Client Background
A smart hardware design/R&D enterprise founded in 2016, headquartered in Beijing Zhongguancun Science Park, with 120+ R&D staff across four core departments: embedded development, industrial design, structural engineering and test validation. It has delivered 80+ smart terminal products for consumer electronics, smart home and medical device clients.
As the business expanded rapidly, the enterprise accumulated massive technical documents — design specs, circuit schematics, BOM lists, test reports, supplier evaluations, project retrospectives — scattered across at least 6 systems: enterprise drive, email attachments, departmental shared folders, Feishu docs, legacy Wiki, and numerous Excel/Word files on personal PCs.
Problems Encountered
"Every time we start a new project, we spend a lot of time re-finding materials, re-validating, and re-making the same mistakes."— the enterprise R&D director described the team core pain point this way.
The specific problems manifested in three aspects:
1. Knowledge search like finding a needle in a haystack; reinventing the wheel became the norm
At project initiation, engineers need historical designs, test data and supplier info. With documents scattered and no unified index, preliminary research takes 3-5 days per project, with literature and material collection alone consuming 15% of the project cycle. Worse, different teams often unknowingly re-validate the same technical solution — in 2025 Q1, at least 7 projects were found highly duplicative of historical work, wasting over 200 person-days.
2. Cross-department information gaps; frequent rework from design to delivery
Industrial design appearance plans only met structural engineering to find they violated process requirements; embedded development chip choices were already known to have failed 2 years ago. This info should have been caught by internal knowledge systems, but repeatedly occurred due to "cannot find" or "did not know". In H1 2025, information gaps caused 34 reworks, about 12% of total project cost.
3. Senior engineers leave with tacit knowledge; long growth cycle for new hires
At the end of 2024, a senior architect in the smart home line left, taking large volumes of unarchived experience notes, supplier communications and troubleshooting manuals. After this tacit knowledge was lost, the successor team spent nearly two months re-learning. New hires took an average of 6 months to independently handle project design.
Solution Process
The R&D director learned of Honesty AI solutions at an industry event and invited the Honesty team for an on-site survey.
Phase 1: Knowledge asset inventory and unified aggregation (1 week)
The Honesty AI team first inventoried the enterprise 6 knowledge storage systems, identifying about 12,000 valid technical documents, 3,800 design drawings and 2,200 test reports. Through a custom data pipeline, documents from drives, email, Feishu, Wiki and personal endpoints were consolidated into the enterprise private knowledge base, achieving:
– Smart dedup: automatically identify and merge duplicate or highly similar document versions, clearing about 2,400 redundant files
– Tagging: auto-generate multi-dimensional tags (product type, tech domain, project phase, key components, etc.), establishing a unified classification
– Permission governance: fine-grained access by department, project group and confidentiality level, ensuring sensitive design drawings are only viewable by authorized staff
Phase 2: AI knowledge base core capability deployment (2 weeks)
Under Honesty guidance, the enterprise deployed the core knowledge base engine based on the Nuo+AI solution, with these key capabilities:
– Semantic SearchEngineers input natural-language descriptions to precisely match historical documents — e.g. "2023 smart home voice module power optimization", the system returns related design specs, test reports and architect notes in seconds, cutting retrieval from hours to under 30 seconds
– Intelligent Q&AAn AI Q&A agent built on the knowledge base can directly answer technical questions — e.g. "Which low-power Bluetooth chips have we used? What were the test results?" — automatically summarizing historical selection records and comparison data
– Related RecommendationsWhen engineers open a design document, the system auto-recommends related supplier evaluations, similar project cases and similar troubleshooting records, forming a "knowledge graph" association network
Phase 3: Scenario-based agent customization (2 weeks)
Honesty customized 4 professional AI agents based on the actual work scenarios of the four core departments:
1. Design Review AgentAfter industrial design submits a new appearance plan, the agent auto-retrieves historical process constraint records and structural feedback, generating a "potential risk report" within 24 hours, flagging details that may violate process requirements
2. Selection Assistant AgentAfter embedded development inputs chip/module selection needs, the agent auto-aggregates comparison results, supplier quotes and past project feedback from the historical selection database, generating a recommendation list
3. Test Archiving AgentAfter the test validation team completes testing, the agent auto-archives test reports into the knowledge base, extracts key metrics and compares trends with historical tests
4. New-hire Mentor AgentProvide interactive knowledge guidance for new engineers, intelligently recommending historical project docs, design standards and FAQs based on their department and skill profile, shortening the new-hire growth cycle
Phase 4: System integration and process embedding (1 week)
Honesty integrated the knowledge base and agent capabilities via API into the existing project management tool (Jira) and collaboration platform (Feishu), achieving:
– Auto-push historical similar project knowledge packages at project initiation
– Embed agent risk-alert nodes in the design review process
– Auto-archive documents and update the knowledge index after approval
– Ask @AI assistant directly in Feishu groups, get knowledge base answers in seconds
Implementation Results
Three months after deployment, the enterprise conducted a comprehensive evaluation:
| Metric | Before | After | Improvement |
|---|---|---|---|
| Preliminary project research time | 3-5 days | 0.5-1 day | Reduced 70%+ |
| Reworks due to information gaps | 34/half-year | 5/half-year | Reduced 85% |
| Projects reinventing the wheel | 7/quarter | 1/quarter | Reduced 86% |
| New hire independent project cycle | 6 months | 2.5 months | Reduced 58% |
| Average knowledge search time | 2-4 hours | <30 seconds | Reduced 99% |
| Overall project delivery cycle | Average 18 weeks | Average 12 weeks | Reduced 33% |
CostThe enterprise planned to hire 2 dedicated knowledge management staff (annual salary ~300k RMB); with Honesty AI solution, annual AI service cost is only ~80k RMB, saving over 73% in labor costs. Plus reduced rework and improved efficiency, indirect savings of ~600k RMB/year.
Knowledge AssetsThe knowledge base accumulated 14,000+ valid documents, 120+ tag dimensions and 3,600+ knowledge relationships, forming the enterprise unique "technology DNA" — even if key staff leave, critical experience and decision paths are preserved.
Client Testimonial
> "Our biggest waste was not lack of technology, but "not knowing what we already know". Honesty AI solution connected our scattered experience — when an engineer opens any design doc, the system is like a senior mentor reminding you "this approach was tried before, here is the result", "this supplier had delivery issues 3 years ago". Tripling project delivery efficiency is not an exaggeration, it is real numbers."
>
> — R&D Director of a smart hardware design/R&D enterprise
—
小诺IT，19年IT服务经验，专注为中小企业提供IT运维、安全防护、AI定制、弱电机房建设一站式服务。服务热线：400-0525-015，邮箱：service@xn11.com
⚠️ Risk ReminderEnterprise knowledge fragmentation is not just an efficiency problem but a risk of core asset loss. When senior engineers leave and project docs scatter, your unique technical accumulation may be silently evaporating. If you face similar challenges — long research time, frequent information gaps, slow new-hire growth — call 400-0525-015 or email service@xn11.com to consult Honesty Nuo+AI solution; we customize knowledge base and agent deployment for you.