Table of contents
Open Table of contents
What the Agent Does
The agent is a conversational AI advisor trained on everything we’ve published about cloud cost optimization, AI economics, and FinOps. You ask a question. It searches our knowledge base, synthesizes a direct answer, and when relevant, points you to the specific service that fits.
It has five capabilities:
- Search knowledge — queries our published analysis on cloud costs, AI strategy, GreenOps, and FinOps practices
- Look up FinOps framework — references the FinOps Foundation’s official capabilities, playbooks, KPIs, FOCUS spec, and maturity criteria
- Get service details — pulls the full scope and engagement model for any of our five services
- Recommend services — matches your specific situation to the right combination of services
- Book a consultation — captures your details so we can follow up directly
This isn’t a chatbot reading a FAQ page. It reasons across our entire body of work to give you the same kind of opinionated, specific guidance you’d get from a senior consultant — just faster. We wrote about why conversational interfaces are transforming FinOps when AWS launched their MCP Server. This agent is us applying that same principle to our own advisory practice.
Powered by Our Open-Source FinOps Agent Skill
The agent’s expertise doesn’t live in a black box. It’s built on our Cloud FinOps Agent Skill — an open-source skill for Claude Code and compatible agent frameworks that packages our advisory methodology into reusable agent tooling.
The skill includes 25 reference files covering FinOps methodology, cloud provider specifics (AWS, Azure, GCP, OCI), AI provider economics (Anthropic, Bedrock, Azure OpenAI, Vertex AI), data platform cost patterns (Databricks, Snowflake), and operational playbooks like tagging governance and GreenOps remediation.
It’s the same knowledge base that powers the web agent, but anyone can install it into their own development environment. A FinOps practitioner running Claude Code gets the same depth of analysis as someone chatting on our site. The methodology follows a six-step reasoning chain: intake, methodology selection, maturity assessment, routing, diagnosis, and structured output.
We open-sourced it because advisory expertise that sits behind a paywall only helps the organizations that can afford it. Packaging that expertise as an agent skill means it scales to anyone who needs it. That’s the model we believe in — and the one we’d advise our clients to follow. If efficiency unlocks demand rather than reducing it, then making expertise more accessible creates more opportunities, not fewer.
Why We Built It
Gartner projects that 40% of enterprise applications will embed task-specific AI agents by 2026 — up from less than 5% in 2025. That’s not a gentle curve. It’s a step change in how organizations deliver expertise.
We advise enterprises on cloud and AI strategy. It would be contradictory to tell clients they should be leveraging AI agents while not building one ourselves. So we did.
The practical motivation is simpler: most cloud cost questions have known answers. The right-sizing patterns, the commitment discount strategies, the tagging frameworks — these aren’t secrets. The value a consultant provides isn’t the knowledge itself. It’s knowing which knowledge applies to your specific situation and how to sequence the work.
An AI agent can handle the first part instantly. It can assess whether your challenge is a visibility problem, an architecture problem, or a governance problem. It can point you to the right framework. And when the situation is complex enough to need human judgment — a multi-cloud migration, a major AI deployment, an organizational FinOps transformation — it can connect you to us directly.
That division of labor means we spend our time on the problems that actually require senior advisory expertise, and technology leaders get useful guidance the moment they need it.
How It Works
The agent runs on Cloudflare Workers with Durable Objects for session management. The knowledge base is built from our blog content and service documentation, chunked and embedded for semantic search. When you ask a question, the agent embeds your query, finds the most relevant content via cosine similarity, and generates a response grounded in that context.
It also carries the full FinOps Foundation framework — all capabilities, maturity criteria, and implementation playbooks — as structured reference material. When you ask about cost allocation maturity or FOCUS spec adoption, it’s pulling from the authoritative source, not improvising.
The architecture is deliberately lightweight. No database writes for message persistence. No heavyweight infrastructure. The cost per conversation runs around three cents. We made the same model selection trade-offs we advise clients to make: choose the model tier that delivers the required quality at the lowest viable cost, and don’t over-provision.
What You Can Ask
The agent handles the range of questions technology leaders typically bring to an advisory engagement:
Cloud cost pressure — “Our AWS bill is $80K/month and growing 15% quarter over quarter. Where do we start?” The agent will assess whether you need visibility tooling, commitment optimization, architectural changes, or all three.
AI cost planning — “We’re adding AI features to our product. How do we budget for inference costs?” It draws on our analysis of hidden AI costs and inference economics to give you a realistic picture.
Post-migration overruns — “Cloud bill doubled after our Kubernetes migration. What went wrong?” It identifies the common architectural anti-patterns that drive post-migration cost spikes.
FinOps maturity — “We have dashboards but nobody looks at them. How do we move to the next level?” It assesses where you are on the maturity curve and what specific capabilities to build next.
Engagement scoping — “What does a FinOps engagement with Suan actually look like?” It pulls the full scope, deliverables, and engagement model for the relevant service.
What It Won’t Do
Transparency matters more than overselling.
The agent won’t access your actual cloud environment. It can’t pull your real spend data, analyze your specific resources, or run optimizations on your behalf. That requires a proper engagement with the right access and governance.
It won’t promise specific savings numbers. Every organization’s cost profile is different, and quoting percentages without seeing the data is how consultants lose credibility.
It won’t replace a full advisory engagement for complex problems. Multi-cloud architectures, organizational FinOps transformations, large-scale AI deployments — these require the kind of sustained, contextual judgment that no agent can provide today.
What it will do is give you a clear, informed starting point. The right frameworks, the right questions to ask internally, and a direct path to deeper expertise when you need it.
Try It
The agent is live at suan.digital/agent. No login, no forms, no waiting — just a quick verification and you’re in.
Ask it whatever you’d ask a cloud cost consultant on a first call. If it can help, it will. If your situation needs more than an AI agent can offer, it’ll tell you that too.
We built this because we believe expert guidance should be available when the question arises — not two weeks later on a scheduled call. Give it a try and let us know what you think.
Sources: Gartner AI Agent Forecast 2026, EY Technology Pulse Poll