This question usually arrives when infrastructure spend has become large enough to be uncomfortable and someone asks whether it needs an owner. The honest answer is that most companies at this size ask too early, and a few ask far too late.
What tools do well
Software is good at the parts that are continuous and mechanical.
Visibility. Consolidating providers into one total with consistent categories. Fully automatable and there is no reason to pay a person to do it.
Detection. Statistical baselines, anomaly alerts, budget thresholds. Software watches every service every day, which no person does reliably.
Attribution. Tagging, grouping, per-team allocation. Configure once, then automatic.
Forecasting. Time-series projection with confidence bands. A model does this better than a spreadsheet and does not skip a month.
Reporting. Recurring summaries to the people who need them.
That covers a large share of what people imagine a FinOps hire would spend their time on.
What a person does that tools cannot
Negotiation. Committed-use discounts, enterprise agreements, provider credits. Vendors negotiate with people, and at meaningful spend this alone can justify the role.
Allocation politics. Deciding how shared costs are split between teams, and making that stick. This is an organisational problem wearing a technical costume, and software cannot resolve it.
Sustained optimisation. Rightsizing, commitment planning, architecture review. Tools surface candidates; someone has to run the change through engineering.
Judgement about trade-offs. Whether a cost increase is worth its benefit is a business decision requiring context no tool holds.
Chargeback and showback. Building a model finance accepts, and maintaining it as the org changes.
The rough thresholds
These are heuristics, not rules, and complexity moves them more than size does.
Under about $50k a month. Tooling plus an existing owner — usually a platform engineer or senior sysadmin with a few hours a week. A FinOps hire at this level costs more than the total savings realistically available. Even an excellent hire saving 20% is saving less than they cost.
$50k to $200k a month. It depends on shape. A single cloud provider with straightforward workloads is still a tooling problem. Multi-cloud, significant committed spend, or a real chargeback requirement starts to need a person — usually part of a platform role rather than a dedicated hire.
Above $200k a month. Generally settled, and the question becomes whether it is one person or a function. Negotiation value alone tends to cover the cost.
The signals that matter more than the number
Spend level is the weakest input. These are better.
Nobody owns the number. If cost is everybody's job it is nobody's, and no tool fixes that. This is an ownership decision, not a purchase.
You are leaving commitment discounts unclaimed. If you are on-demand for predictable baseline workloads, there is money on the table that requires analysis and a negotiation.
Allocation disputes are recurring. Teams arguing about who owns shared spend is a governance problem that needs a referee.
Cost decisions are being deferred for lack of analysis. If "should we move this workload" keeps getting postponed because nobody has time to model it, that is a person-shaped gap.
The order that works
Buy visibility first, in every case. It is cheap, it is fast, and it tells you whether you have a hiring problem or a tooling one.
The common expensive mistake is hiring first. A FinOps hire arriving without consolidated data spends their first quarter building spreadsheets — doing manually what software does continuously, and burning the goodwill they need for the negotiation and allocation work that actually justifies the role.
Give them the data on day one and they start on the parts only a person can do.
The honest answer for most readers
If you are at 20–200 people and asking this question, you probably do not need the hire yet. You need one consolidated view, a daily signal, and someone whose job explicitly includes looking at it — an hour a week, not a headcount.
That is what StackSpend is built for: read-only connection to every provider, one total, anomaly detection against per-service baselines, and a daily summary in Slack. When you do eventually hire, they inherit a working system instead of building one.
FAQ
At what spend level should I hire a FinOps engineer?
As a rough guide, below $50k a month tooling and an existing owner is better value; above $200k a month a dedicated role usually pays for itself in negotiation alone. Between the two it depends on multi-cloud complexity and chargeback requirements rather than the number.
What can a FinOps hire do that software cannot?
Negotiate commitments and enterprise agreements, resolve allocation disputes between teams, drive optimisation projects through engineering, and apply business judgement to trade-offs. Software handles visibility, detection, attribution and forecasting.
Should I buy tooling before hiring for FinOps?
Almost always. A hire without consolidated data spends their first quarter building reports manually. Visibility is cheap and quick, and it also reveals whether the problem is actually a headcount one.
Can a platform engineer own cloud cost management part-time?
Yes, and at most companies under 200 people that is the right answer — provided it is explicitly in their remit and they have a daily signal rather than a dashboard they are expected to remember to check.
