Home / Services / FinOps consulting
FinOps Waste map Reporting
A waste map, the levers you can actually pull, and reporting your finance owner can read. Start with Discovery $99, then Consulting $499, Starter $2,500, or Growth $8,000/mo. We will not invent a savings percentage before we have seen the bill.
Discovery $99 Consulting $499 Starter $2,500 Growth $8,000/mo Response within 24h
Discovery names the waste on your invoice — not a generic checklist. These are the patterns we look for first:
Always-on instances, oversized VMs, and forgotten autoscaling groups that never scale in. We map hours and owners, not a made-up savings rate.
Unattached volumes, leftover IPs, unused load balancers, and snapshots nobody can name. Orphans are the cheapest cuts because nobody is using them.
Dev, staging, and “temporary” accounts with no named operator. If nobody owns the spend, it does not get turned off.
Untagged resources cannot be allocated to a team or product. Tagging is the precondition for a report finance will trust.
Model calls, retries, and embeddings without a spend envelope. Token lines belong on the same board as compute — deepened below, and in AI model cost governor Pay vs Consulte. See autonomous agents when agents are in the design.
On-demand where a commitment would fit — or commitments that no longer match the workload. We will not recommend a term until the usage is stable enough to justify it.
FinOps work sequences the same way every time. We will not invent a savings percentage before the bill is mapped.
Related write-ups: FinOps consulting services · Kubernetes cost optimizer Pay vs Consulte · Multi-cloud cost optimizer Pay vs Consulte · AI model cost governor Pay vs Consulte.
Map the waste. Add guardrails. Pilot only when the scope is written.
Discovery $99 Consulte Starter $2,500 (ZG06) All plansToken and API bills spike for the same reason cloud bills do: no owner, no tag, and no stop rule. This landing owns that work. There is no separate /genai-cost-optimization/ or /genai-finops/ service URL. The core write-up is AI model cost governor Pay vs Consulte. Multi-team and multi-agent failure modes are in GenAI FinOps at scale.
Shared or untagged keys, chatty retry loops, a large model on a small job, unscoped retrieval, and shadow plugins. The first deliverable is a spend register — keys, workflows, who can pause — not a savings percentage.
Usage is the vendor invoice (tokens, embeddings, tool traffic). The managed layer is design, policy, monitoring, and the people who stop a runaway job. Mixing them is how teams buy a platform to fix a key that should have been rotated.
Per-workflow keys and tags, budgets and anomaly alerts, model routing, context and tool budgets, and human gates on irreversible actions. A cheaper model swap with no tags is not FinOps.
Put model and API lines on the waste register next to idle compute and orphan storage. Do not buy reserved capacity or a new AI platform to cover a workflow you have not tagged yet. If the workload is an agent, pair this with autonomous AI agents.
One tagged key is a start. Several teams and several agents on shared keys is how the invoice becomes unreadable. This landing owns that work — there is no /genai-finops/ URL. Longer write-ups: Kubernetes cost optimizer Pay vs Consulte and AI model cost governor Pay vs Consulte.
Shared keys across products. Shadow agents a team stood up without a spend owner. Retry storms on a tool that never returns. A large model on a classification job. Unscoped retrieval that pulls the whole corpus for every call. Duplicate embeddings nobody retired.
Cost per completed workflow, per ticket, or per approved action — using your baseline, not a vendor slide. Usage (tokens, embeddings, tool traffic) and the managed layer (design, policy, people who stop a runaway job) stay two lines. We will not invent a savings percentage or a “typical” token cut.
Per-workflow keys and tags. Budgets and anomaly alerts per team, not one org-wide surprise. Model routing (small model first). Context and tool budgets. Human gates on irreversible actions. A kill switch that does not take the rest of the stack down. Chargeback or showback finance will actually read.
Each agent is a workflow with an envelope, an owner, and a step log. Shared memory or a shared key across agents is a cost and a governance incident. Pair this section with autonomous AI agents when agents are in the design, and with AI consulting when the first process is still unranked.
Map keys, owners, and envelopes before a second team scales tokens.
Cloud spend diagnostic — $99 ContactWhat changed, what is still open, and who owns the next cut. No invented savings percentage.
Untagged or unowned spend stays on the open list until someone is named. Tagging is a lever, not a slide.
Starter delivers the first report for the mapped waste. Growth is the monthly loop and change control — prices on /plans/.
| Step | Price | What you get |
|---|---|---|
| Discovery | $99 | Cloud spend diagnostic: waste map, GenAI keys/owners, 30-min readout |
| Consulting | $499 | Prioritized levers, at-scale controls, build / wait / do-not-cut |
| Starter | $2,500 | First FinOps implementation for the mapped waste (compute or model lines) |
| Growth | $8,000/mo | Managed FinOps cadence, multi-team envelopes, written SLA |
Idle and orphan cuts can show on the next invoice cycle. Commitments and rightsizing take a measurement window. We will not quote a savings percentage in Discovery — the waste map comes first.
For a useful waste map, yes — read-only billing and inventory is enough to start. Discovery can still map owners and questions from invoices if console access is not ready.
This landing owns the commercial “FinOps consulting” path. Prices live on /plans/ (/pricing/ soft noindex alias).
We treat model spend as a FinOps line: owner, envelope, retries, and the same waste register as idle compute. Measurement comes before a commitment or a platform purchase. Shared or untagged keys, chatty retry loops, and a large model on a small job are the first items we name. Longer write-up: AI model cost governor Pay vs Consulte. If the workload is an agent, pair this page with autonomous AI agents and AI consulting.
No. We do not invent a savings percentage. Discovery $99 is a cloud spend diagnostic: visibility, owners, and which controls are missing. Any later number is your baseline versus closed items — including cost per completed workflow once you have a baseline.
Several teams or agents on shared keys, with no per-workflow tag, budget, or kill switch. The work is unit economics you can defend (cost per workflow, not a fake %), plus controls that survive more than one owner. Detail: GenAI FinOps at scale and Kubernetes cost optimizer Pay vs Consulte.
No. At-scale GenAI spend deepens this same FinOps landing. Do not look for a thin SKU. The commercial path stays Discovery $99 → Consulting $499 → Starter $2,500 → Growth $8,000/mo.
Each agent is a workflow with an envelope, a named owner, and a step log. Shared keys or shared memory across agents are mapped as both a cost and a governance gap. See autonomous AI agents for the HITL and kill-switch rules those agents still need.
Primary path is /discovery/ (live Stripe BRL+USD). For invoice or unclear scope, use /contact/ or kleber@ziontechgroup.com. When the waste map is clear, Starter is ZG06. We respond within 24 hours.
See also: FinOps consulting services · Visibility → Starter · Autonomous AI agents · AI consulting services · Solutions · Discovery · Plans · Pricing · GenAI FinOps at scale · Kubernetes cost optimizer Pay vs Consulte · Multi-cloud cost optimizer Pay vs Consulte · AI model cost governor Pay vs Consulte · Managed IT