HomeProductUse casesPricingAboutHelp center
MVP in private beta

A sales team'sAI customer relationship graphshould feel this light

See a client, just say a few words. AI recognizes the people, companies and projects and weaves them into a backlink network; the more you log, the sharper the profile, until Youdian can predict your next move.

Built for real, complex B2B sales · No forms to fill · Your data is portable anytime
Youdian = start from a single point
// say a few words -> customer graph -> predict next step
Voice capture
Hold to talk, filed in 30s
Auto backlinks
People, companies, projects auto-linked
Customer profile
Understands your clients more over time
AI prediction
Thinks of the next step first
The Problem

Customer knowledge is scattered across everyone
It is not that no one records it — it is that it does not stick

Problem 01

CRM entry is too heavy

You are already worn out after the meeting; nobody wants to fill in a dozen fields. The realest information stays in WeChat, voice notes and your head — everywhere but the CRM.

Problem 02

Customer knowledge cannot be reused

For the same client, every rep knows something different. History, quotes, key people and files sit in separate chats and folders, and no one can reuse what another rep knows.

Problem 03

Managers cannot see the process

Weekly reports only show results. Where a deal is stuck, who the decision-maker is, what is next — you ask one by one, and it still does not line up.

How it works

From one WorkLog a day
to a customer graph that predicts your next step

Point

One WorkLog a day

Visits, calls, meetings, quotes — say what happened today, 30 seconds is enough.

Line

AI extracts and auto-backlinks

AI recognizes events, people, companies and projects and builds backlinks automatically — like Obsidian, but hands-free.

Plane

Builds into customer profiles

One entry lives at once on the company, contact and project pages, and the profile grows more complete with each day of work.

Future

AI predicts the next step

What to do next, which project is stalling, when to follow up with whom — AI thinks ahead, with the reasons attached.

Two Flywheels

Two flywheels make the system smarter the more you use it

Records accumulate and knowledge accumulates — two flywheels mesh on the same customer network and spin faster together.

Record Flywheel

WorkLog Loop
  1. 1Say today's WorkLog out loud
  2. 2AI backlinks it to people, companies and projects
  3. 3Customer profiles grow more complete
  4. 4AI predicts the next action
Act on the suggestions, and new records make the predictions sharper

Knowledge Flywheel

Knowledge Loop
  1. 1Upload PPT / PDF materials
  2. 2AI splits them into new cards and updates old ones
  3. 3Q&A cites them any time, each with a source
  4. 4Save a good answer as a new card in one click
The more you ask, the thicker it gets; the thicker it gets, the more useful it is
Core capabilities

Covers everything from capture to prediction

Every capability is a real, usable product screen — not an abstract AI concept.

Automatic backlinks, like Obsidian

Log once, update everywhere: one WorkLog shows up on the person, company and project pages, with every backlink captured — all done automatically by AI.

See example
WorkLog · automatic backlinks
Linked to 4 pages
2026-06-08 · Voice transcript

Visited United Nutrition Tech in the afternoon and met Lin Lan. Discussed the Q3 supply and price lock-in for the HN-021 purchase project. She wants to see the final stability report next week; budget approval is in progress.

AI created 4 backlinks automatically

30-second voice capture

Hold to talk; transcription, summary, extraction and linking happen in one go, even offline on a weak connection.

See example
9:41
Recording client visit
00:27
Live transcript

Visited United Nutrition Tech today and met Lin Lan. Talked about the HN-021 purchase project. They care about Q3 supply; need to send the stability report next week…

Customer profiles that grow themselves

Every record feeds the profile: objective facts are merged and de-duped, subjective judgments keep their source, history is never overwritten.

See example
Company Profile
Active client

United Nutrition Tech

Functional food ingredientsShanghai · 200–500 staffLead: Lin Lan
Objective info
Linked projects3 in progress
Total quoted¥ 2.4M
Last touch2026-06-08
Touchpoints47
Key contacts
L
Lin Lan
Head of Procurement · Decision-maker
Z
Zhou Mingzhe
R&D Lead · Influencer
C
Chen Siyuan
QC Manager · User
AI synthesis

The client is focused on Q3 supply stability and price lock-in. R&D has passed sample testing; the deal is stuck at the annual budget approval by the Head of Procurement.

Source: 3 WorkLogs · 2 meeting notes · 1 quote

A growing knowledge base

PPT/PDF are split into knowledge cards, good answers settle into new cards, and AI runs regular check-ups — the knowledge base gets richer and fresher the more you use it.

See example
Knowledge base · Knowledge Cards
Health check passed · today128 knowledge cards
Upload PPT / PDFAuto-split into cardsCited in AI Q&A Answers settle into new cards
Product doc
HN-021 spec sheet
  • Purity ≥ 99.2% · water-soluble
  • Q3 capacity 40T / month
From PDF · split into 12 new cards, 3 updated
AI-distilled
HN-021 vs competitors
  • 18% higher solubility
  • 6% higher price, better stability
Distilled from one AI Q&A
Regulation
GB 28050 labeling essentials
  • Nutrition-claim threshold reference
  • Label review checklist
From PPT · split into 6 cards

Got a question? Just ask your knowledge base

The quote from last year? Where is the deal stuck? Ask in one line, and every answer cites its source.

See example
Ask Youdian
What prices did we discuss with United Nutrition Tech last year? Where is the deal stuck now?
Youdian AI

Two quotes last year: ¥1.4M in May, ¥1.2M in November. The HN-021 deal is currently in Quoting, R&D has passed sample testing, and the blocker is the annual budget approval by Lin Lan, Head of Procurement.

Quote record · 2025-11-14
WorkLog · Zhou Mingzhe visit · 2026-05-30
Meeting notes · Quarterly review · 2026-06-02
Keep asking — answers settle into the customer knowledge base…

AI prediction and reminders

Who to follow up, when, which deal is at risk — AI reminds you proactively, each with a reason.

See example
AI predictions · Next Steps
Predictions from 312 records and profilesUpdated daily
Suggest revisiting Lin Lan this week to push budget approval
She mentioned approvals usually close by mid-June · based on 3 records
View basis
Haiyue Foods deal risks stalling
No reply for 14 days after quoting; similar past deals often stall here
Follow up
Mingyuan Health is entering a reorder window
Historical cycle is about 90 days; this week is a good time to quote
Prepare quote
Built for every role

Built for reps, managers and owners

Sales

Sales rep

Capture fast, look up fast, give time back to seeing customers.

  • Log today's visit by voice in 30 seconds
  • Look up history, quotes and key people on the fly
  • Own your originals, share activity with the team
Leader

Sales manager

Deal progress, account coverage, risk alerts — all on one screen, no chasing weekly reports.

  • Team WorkLog summaries and a deal board
  • Lint flags stuck and stalled deals automatically
  • Team activity and account coverage at a glance
Owner

Founder / owner

Customer assets stay with the company; people come and go, the knowledge stays.

  • Company-wide customer assets and relationship graph
  • Ongoing tracking of key deals and decision chains
  • Access audit and smooth offboarding handover
Security & Permissions

Sales collaboration needs sharing — and boundaries

Youdian separates WorkLog originals, AI summaries, objective facts and subjective judgments. Different roles see different layers — collaboration and personal boundaries, both intact.

CapabilitySalesLeaderOwner
Write WorkLogSelfSelfAll
View own originals
View subordinate originals
View subordinate summaries
View shared-entity activity
View team activityOwn teamCompany
Manage team structureOwn teamCompany

Field-level encryption

WorkLog originals are field-level encrypted, and source files are encrypted at rest.

Subjective-info isolation

Separates objective facts from subjective judgments, visible in layers by role.

Access-log audit

Sensitive actions are fully logged, and the owner can audit access records.

Offboarding data handling

When a member leaves, customer assets stay with the company and hand over smoothly by policy.

Roadmap

From lightweight capture to a full sales knowledge system

Phase 1
In progress · MVP

Lightweight capture

  • Voice / text capture
  • WorkLog summaries
  • Entity recognition and automatic backlinks
  • Basic customer profiles
Phase 2
Knowledge capture

Docs and Q&A

  • PPT / PDF upload
  • Auto-split into knowledge cards
  • AI Q&A with citations
  • Answers saved as new cards
Phase 3
Team collaboration

Multi-role collaboration

  • Multi-role permissions
  • Deal board
  • Admin console
  • Lint reports
Phase 4
Compounding loop

Profiles and prediction

  • Deeper company / customer profiles
  • AI predictions and action suggestions
  • Proactive reminders
  • Organizational knowledge assets
Start now

Turn customer relationships into
a compounding company asset

Today's visit is still searchable next year; one person's work becomes the whole team's. Start with a single WorkLog.

No complex setup Build with early users Portable data