We do not sell you a tool, we embed one
AI coding tools fail in predictable places. Licenses get bought, a few people try it, the development process stays exactly the same, and six months later nobody can produce a number at the renewal meeting. Adopting Marblo is a change to how work is broken down, not a tool install, which is why assessment, training, PoC, and embedding have to sit on one ladder. This page is that whole ladder.
In-house capability, not outsourcing
We are not here to run it for you. The goal is that one of your own PMs operates the agent team, so we run the first PoC together and you run the second one alone. Staying attached would grow our revenue and not your capability.
Seven rungs, from a free call to enterprise rollout
You do not buy the whole thing at once. Start at the bottom and decide at each rung whether to continue. The assessment fee is credited in full against an adoption contract, so in practice everything up to and including the assessment is the no-risk part.
- 01
Free Fit Consultation
First-pass adoption fit · direction for a priority PoC candidate
- Starts from
- Free
- Duration
- 30 min
- 02
Marblo Diagnostic Workshop
Suitability report · 1 priority PoC candidate · estimated ROI
- Starts from
- From ₩1.5M · credited on adoption
- Duration
- Half-day + 1-week report
- 03
Enterprise AI Agent Training
Per-team application candidates · PoC priority · ops guide draft
- Starts from
- From ₩3M
- Duration
- 1–2 days
- 04
Marblo Team PoC
Single-workflow validation · workspace setup · impact report
- Starts from
- From ₩5M
- Duration
- From 1–2 weeks (by scale)
- 05
Team Adoption Package
First agent team rollout · in-house manual · training · ops setup
- Starts from
- Custom quote after diagnostic
- Duration
- 4–8 weeks
- 06
Enterprise Deployment
On-prem · SSO/audit logs · AI governance · SLA
- Starts from
- Custom quote after diagnostic
- Duration
- 8–12+ weeks
- 07
Operations Support
Usage/cost monitoring · regular reviews · improvement support
- Starts from
- From ₩1M/mo
- Duration
- Monthly
Figures shown are starting prices; the final quote is tailored after the diagnostic based on org size and requirements.
What actually happens
The ladder is what you buy. This is what we do inside your organization during that time. The full cycle runs about three to four months.
- 1-2 weeks
1. Diagnose
Analyze workflows, roles, and systems. Identify the best entry points for agents.
- 2 days
2. Workshop
Train your team on Marblo and the agentic teamwork methodology, hands-on.
- 4-6 weeks
3. PoC Together
Co-design and build the first workflow, then hand operations off to your PM.
- 3 months
4. Settlement
Monthly review & tuning sessions. Co-author governance documentation.
If the people do not change, the tool does not stay
Training is not an accessory to the ladder, it is the condition for embedding. Parallel agents change how work gets cut into tickets, so if the person doing the cutting does not know the method, all you have is an installed tool. One or two intensive days, 30% theory and 70% hands-on.
Curriculum
- 01
Understanding AI coding & agent trends
How agent coding changes the way teams work, and where the field is going.
- 02
Comparing the agent tooling landscape
Strengths, weaknesses, and use cases of Claude Code, Codex, Gemini CLI, Cursor, and Grok.
- 03
Hands-on: decomposing work into multi-agents on Marblo
Break real work into tasks and distribute them across heterogeneous models.
- 04
Hands-on: Kanban-based agent collaboration
Direct an agent fleet through a TODO→IN_PROGRESS→REVIEW→DONE flow.
- 05
Designing security, permission, logging & approval policy
Design isolated execution, permission scope, audit logs, and human approval gates.
- 06
Identifying your company's PoC candidates
Surface per-team application candidates and prioritize them.
Adjusted by role
The same course needs different depth for different roles. We shift the weighting to match who is in the room.
- Engineering
- Integrate agent coding tools (Claude Code, Codex, Cursor) into real workflows
- Product & Planning
- Decompose work into multi-agents and direct them on a Kanban board
- Digital Transformation
- Company-wide AI agent adoption strategy and internal operating model
- Marketing & Data
- Apply agents to analytics, content, and report automation
- Executives
- ROI and governance decision-making for AI agent adoption
What you leave with
A certificate is not an outcome. Training earns its cost when you walk out with documents and a workspace you can use next week.
- Per-team AI agent application candidates
- A list of work each team can automate or assist
- PoC priority list
- First PoC candidates ranked by impact and difficulty
- Draft internal operations guide
- An internal doc draft covering permission, approval, and logging policy
- Marblo hands-on workspace
- A working workspace you built during the program
If an agent writes the code, who reviews the PR
Adoption reviews stall on this question far more often than on performance. Without a trust hierarchy designed in advance, one incident stops the rollout across the company. We document these five principles against your own internal policies during the engagement.
- Privilege Separation
- Separate system access by agent role. Apply the principle of least privilege.
- Audit Logs
- Record every agent action. Trace who, when, and why for every decision.
- Rollback Paths
- Always keep a reversible path for every agent decision. Git revert · DB snapshot.
- Measurable KPIs
- Define cost, accuracy, and time-saved KPIs. Measure on a regular cadence.
- Team Training
- Agents are both tool and colleague. Train your team to collaborate with them.
That covers the rules you set inside your organization. What the app itself transmits and never transmits is on its own page — six transmission categories, where the switches live, and the full outbound host list. You can hand that link straight to your security team.
See transmission scope and the security Q&AWhat came out of it
Client projects and internal pilots are labeled separately. Blending them makes the numbers look better and collapses in the first meeting.
- Law Firm A — RAG-based First-touch Consultation Agent
ChallengeLawyers were handling 500+ monthly consultations directly. Repetitive simple questions consumed their time; complex cases got insufficient attention.
SolutionA Flask + GCP Cloud Run + OpenAI RAG chatbot deployed on the firm's internal system. Their case database was vectorized to automate first-touch responses. Our team ran PoC and settlement on Marblo to ship fast.
82%First-response automation rate-70%Lawyer consultation time8 weeks (4 PoC + 4 settlement)Duration- Medical Devices B — In-house Marblo Adoption for Ops Automation
ChallengeSales, regulatory, and customer service document work was scattered. Each new product launch required rewriting the same materials from scratch.
SolutionMarblo adopted in-house. Sales proposals, regulatory docs, and FAQs each have a dedicated drafting agent; in-house PMs review and publish from the kanban board. Our team led 6-8 weeks of adoption consulting, training, and settlement.
-75%Doc drafting time3.5×Monthly throughput6-8 weeks (in-house adoption)Duration- Fashion E-commerce C — Product Ops & CS Automation System Build
ChallengeNew product registration, detail pages, and CS responses were all manual. Seasonal launches forced overtime across the ops team.
SolutionA FastAPI + GCP + OpenAI/Claude automation pipeline built and delivered. Integrated with Cafe24/Shopify APIs; the LLM generates copy, SEO meta, and image tags. Ops only approves. Our team designed, built, and delivered in 5 weeks on Marblo.
5×New product throughput-60%Ops team hours5 weeksDuration- Fintech D — 24/7 CS Agent Build
ChallengeNight and weekend responses delayed. Hiring more CS staff was costly. Simple lookups and FAQs consumed senior agent time.
SolutionA FastAPI + Claude API finance-specialized CS chatbot. Account lookups and FAQs handled automatically; complex inquiries escalate to the internal CRM. Audit logs, rollback paths, and compliance modules built in. Our team ran PoC on Marblo; production lives on the client's in-house infrastructure.
< 2 secAverage response time-60%CS operating cost6 weeksDuration- Solo Developer (self-case) — Full-stack in Parallel on Marblo
ChallengeA solo developer handling backend, frontend, tests, and docs alone. Single-model tools (Cursor, etc.) can't run things in parallel.
SolutionA Marblo workspace running Claude Code (backend), Codex (frontend), and Grok (tests) concurrently. One sentence to the natural-language orchestrator decomposes the work and routes it to 3 agents. Each runs in an isolated worktree, and a human approves at the REVIEW gate before anything merges.
100+Weekly commits5× fasterFeature PR cycleOngoingDuration- EdTech F — In-house Marblo for Course Content Pipeline
ChallengeInstructors had to personally produce scripts, quizzes, and materials. Launching a new course took 3-4 weeks.
SolutionMarblo adopted in-house for the content pipeline, operated directly by the course team. Outline, script, quizzes, and design guidance distributed across agents; saved as a reusable flow template. Our team led 4 weeks of adoption consulting and training.
3 weeks → 5 daysProduction time per course1 → 4New courses per month4 weeks (in-house adoption)Duration
The figures were measured in those specific engagements and may not reproduce in a different industry or organization. The assessment workshop produces an estimate against your own environment.
Frequently asked
The six that come up most while a team is deciding.
- What if we don't have AI experts internally?
- That's fine. You only need one in-house PM or operations lead to start. The workshop and PoC stages build the capability together with you.
- Is Marblo adoption mandatory?
- STEP 1 (diagnostic) is tool-neutral. STEP 2+ (settlement and training) is designed around Marblo. Marblo is the heterogeneous agent orchestration platform — multi-model, kanban-integrated, MCP-native — capabilities other tools don't offer in one place.
- Can we integrate with our existing systems (ERP, CRM, etc.)?
- Yes. Through MCP (Model Context Protocol) we can connect to filesystems, databases, APIs, Git, and most systems. Custom MCP server development can be included in STEP 2 or 3.
- How are security requirements handled?
- Marblo is a local desktop app — code and data don't leave your machine. STEP 3 enterprise deployment includes on-prem options, SSO/SAML, audit logs, and SLA.
- How is pricing determined?
- The 30-minute fit consultation is free. The full diagnostic workshop (STEP 1) starts from ₩1.5M and is fully credited if you proceed to a PoC or adoption contract. STEP 2·3 are quoted separately based on org size and requirements, and we present a reasonable budget plan during the diagnostic.
- How long does adoption take?
- The full cycle: diagnostic 1-2 weeks + workshop 2 days + PoC 4-6 weeks + 3-month settlement support = approximately 3-4 months. Add 8-12 weeks for enterprise deployment.
Start with the fit call
Thirty minutes, free. The purpose is not to sell but to work out together whether this is the right stage for you. If it is not, we say so.
- Team size and how the engineering org is structured
- The AI coding tools you use today and how well they are working
- One or two workflows you want to automate
- Anything security or compliance needs cleared up front