榫卯Tenon AI
Contact us

Services

Choose the service you need

Our services are built around enterprise digital and AI transformation, in five categories: digital foundations, business implementation, in-house AI environment, software engineering, and consulting and training. Each one can be commissioned on its own. If you are not sure where to start, begin with a free assessment.

Contact usSee pricing
Digital foundationsGet work and data online

Move work online

We move work that still lives in spreadsheets, paper forms and group chats into an online system.

Includes

  • Map one process and its data
  • Build it in Feishu Base or a lightweight system
  • Migrate past data and switch over once it checks out
  • Set up forms, approvals, permissions and reminders
  • Train the people who use it and hand over an operating manual
  • We handle issues for a year after delivery

Data integration and reporting

We connect your existing systems, sync data on a schedule and build one operating dashboard.

Includes

  • Define the metrics and agree how each number is calculated
  • Connect ERP, e-commerce back offices, finance, CRM and other systems through their APIs
  • Sync and validate data on a schedule, with alerts when something is off
  • Build an operating dashboard with role-based access
  • Hand over a data dictionary and interface notes
  • We handle issues for a year after delivery
Business implementationTurn business work into workflows

AI agent implementation

Engineers work with your staff to turn one repetitive job into a reliable workflow.

See how it is delivered

Includes

  • Map the job with your staff and agree on the acceptance records
  • Build the workflow and connect your existing systems
  • The owner confirms the important steps
  • Accepted against real records, one by one
  • Hand over the operating manual and training material
  • We handle issues for a year after delivery
In-house AI environmentGive staff and systems access to AI

AI agent deployment

We deploy AI agents for your staff’s everyday use.

Explore AgentHub

Includes

  • Private deployment of AgentHub or another agent platform
  • Connect Feishu and model services
  • Create agents for staff and projects, with accounts and permissions
  • Administrators can see status and conversations
  • Train administrators and staff and hand over a user guide
  • Agree on upgrades and maintenance

AI model gateway deployment

We deploy one internal model API entry point to manage keys, quotas and usage centrally.

Includes

  • Deploy the gateway and connect the model services you need
  • Give each team and application its own key, model scope and quota
  • Help existing applications move to the gateway address
  • Check calls, quota limits and usage figures before launch
  • Review each team’s requests and usage
  • Train administrators and hand over operating notes

AI infrastructure

We build the infrastructure needed to run AI in-house.

Includes

  • Model selection and hardware sizing advice
  • Deploy local models and inference services
  • Plan and schedule GPUs and other compute for several teams to share
  • Set up vector search, logging and monitoring
  • Load test against the agreed concurrency and latency targets
  • Hand over an operations manual and train the operators
Software engineeringBuild or adapt software

Open-source customization and deployment

We deploy existing open-source projects and customize them as needed.

Includes

  • Compare candidate projects and recommend one
  • Check the open-source license and confirm how it may be used
  • Deploy open-source projects such as Dify and knowledge bases
  • Modify features and connect your existing systems
  • Hand over the modified source code and usage notes
  • Agree on how upgrades are handled

Software development

We build software from scratch to your requirements, including internal platforms.

See our products

Includes

  • Websites, internal platforms, business systems and admin tools
  • Web apps and iPhone apps
  • Work through the requirements and agree on a prototype and interface design first
  • Works properly on computers and phones
  • Deploy it and hand over all source code and documentation
  • We handle issues for a year after delivery
Consulting and trainingAssess plans and train teams

AI coding adoption

We help engineering teams make AI coding tools part of daily development. We build our own website and products the same way.

See our products

Includes

  • Assess where the team stands and choose coding tools and models
  • Agree on repository conventions and write the project notes the AI reads
  • Adapt code review and CI, and set security boundaries
  • Work alongside your engineers in your real repositories
  • Train the engineers and hand over training material
  • Set up measures of impact and agree on how to review them

Technical consulting

Assess feasibility, model choice and cost before you commit to building.

Start with a free assessment

Includes

  • Assess where you stand digitally and which jobs suit AI
  • Set the order and draft a transformation roadmap
  • Choose models and tools, and estimate build and running costs
  • Review an existing plan or AI project and recommend changes
  • Deliver a written report
  • Present the conclusions to management and answer questions

Typical scenarios

From software development to AI infrastructure

Six typical scenarios, from simple to advanced. Find the one closest to where you are.

01 · Software development

Internal business platform

Turn business data scattered across spreadsheets and group chats into one internal platform.

  • Map the business objects, roles and permissions
  • Build the web app and admin tools
  • Deploy it and hand over the source code and documentation

How it’s acceptedAccepted item by item against the agreed feature list.

See this service
Internal order platform
OrdersClientsReportsPermissions
SalesWarehouseFinance
OrdersClientsStatus
SO-2041Xinghe ManufacturingTo ship
SO-2040Yuanhang LogisticsShipped
SO-2039Mingchuan ElectronicsTo invoice
Illustration

02 · AI agent deployment

Agents for everyday staff use

Give staff and projects their own agents, used directly in Feishu and managed by administrators.

DeploymentAgentHub deployed privately, with agents for staff and projects
UseIn Feishu group and direct chats
ManagementPermissions by ownership, with status and conversations visible to administrators
AcceptanceStaff complete the agreed daily tasks in Feishu
See this service
Engineering assistantFeishu

Can you find out why this endpoint returns 401?

The token is not refreshed after it expires. The problem is on line 42 of auth.ts, and the fix is ready.

Commit the fix.

Committed, ready for your review.

Illustration

03 · AI model gateway deployment

One entry point for models

When teams buy models and hold keys separately, a gateway brings access, quotas and usage into one place.

  • Connect several model services
  • Give teams and applications their own keys and quotas
  • Review each team’s requests and usage

How it’s acceptedExisting applications move to the gateway address, then calls, quota limits and usage figures are checked.

Engineering tools
Support system
Internal agents
Model gatewayOne API · Separate keys · Quotas
Model service A
Model service B
Local model

Usage this month

Engineering72%
Support45%
Data team18%
Illustration

04 · AI agent implementation

Customer follow-up

Pulls requirements, reply drafts and next steps from meeting notes, chats and email, and sends them after the owner approves.

How it’s acceptedTen real conversations are checked for complete key points and usable drafts.

See this service
  1. Read the conversation recordsMeeting notes, chats and email
  2. List the requirements3 requirements, 2 open questions
  3. Draft a replyFrom product material and pricing rules
  4. Owner approvalReply draft and next stepsPending
Illustration

05 · AI agent implementation

Multi-agent collaboration

Split a complex job across several agents working in parallel, then combine the results for the owner to approve.

Parallel workJoin and waitConditional branchesHuman approval

How it’s acceptedTwenty past contracts are checked for missed risks and usable review notes.

Purchase agreement40 pages
Clause extraction agentDone
Risk review agentDone
Quote check agent1 discrepancy found
Summary agentDone
Legal approvalAwaiting approval
Illustration

06 · AI infrastructure

Internal model platform

When data cannot leave your network, run models on your own servers and offer one inference service.

  • Choose and deploy local models and inference services
  • Schedule GPUs so several teams can share them
  • Set up vector search, logging and monitoring

How it’s acceptedLoad tested against the agreed concurrency and latency targets, and checked that data stays inside the network.

See this service

Private network

ApplicationsAgents · Internal systems
InferenceOne API · Load balancing
ModelsLocal models · Embedding models
Compute
GPU 078%
GPU 164%
GPU 281%
GPU 352%
Illustration

AI agent implementation

Turn one repetitive job into a workflow

Engineers work directly with your staff: they map the job, build it as a workflow, connect it to your systems and hand it over. The industry calls this forward deployed engineering (FDE).

WorkflowRuns in your existing systems. Steps are decoupled, so changing one does not affect the others
Operating manualCovers daily use, troubleshooting and who to contact
Training materialWe train the people who use it and leave material new staff can learn from
Acceptance recordResults of checking real records one by one, confirmed by the owner
One year of maintenanceWe handle bugs, running issues and small adjustments for a year after delivery

Delivery steps

How a workflow is delivered

It starts with agreed acceptance criteria and ends when acceptance passes. A year of maintenance follows.

  1. 01

    Map the job

    We sit down with the owner and the people who do the job, map how it works today, and agree on the goal and the records to check.

    You provide: one job, one owner, a few real records

  2. 02

    Build

    We choose models and tools, split the job into decoupled steps, and set up agents, material and permissions.

    We handle: design, build and testing

  3. 03

    Acceptance and training

    We check each agreed record, train the people who will use it, and hand over the operating manual and training material.

    The balance is due after acceptance

  4. 04

    Maintain

    For a year after delivery we handle bugs, running issues and small adjustments.

    New work is quoted at the returning-client price

The first step toward digital and AI transformation

Start with a free assessment and get a transformation roadmap. Whether to go ahead is your call. We reply within 1 business day.

Contact usEmail us