2024Life insurer
AI coaching system demo
We built a working AI coaching demo in about three weeks, so the client could decide whether to continue based on results.
- Prototype stage
- Rapid validation
- Insurance
Forward deployed team AI implementation team (FDT) Not just advice.
A whole team takes your AI to production.
A team of a PM, designer, front-end, back-end and AI engineers works inside your workflow to find the use cases worth building with AI. We start with a PoC or MVP and you accept it on results; once it proves useful, the same team takes it to production and can maintain it long term.
Free requirements review · Reply within 1 business day · NDA available
JoinX experience
- projects delivered
- 520+
- clients served
- 220+
- years of development
- 10+
- legacy systems taken over
- 100+
What is a forward deployed team (FDT), and how is it different from an AI consultant or an FDE?
A forward deployed team (FDT) places a whole team inside a company's workflow, and the same people own everything from finding the use case and building a prototype to launching in production. AI consultants usually deliver recommendations and plans, and you hire someone else to build. A forward deployed engineer (FDE) is an engineer who learns your process on site while building, but design, project management and maintenance are usually left to you. An FDT puts all these roles in one team, so nothing gets lost between advice and delivery, and someone stays to maintain the system after launch.
Real projects AI projects we have taken from prototype to production
Build it, accept it on results, then decide what is next.
These are AI projects we actually developed. Some started with a prototype built in a few weeks; others ran in stages over several years.
2022–2025National Atomic Research Institute (NARI)
AI eddy current signal analysis and automation
Developed in stages, from single-channel to multi-channel, with AI recognition and automation built as two separate systems to reduce risk; the work continued through 2025.
- Staged validation
- AI analysis
- Long-term partner
Linking Digital AiCast
AI text analysis and character voice acting
We set acceptance thresholds first, 90% character recognition and 75% text segmentation, then automated the whole audiobook production flow. The platform won an iF Design Award.
- Acceptance thresholds
- Generative AI
- In production
2026Insurance company
AI agent interface design and build
We designed the AI agent's conversational interface and workflows around how frontline staff work; the AI looks things up and organizes data, and people make the final call.
- AI agent
- Interface design
- Finance
2026GoodParty
AI search added to a platform we maintain
We built and still maintain the platform, and added AI search directly to it, reusing the existing data and back office.
- Extending existing systems
- AI search
- Long-term maintenance
Since 2024Health tech startup
Taking over an AI health app
We took over maintenance of an AI health app already in production and built a new back-end API, supporting the product into its next version.
- Takeover
- Back-end API
- Health tech
* Projects without a client name are described by industry under confidentiality agreements.
What the team does What the team does
One team, from finding the use case to life after launch.
We deploy a team: a PM, UI/UX designer, front-end and back-end engineers and an AI engineer, plus QA and operations staff as the project needs.
Find the use cases worth building
We interview frontline staff and look at real data and workflows to pick scenarios where data is available, people can review the results, and mistakes are recoverable.
Build a PoC or MVP together
We agree on acceptance thresholds first, then build a working prototype with real data; if it misses the threshold, we adjust or stop rather than push on.
Launch in production
We add permissions, integrations, monitoring and testing to turn the prototype into a system people use every day, with the same team and no handover.
Connect to your existing systems
The AI reads ERP, CRM, order or document data through APIs and reuses your existing account permissions, without rebuilding everything.
Training and handover
Training is part of the service: how to use the system, how to judge AI results and how to update the knowledge base, with no separate course fees.
Maintain, or take over systems already live
We keep tuning and maintaining after launch; AI or other systems built by someone else can be taken over after we review the code and architecture.
Which one Which one
AI consultant, FDE, FDT or outsourcing: what is the difference?
The difference is who comes in, what they deliver, and who is responsible afterwards.
| Approach | Who comes in | What is delivered | Watch out for |
|---|---|---|---|
| AI consultant | Who comes in: Consultants | What is delivered: Assessment reports, implementation plans, training courses | Watch out for: You need another team to build it, and plans often get lost before delivery |
| FDE (forward deployed engineer) | Who comes in: One or a few engineers | What is delivered: Features built on site while learning your process | Watch out for: Design, project management and maintenance are usually left to you |
| FDT (forward deployed team) | Who comes in: PM, designer, front-end, back-end and AI engineers | What is delivered: A PoC or MVP, then production launch and maintenance | Watch out for: You need an internal contact involved so the right use case is chosen |
| Standard outsourcing | Who comes in: A team that builds to spec | What is delivered: A system built to the spec | Watch out for: You must define the spec yourself; AI projects often change once you see results |
Process How we work
Two paid stages, each accepted on results.
Prove the AI is useful in the prototype stage before committing the production budget.
- STEP 01
Free requirements review
1–2 meetings
We hear which workflow you want to fix and what data you have, judge whether AI fits, and suggest where to start.
- STEP 02
Prototype stage (PoC/MVP)
4–12 weeks
The team maps your workflow and data, builds a PoC or MVP, and you accept it against thresholds agreed up front.
- STEP 03
Delivery stage (production)
2–4 months
We add permissions, integrations, monitoring and testing, launch, and complete training and handover.
- STEP 04
Maintain and expand
Ongoing after launch
We tune based on usage, track cost and accuracy, and expand to the next use case when you are ready.
Pricing Price ranges
How much does an AI implementation team cost?
Based on custom projects in Taiwan in 2026, priced by stage. Actual cost depends on your data, the systems to connect and acceptance requirements.
| Stage | Price range | Includes | Timeline |
|---|---|---|---|
| Requirements review | Free | Understand your workflow and data, judge whether AI fits, suggest the first use case | 1–2 meetings |
| Prototype stage (PoC/MVP) | NT$300K–1.2M | PoC about NT$300K–800K, MVP about NT$500K–1.2M, accepted on results, training included | 4–12 weeks |
| Delivery stage (production) | NT$800K–2.5M | Permissions, integrations, monitoring, testing and deployment, with handover and training | 2–4 months |
| Cross-department, multi-system | NT$2.5M+ | Multiple use cases or AI agents, cross-system workflows and full governance | 4+ months |
AI model API and cloud hosting costs are billed by actual usage; maintenance after launch, and taking over systems already in production, are contracted separately.
FAQ FAQ
AI implementation team FAQ
How is an AI implementation team different from an AI consultant?
AI consultants usually deliver assessment reports and plans, and you hire another team to build. Our team goes straight from assessment to a PoC or MVP, you accept it on results, and the same people take it to production, so nothing gets lost between advice and delivery.
What is the difference between an FDT and an FDE?
A forward deployed engineer (FDE) is an engineer placed inside your company who learns the process while building. A forward deployed team (FDT) sends a whole team: besides engineers, a PM, a designer and AI engineers, so finding the use case, interface design, launch and maintenance all happen within one team.
How is it priced?
By stage. The requirements review is free; the prototype stage costs about NT$300K–800K for a PoC or NT$500K–1.2M for an MVP; the delivery stage (production) about NT$800K–2.5M; cross-department, multi-system work usually starts at NT$2.5M. AI model API and cloud costs are billed by usage, and maintenance is contracted separately.
How is the prototype stage accepted? What if it does not work?
Before we start, we agree on acceptance thresholds such as accuracy, processing time or whether users can complete a task, then test with real data. If it misses the threshold, we decide together whether to change the approach, switch use cases or stop, rather than forcing it into production.
Is training included, or does it cost extra?
Training is included in every stage, with no separate course fees. In the prototype stage we show your team how to use the prototype and judge AI results; at launch we hand over operations and knowledge base updates.
Can you take over a system or AI built by someone else?
Yes. We first review the code, architecture and data flows, explain the current state and risks, then discuss maintenance or expansion. We also maintain systems we built over the long term, such as the GoodParty platform and a health tech startup's AI health app.
Will our data be used to train AI?
No. We use enterprise AI services whose contracts prohibit training on your data; passwords and personal data are never put into AI tools, and data access is set by role.
Get started Start with one workflow
Bring a workflow you want AI to fix, and we will tell you for free whether it fits.
No spec needed. Tell us which department and which workflow takes the most effort, and where the related data lives today.
- 1A sales manager replies within 1 business day
- 2Sign an NDA first, then discuss your workflow and data
- 3We discuss workflow and scope first, and quote once the spec is clear