bill, a flat percentage of cloud spend, falls as the estate gets cleaner.
The Human in the Loop
The ‘human in the loop’ concept has been widely discussed in AI circles, but rarely applied well in enterprise software. Most implementations use it as a euphemism for ‘the AI gives you a list and you do the work.’ HALO inverts this: the AI does the work, and the human sets the boundaries within which that work can happen.
On onboarding, every customer defines their governance parameters. Which resource types can be autonomously rightsized? What is the acceptable risk tolerance for commitment purchasing? Are there accounts, environments, or workloads that must remain human-approved regardless of the savings opportunity? These parameters are set once and can be adjusted at any time through the Opactiv dashboard. HALO operates within them absolutely – it will never cross a line the customer has drawn.
Opactiv operates like a competent operations team that never sleeps, never gets distracted, and never misses a savings opportunity because it was heads-down on something else. The CFO sees the bill going down. The engineering team is not interrupted for routine optimisations. The FinOps practitioner, where the role exists, shifts from firefighting to strategy.
Why This Matters for FinOps Maturity
The FinOps Foundation’s maturity model describes three stages: Crawl, Walk, and Run. Most organisations have been stuck at Crawl, not for lack of ambition, but because every step up the maturity ladder requires more human effort, more process, and more organisational change. FinOps has historically scaled linearly: more cloud spend means more FinOps headcount, more tooling, more governance overhead.
HALO breaks this relationship. Because the optimisation loop runs autonomously, maturity scales with the platform rather than with headcount. An organisation with one part-time FinOps practitioner and Opactiv can achieve run-stage outcomes – continuous waste elimination, active commitment management, cross-platform governance – that previously required a dedicated team of four or five people and years of institutional knowledge-building.
This is not a marginal improvement. It is a structural shift in how FinOps value is delivered.
The conventional path to FinOps maturity is cultural. Build the team. Train the engineers. Establish the rituals. Run weekly waste reviews. Wait two years for the savings to compound. The advice is not wrong, but it assumes the bottleneck is decision-making, when the real bottleneck is the work that follows the decision.
A recommendation surfaced on Tuesday is not a saving. It is a ticket.
Industry estimates of cloud waste sit at around 28% of spend on average. That is the gap between what an organisation pays the hyperscaler and what the workload actually needs.
HALO closes that gap autonomously, within the boundaries the customer has drawn, from the moment the platform is connected. The savings start on day one of the trial, not month four of an implementation project. The culture follows the results.
The Channel Dimension
One aspect of HALO is rarely discussed: what it means for the channel. Opactiv pays its channel partners 50% of all gross revenue, a number that has no precedent in the FinOps market, where 15–20% reseller margins are considered generous and most platforms sell direct.
HALO is the product that makes this model defensible. Because the optimisation loop runs autonomously, there is no ongoing professional services burden on the channel partner. They do not need to staff FinOps engineers to deliver outcomes. They connect their customer to Opactiv, complete the 30-day free trial, and the platform does the rest. The partner earns recurring revenue on a growing base, customer cloud estates typically grow 20–25% per year, without the cost of delivery growing at the same rate.
For MSPs and cloud resellers adding FinOps to their portfolio without building a practice from scratch, HALO is something new: a product with the commercial model of a software platform and the outcome delivery of a managed service.
What Comes Next
HALO is Opactiv’s first domain, but the architecture is designed to extend. The same Hybrid AI Loop – observe, classify, act within governed boundaries, applies equally to SaaS licence management, data pipeline optimisation, security posture cost reduction, and energy efficiency. As trust is established in each domain, the loop expands.
The FinOps Foundation’s 2026 framework update made this expansion official with ‘Cloud+’, FinOps now encompasses SaaS, licensing, private cloud, AI workloads, and even labour costs as part of the technology value management discipline. Opactiv built HALO for exactly this era. The loop is domain-agnostic. The governance model is consistent. The pricing, a flat percentage of spend under management, scales across every category.
The question for organisations running cloud workloads in 2026 is no longer whether they need FinOps. It is whether they want FinOps that works like it is 2019 – dashboards, recommendations, tickets, waiting, or FinOps that works like the problem deserves. A loop that closes. Every time.
