Back to Details

company-research

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

KServe Company Research Skill

KServe 公司研究技能

This skill produces a comprehensive BD intelligence report on a prospect company so KServe's business development team can reach out to the right people with the right message. The output is a structured chat summary with verified sources for every data point.
Compatible with: Claude.ai · Claude Code · Cowork · OpenCode · Codex · Any AI agent platform
此技能可生成关于潜在客户公司的全面BD情报报告，以便KServe的业务开发团队能够用恰当的话术对接正确的决策人。输出内容为结构化的对话摘要，且每个数据点都附有经过验证的来源。
兼容平台： Claude.ai · Claude Code · Cowork · OpenCode · Codex · 任何AI Agent平台

About KServe

关于KServe

KServe is an AI-powered Business Process Outsourcing (BPO) company headquartered in Thane, Maharashtra, India. KServe helps businesses grow and operate more efficiently by taking over key business functions — powered by integrated AI technology that delivers faster turnaround, higher accuracy, and better outcomes than traditional BPO.
KServe是一家基于AI的业务流程外包（BPO）公司，总部位于印度马哈拉施特拉邦的塔那市。KServe通过接管关键业务职能来帮助企业更高效地发展和运营——其集成的AI技术相比传统BPO能提供更快的周转速度、更高的准确性以及更优的结果。

Services

服务内容

ServiceWhat KServe does
Lead GenerationIdentifies and sources potential customers for the client's sales pipeline
Lead QualificationEvaluates leads to determine fit, intent, and readiness to buy — so the client's sales team focuses only on high-value prospects
Customer OnboardingManages the end-to-end process of welcoming and activating new customers on behalf of the client
Staff AugmentationProvides trained, dedicated staff who work as an extension of the client's own team — without the overhead of in-house hiring
Customer ServiceHandles inbound and outbound customer interactions across voice, chat, email, and other channels
Back-Office OperationsTakes over internal processing tasks — data entry, documentation, verification, and admin workflows
CollectionManages payment follow-ups, outstanding dues, and recovery processes on behalf of the client
Market ResearchGathers competitive intelligence, customer insights, and market data to support the client's business decisions
All services can be augmented with KServe's AI technology — enabling automation, smarter routing, predictive insights, and higher throughput at lower cost.
服务KServe 服务内容
Lead Generation为客户的销售渠道识别并挖掘潜在客户
Lead Qualification评估潜在客户以确定匹配度、购买意向和转化准备度——让客户的销售团队仅聚焦于高价值潜在客户
Customer Onboarding代表客户管理新客户的全流程迎新与激活工作
Staff Augmentation提供经过培训的专属人员，作为客户内部团队的延伸——无需承担内部招聘的管理成本
Customer Service处理语音、聊天、邮件及其他渠道的 inbound 和 outbound 客户互动
Back-Office Operations接管内部处理任务——数据录入、文档处理、验证及行政工作流
Collection代表客户管理付款跟进、逾期账款及回款流程
Market Research收集竞争情报、客户洞察和市场数据，为客户的业务决策提供支持
所有服务均可通过KServe的AI技术进行增强——实现自动化、智能路由、预测性洞察，以更低成本提升处理量。

Target Industries

目标行业

BFSI · NBFC · Banking & Securities · Insurance · eCommerce · Education / EdTech · Automobile · Energy & Utilities · Healthcare · Media & Entertainment · Real Estate · Retail · Manufacturing · Tours & Travel · Hospitality · Agriculture · Immigration · Accounting · Fintech · Food & Beverages · Supply Chain Management · Logistics
BFSI · NBFC · 银行与证券 · 保险 · 电子商务 · 教育/EdTech · 汽车 · 能源与公用事业 · 医疗健康 · 媒体与娱乐 · 房地产 · 零售 · 制造业 · 旅游 · 酒店业 · 农业 · 移民 · 会计 · 金融科技 · 食品与饮料 · 供应链管理 · 物流

Platform Execution Mode

平台执行模式

Detect your execution mode before starting. Apply it consistently throughout.
ModeWhen to usePlatforms
PARALLELYou can spawn independent subagents that run simultaneouslyClaude Code (
Task
tool) · OpenCode (
Task
) · Codex agent (
spawn_agent
) · Any multi-agent platform
SEQUENTIALSingle-thread only — one step at a timeClaude.ai · Cowork · Codex chat · Any single-thread assistant
If unsure, default to SEQUENTIAL — it is always safe, just slower.
PARALLEL: Spawn Workers 2–15 all at once after user confirms. Each Worker runs its own Checker loop. Orchestrator assembles the report once all Workers complete. Note: Step 10 (KServe Fit) depends on Steps 2–9 — spawn it last.
SEQUENTIAL: Run Steps 2–15 in order. Complete each Worker → Checker loop before advancing. Assemble and present the full report after Step 15.
Tool naming across platforms:
  • Web search: 
    web_search
    , 
    WebSearch
    , 
    search
    , 
    browse
    , or equivalent
  • File write: 
    write_file
    , 
    Write
    , 
    fs.write
    , or equivalent
  • Subagents: 
    Task
    , 
    spawn_agent
    , 
    create_agent
    , or platform equivalent
开始前请检测执行模式，并在整个过程中保持一致。
模式适用场景对应平台
PARALLEL可生成独立的子Agent同时运行Claude Code（
Task
工具）· OpenCode（
Task
）· Codex agent（
spawn_agent
）· 任何多Agent平台
SEQUENTIAL仅单线程——按步骤依次执行Claude.ai · Cowork · Codex chat · 任何单线程助手
若不确定，默认使用SEQUENTIAL模式——该模式始终安全，只是速度较慢。
PARALLEL模式： 用户确认后，同时生成2-15个Worker。每个Worker运行各自的Checker循环。所有Worker完成后，由Orchestrator组装报告。注意：步骤10（KServe匹配度）依赖步骤2-9的结果——最后生成该Worker。
SEQUENTIAL模式： 按顺序执行步骤2-15。完成每个Worker→Checker循环后再推进到下一步。步骤15完成后，组装并呈现完整报告。
跨平台工具命名：
  • 网页搜索：
    web_search
    、
    WebSearch
    、
    search
    、
    browse
    或等效工具
  • 文件写入：
    write_file
    、
    Write
    、
    fs.write
    或等效工具
  • 子Agent：
    Task
    、
    spawn_agent
    、
    create_agent
    或平台等效功能

Research Flow

研究流程

Phase 1 — Verification (always first, on every platform)

第一阶段：验证（所有平台均需首先执行）

Search the web for the company. Present the user with:
  • Company name (as found)
  • Website URL
  • Registered / primary address
  • Brief one-line description
Stop and wait for the user to confirm this is the right company before proceeding. If the user provided a website or address, use it to narrow the search.
Example:
I found the following. Is this the company you mean?

**Name:** Reliance Retail Ltd.
**Website:** https://www.relianceretail.com
**Address:** 3rd Floor, Court House, Lokmanya Bal Gangadhar Tilak Marg, Mumbai – 400002
**About:** India's largest retail chain across grocery, fashion, and electronics.

Please confirm and I'll run the full research.
通过网页搜索目标公司，向用户展示以下信息：
  • 公司名称（搜索结果中的名称）
  • 网站URL
  • 注册/主要地址
  • 简短的一句话介绍
暂停并等待用户确认这是目标公司后再继续。若用户已提供网站或地址，可利用其缩小搜索范围。
示例：
I found the following. Is this the company you mean?

**Name:** Reliance Retail Ltd.
**Website:** https://www.relianceretail.com
**Address:** 3rd Floor, Court House, Lokmanya Bal Gangadhar Tilak Marg, Mumbai – 400002
**About:** India's largest retail chain across grocery, fashion, and electronics.

Please confirm and I'll run the full research.

Phase 2 — Full Research (after user confirms)

第二阶段：全面研究（用户确认后执行）

Run all 14 research steps (Steps 2–15) using the execution mode detected above. Each step follows the Worker → Checker → Orchestrator pattern:
  1. Worker gathers data for that step using available web/search tools
  2. Checker validates the output against the five criteria (see Checker Instructions)
  3. If anything fails, Checker returns specific feedback to Worker — loop repeats
  4. Once approved, output passes to the Orchestrator
  5. Orchestrator assembles the final report once all steps are complete
根据检测到的执行模式，运行全部14个研究步骤（步骤2-15）。每个步骤遵循Worker→Checker→Orchestrator模式：
  1. Worker：使用可用的网页/搜索工具收集该步骤的数据
  2. Checker：根据五项标准验证输出结果（详见检查器说明）
  3. 若有任何标准未通过，Checker向Worker返回具体反馈——重复循环
  4. 验证通过后，输出结果传递给Orchestrator
  5. Orchestrator：所有步骤完成后组装最终报告

Core Research Principles

核心研究原则

These principles apply to every step and every platform. Read them before executing any step.
Recency first. Prioritize sources from the last 12 months. If only older data is available, use it but note in report: 
⚠️ Most recent available: [FY/date]. Newer data may not yet be public.
Every data point needs a source. Never present a fact without a URL or document reference. If something cannot be sourced, write "Not publicly available" — do not guess.
MCA is ground truth for Indian companies. For these fields, always use MCA (mca.gov.in) as the primary source:
  • Incorporation date (Step 5)
  • Current directors (Step 6)
  • Registered address (Step 4)
  • Financial filings / turnover (Step 3)
Tofler, Zauba Corp, and similar aggregators pull from MCA and are acceptable secondary sources.
BD framing throughout. Every section must be written with the lens of: "How does this help KServe win this account?" — not raw data, but insight.
Graceful degradation. If a tool or data source is unavailable, note it clearly in that section and move on. Never halt the entire report because one step hit a wall.
这些原则适用于所有步骤和平台。执行任何步骤前请仔细阅读。
时效性优先。 优先使用过去12个月内的来源数据。如果仅有较旧的数据可用，可使用但需在报告中注明：
⚠️ 最新可用数据：[财年/日期]。更新的数据可能尚未公开。
每个数据点都需要来源。 不得在无URL或文档参考的情况下呈现事实。若无法找到来源，请注明“未公开”——切勿猜测。
MCA是印度公司的权威来源。 对于以下字段，始终以MCA（mca.gov.in）作为主要来源：
  • 成立日期（步骤5）
  • 当前董事（步骤6）
  • 注册地址（步骤4）
  • 财务申报/营业额（步骤3）
Tofler、Zauba Corp等类似聚合平台的数据源自MCA，可作为可接受的次要来源。
全程以BD视角撰写。 每个部分的撰写都需围绕：“这如何帮助KServe赢得该客户？”——并非原始数据，而是有价值的洞察。
优雅降级。 若工具或数据源不可用，请在该部分明确注明并继续推进。切勿因某一步骤受阻而中断整个报告。

Source Priority Reference

来源优先级参考

Use this table for every step. Each step lists which sources to try in order of preference.
StepPrimarySecondaryFallback
2 — Line of BusinessCompany website (About page)LinkedIn company pageNews articles · Industry directories
3 — TurnoverMCA filings (AOC-4 Annual Return, MGT-7 Board Report)Tofler · Zauba CorpNews articles · Annual reports
4 — Head OfficeMCA registered addressCompany websiteGoogle Maps Business listing
5 — Years in ExistenceMCA company master dataCompany website (Our Story / About)LinkedIn Founded year · Wikipedia
6 — DirectorsMCA director listingToflerCompany website (Leadership) · LinkedIn
7 — BranchesCompany websiteGoogle MapsNews · LinkedIn (employees by location)
8 — ReviewsGoogle Business · Glassdoor · AmbitionBox (see step for industry table)Trustpilot · Justdial · IndiaMartApp Store / Play Store reviews
9 — RatingSynthesized from Step 8 output——
10 — KServe FitSynthesized from Steps 2–9 output——
11 — Customer CareCompany websiteGoogle Business · JustdialApp Store / Play Store listing
12 — Social MediaDirect platform search (LinkedIn, Instagram, Facebook, X, YouTube)Social Blade (trends)Company website social links
13 — TracxnTracxn.comCrunchbase (fallback if Tracxn locked)—
14 — M&ANews (last 12 months)Tracxn · Crunchbase · MCA filingsET · Mint · Business Standard
15 — BD BriefingSynthesized from Steps 2–14 output — no new searches——
所有步骤均需使用此表。每个步骤列出了按优先级排序的尝试来源。
步骤主要来源次要来源备用来源
2 — 业务范围公司官网（关于我们页面）LinkedIn公司主页新闻文章·行业目录
3 — 营业额MCA申报文件（AOC-4年度报告、MGT-7董事会报告）Tofler·Zauba Corp新闻文章·年度报告
4 — 总部MCA注册地址公司官网Google Maps商家列表
5 — 成立年限MCA公司主数据公司官网（我们的故事/关于我们）LinkedIn成立年份·维基百科
6 — 董事MCA董事列表Tofler公司官网（领导层）·LinkedIn
7 — 分支机构公司官网Google Maps新闻·LinkedIn（按地点划分的员工）
8 — 评价Google Business·Glassdoor·AmbitionBox（详见对应行业表格）Trustpilot·Justdial·IndiaMartApp Store/Play Store评价
9 — 评分基于步骤8的输出结果综合生成——
10 — KServe匹配度基于步骤2-9的输出结果综合生成——
11 — 客户服务公司官网Google Business·JustdialApp Store/Play Store列表
12 — 社交媒体直接平台搜索（LinkedIn、Instagram、Facebook、X、YouTube）Social Blade（趋势分析）公司官网社交链接
13 — Tracxn档案Tracxn.comCrunchbase（Tracxn无法访问时的备选）—
14 — 并购活动新闻（优先过去12个月）Tracxn·Crunchbase·MCA申报文件ET·Mint·Business Standard
15 — BD情报简报基于步骤2-14的输出结果综合生成——无需新的搜索——

Research Steps (Steps 2–15)

研究步骤（步骤2-15）

Step 2 — Line of Business

步骤2 — 业务范围

Find: industry, core products/services, business model (B2B / B2C / B2G), key customer segments.
查找：行业、核心产品/服务、商业模式（B2B/B2C/B2G）、关键客户群体。

Step 3 — Turnover (₹ Crores)

步骤3 — 营业额（₹ 千万卢比）

Find annual revenue/turnover in Indian Rupees (Crores). Always include the financial year (e.g., FY2023-24).
For Indian-registered companies: use MCA annual filings → Tofler/Zauba → news. For non-Indian companies or Indian subsidiaries of foreign entities: report in original currency, convert to INR at filing-date exchange rate, and note in report: 
Revenue in [currency]; converted to INR at [rate] as of [date].
If not publicly available: write "Private company — turnover not publicly disclosed."
查找以印度卢比（千万卢比）为单位的年收入/营业额。请务必注明财年（例如：FY2023-24）。
对于印度注册公司：使用MCA年度申报文件→Tofler/Zauba→新闻。 对于非印度公司或外国实体在印度的子公司：以原货币报告，按申报日期的汇率转换为INR，并在报告中注明：
收入以[货币]计价；按[日期]汇率转换为INR。
若数据未公开：注明“私营公司——营业额未公开。”

Step 4 — Head Office Location

步骤4 — 总部位置

Find the primary registered office address. Cross-reference MCA registered address against company website — they sometimes differ. If different, report both: 
Registered (MCA): [address] | Current operations (website): [address]
查找主要注册办公地址。交叉比对MCA注册地址与公司官网地址——两者有时会不同。若不同，需同时报告：
注册地址（MCA）：[地址] | 当前运营地址（官网）：[地址]

Step 5 — Years in Existence

步骤5 — 成立年限

Find the incorporation / founding year. Calculate age from today.
查找公司成立/创立年份。计算至今的运营时长。

Step 6 — Directors

步骤6 — 董事

Pull current directors from MCA. For each: Full name · Designation (MD, Director, Independent Director, etc.) · DIN (Director Identification Number).
For BD outreach, identify for BD outreach: directors likely to be decision-makers for outsourcing (MD, COO, CFO, VP Operations).
从MCA获取当前董事信息。每位董事需包含：全名·职位（MD、董事、独立董事等）·DIN（董事识别号）。
针对BD拓展，识别可能的外包决策人（MD、COO、CFO、运营副总裁）。

Step 7 — Branches & Offices

步骤7 — 分支机构与办公地点

Find: total number of offices/branches/locations · key cities/states · any international presence.
查找：办公地点/分支机构总数·主要城市/邦·是否有国际业务。

Step 8 — Reviews & Reputation (Last 12 Months)

步骤8 — 评价与声誉（过去12个月）

Search for reviews from the last 12 months on platforms relevant to the company type:
Company typePlatforms to check
Consumer brand / eCommerceGoogle Business · Trustpilot · Flipkart · Amazon · Myntra
Employer brandAmbitionBox · Glassdoor · Indeed
B2B / GeneralGoogle Business · Justdial · IndiaMart
Finance / InsuranceGoogle Business · consumer forums
Synthesize into:
  • Top 5 Positives — recurring themes across multiple reviews (note frequency)
  • Top 5 Negatives — recurring pain points (these are BD opportunities for KServe)
  • Platforms checked — list each with review count and date range
If less than 12 months of reviews exist, include older reviews and note in report: 
⚠️ Full 12-month data unavailable; includes reviews from [date range].
根据公司类型，在相关平台搜索过去12个月内的评价：
公司类型需检查的平台
消费品牌/电子商务Google Business·Trustpilot·Flipkart·Amazon·Myntra
雇主品牌AmbitionBox·Glassdoor·Indeed
B2B/通用型Google Business·Justdial·IndiaMart
金融/保险Google Business·消费者论坛
综合整理为：
  • Top 5 正面评价——多个评价中反复出现的主题（注明出现频率）
  • Top 5 负面评价——反复出现的痛点（这些是KServe的BD机会）
  • 已检查平台——列出每个平台的评价数量和日期范围
若无法获取12个月的评价，可纳入较早的评价并在报告中注明：
⚠️ 无法获取完整12个月的数据；包含[日期范围]的评价。

Step 9 — Overall Business Rating (out of 10)

步骤9 — 整体业务评分（满分10分）

Assign a synthesized reputation score — NOT an average of star ratings. Base it on the review themes from Step 8.
Rating anchors:
ScoreLabelCriteria
9–10ExcellentFew complaints, strong positive trends, company actively responds to feedback
7–8GoodMostly positive, some recurring but minor issues
5–6FairMixed reviews, notable pain points alongside positives
3–4PoorMajority negative, serious issues (e.g., unfulfilled orders, unresolved complaints), low responsiveness
1–2Very PoorSevere, consistent failures across multiple platforms
Provide a 2–3 sentence rationale. Note: a lower score often signals more BPO opportunity for KServe.
综合给出声誉评分——并非星级评分的平均值。评分基于步骤8中的评价主题。
评分参考标准：
分数等级标准
9–10优秀投诉极少，正面趋势明显，公司积极回应用户反馈
7–8良好整体正面，存在一些反复出现但影响较小的问题
5–6一般评价好坏参半，存在显著痛点但也有正面反馈
3–4较差负面评价占多数，存在严重问题（如未履行订单、未解决的投诉），响应度低
1–2极差多个平台均出现严重、持续的问题
提供2-3句话的评分理由。注意：较低的评分通常意味着KServe有更多的BPO合作机会。

Step 10 — KServe Services Fit

步骤10 — KServe 服务匹配度

Depends on: Steps 2–9 (LoB, size, reviews, directors). In PARALLEL mode, run this step last — after all other workers complete.
Based on the full research picture, recommend 3–5 services (not all 8) with explicit fit levels:
  • ⭐ HIGH FIT — service directly addresses a visible pain point found in reviews or news
  • ✅ MEDIUM FIT — service aligns with company strategy, size, or industry norms
Omit LOW FIT services entirely — only include what is genuinely relevant.
Format each as: 
[Service] — [Fit level] — [Specific evidence from research]
Example:
Customer Service ⭐ HIGH FIT
Glassdoor reviews cite "2-hour wait times" and "unresponsive support" — a direct signal
that in-house customer ops are stretched. KServe's AI-powered CX management addresses this.

Lead Generation ✅ MEDIUM FIT
Company is expanding into 3 new cities (per recent news). Qualified outbound lead gen
could accelerate market entry without growing headcount.
依赖： 步骤2-9的结果（业务范围、规模、评价、董事）。在PARALLEL模式下，此步骤最后运行——需等其他所有Worker完成后再生成。
基于全面研究结果，推荐3-5项服务（无需全部8项）并明确匹配度：
  • ⭐ 高匹配——服务直接解决评价或新闻中发现的明确痛点
  • ✅ 中匹配——服务与公司战略、规模或行业常规需求相符
完全省略低匹配服务——仅纳入真正相关的服务。
格式示例：
Customer Service ⭐ HIGH FIT
Glassdoor reviews cite "2-hour wait times" and "unresponsive support" — a direct signal
that in-house customer ops are stretched. KServe's AI-powered CX management addresses this.

Lead Generation ✅ MEDIUM FIT
Company is expanding into 3 new cities (per recent news). Qualified outbound lead gen
could accelerate market entry without growing headcount.

Step 11 — Customer Care Number

步骤11 — 客户服务电话

Find their publicly listed customer support / helpline number.
BD insight: presence of a published number signals a formal support structure. Absence may indicate underdeveloped customer ops — a potential KServe entry point. If no number is found, note in report: "No published support number found."
查找其公开的客户支持/热线电话。
BD洞察：公开电话的存在意味着公司拥有正式的支持架构。若无公开电话，可能表明客户运营体系不完善——这是KServe的潜在切入点。若未找到电话，需在报告中注明：“未找到公开的支持电话。”

Step 12 — Social Media Followers

步骤12 — 社交媒体粉丝数

Pull current follower counts: LinkedIn · Instagram · Facebook · Twitter/X · YouTube (if applicable).
Engagement signal (check the main platform — LinkedIn for B2B):
  • Review last 5–10 posts
  • Note if engagement rate appears low (<2% likes+comments/followers) or posting frequency is sparse (<1×/month)
  • Flag low engagement in report as: 
    Low engagement — [platform]: [observation]
获取当前粉丝数：LinkedIn·Instagram·Facebook·Twitter/X·YouTube（若适用）。
参与度信号（重点检查核心平台——B2B企业重点看LinkedIn）：
  • 查看最近5-10条帖子
  • 注意参与率是否偏低（点赞+评论/粉丝数 <2%）或发帖频率稀疏（<1次/月）
  • 在报告中标记低参与度：
    低参与度——[平台]：[观察结果]

Step 13 — Tracxn Profile

步骤13 — Tracxn 档案

Search Tracxn.com for the company. Report: Tracxn Score (0–100 scale, if available) · category/sector tags · funding stage · investors · notable badges.
If company is not on Tracxn (common for traditional/non-VC companies): note in report 
Not on Tracxn — likely private/bootstrapped.
and check Crunchbase as fallback.
If Tracxn profile requires a paid subscription to view detail: note in report 
Tracxn profile exists but detail is gated.
在Tracxn.com搜索该公司。报告内容：Tracxn评分（0-100分，若有）·分类/行业标签·融资阶段·投资者·知名徽章。
若公司未在Tracxn上收录（传统/非风投公司常见）：在报告中注明
未收录于Tracxn——可能为私营/自筹资金企业。
并以Crunchbase作为备选。
若Tracxn档案需付费订阅才能查看详情：在报告中注明
Tracxn档案存在，但详情需付费查看。

Step 14 — Acquisitions & M&A Activity

步骤14 — 收购与并购活动

Search for any recent (last 12 months preferred): acquisitions · being acquired · mergers · major investment rounds · PE/VC backing changes.
BD signals:
  • Being acquired → may freeze vendor decisions (note in report)
  • Fresh funding raised → likely expanding, open to outsourcing (highlight as trigger signal for Step 15)
搜索近期（优先过去12个月）的任何活动：收购·被收购·合并·重大融资轮次·PE/VC支持变化。
BD信号：
  • 被收购→可能冻结供应商决策（需在报告中注明）
  • 获得新融资→可能正在扩张，对外包持开放态度（在步骤15中作为触发信号重点标注）

Step 15 — BD Intelligence Briefing

步骤15 — BD情报简报

Most important step. Synthesize findings from Steps 2–14 into actionable outreach intel. Do not run new web searches — use only what was gathered in prior steps.
A. Things to Know Before Reaching Out (3–5 bullet points) Current strategic focus · key decision-makers · recent challenges visible in research.
B. Conversation Starters (3–5 specific, recent hooks) Based on actual events found in research (expansion, funding, product launch, leadership hire, negative reviews). Format: "[Company] recently [event] — we've helped similar companies with [KServe service] in situations like this."
C. Trigger Signals — Why Reach Out Now (top 2–3 only) Select the most compelling from:
  • Rapid hiring (scaling pain) · Geographic expansion · New product/service launch
  • Negative reviews spiking · Funding round closed · Leadership change
D. Potential Objections & Responses (2–3 only) Based on company profile, anticipate likely pushbacks and provide a suggested KServe response for each.
最重要的步骤。 将步骤2-14的研究结果综合为可执行的拓展情报。无需进行新的网页搜索——仅使用此前收集的内容。
A. 拓展前需了解的信息（3-5个要点） 当前战略重点·关键决策人·研究中发现的近期挑战。
B. 对话切入点（3-5个具体、近期的话题） 基于研究中发现的实际事件（扩张、融资、产品发布、领导层变动、负面评价）。 格式示例："[公司]近期[事件]——我们曾在类似场景中帮助同类企业提供[KServe服务]。"
C. 触发信号——为何现在拓展（仅列出前2-3个最具说服力的信号） 从以下选项中选择：
  • 快速招聘（扩张痛点）·地域扩张·新产品/服务发布
  • 负面评价激增·完成融资轮次·领导层变动
D. 潜在异议与应对方案（仅列出2-3个） 基于公司概况，预判可能的反对意见，并为每个意见提供建议的KServe应对话术。

Output Format

输出格式

Present the final report using this template:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🏢 KSERVE BD RESEARCH REPORT
Company: [Name]
Research Date: [Date]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

✅ VERIFICATION
[Website | Address | Confirmed by user]

📋 LINE OF BUSINESS
[Summary]
Source(s): [URL] | [URL]

💰 TURNOVER
[₹ X Crores | FY XXXX-XX]
Source(s): [URL]

📍 HEAD OFFICE
[Address]
Source(s): [URL]

📅 YEARS IN EXISTENCE
[Founded XXXX | X years old]
Source(s): [URL]

👔 DIRECTORS
[Name — Designation — DIN]
[Name — Designation — DIN]
Source(s): [MCA URL]

🗺️ BRANCHES & OFFICES
[X locations | Key cities]
Source(s): [URL]

⭐ REVIEWS & REPUTATION (Last 12 months)
Top 5 Positives:
1. ...
Top 5 Negatives:
1. ...
Platforms checked: [Platform — X reviews — date range — URL]

🎯 OVERALL RATING: X/10 — [Label]
[2–3 sentence rationale]

🤝 KSERVE FIT ASSESSMENT
[Service — Fit level — Evidence]

📞 CUSTOMER CARE NUMBER
[Number or "Not published"] | Source(s): [URL]

📱 SOCIAL MEDIA FOLLOWERS
LinkedIn: X | Instagram: X | Facebook: X | Twitter/X: X | YouTube: X
Source(s): [URLs]

📊 TRACXN PROFILE
[Score / Not listed / Gated]
Source(s): [URL]

🔀 M&A & FUNDING ACTIVITY
[Summary or "No recent M&A activity found"]
Source(s): [URL]

🧠 BD INTELLIGENCE BRIEFING

Things to Know:
• ...

Conversation Starters:
• ...

Trigger Signals:
• ...

Potential Objections:
• [Objection] → [Suggested response]

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📝 DATA QUALITY
Confidence: High (MCA-verified) / Medium (aggregators + news) / Low (partial data)
Data age: [All within 12 months / Mixed — oldest source: date]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
请使用以下模板呈现最终报告：
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🏢 KSERVE BD RESEARCH REPORT
Company: [Name]
Research Date: [Date]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

✅ VERIFICATION
[Website | Address | Confirmed by user]

📋 LINE OF BUSINESS
[Summary]
Source(s): [URL] | [URL]

💰 TURNOVER
[₹ X Crores | FY XXXX-XX]
Source(s): [URL]

📍 HEAD OFFICE
[Address]
Source(s): [URL]

📅 YEARS IN EXISTENCE
[Founded XXXX | X years old]
Source(s): [URL]

👔 DIRECTORS
[Name — Designation — DIN]
[Name — Designation — DIN]
Source(s): [MCA URL]

🗺️ BRANCHES & OFFICES
[X locations | Key cities]
Source(s): [URL]

⭐ REVIEWS & REPUTATION (Last 12 months)
Top 5 Positives:
1. ...
Top 5 Negatives:
1. ...
Platforms checked: [Platform — X reviews — date range — URL]

🎯 OVERALL RATING: X/10 — [Label]
[2–3 sentence rationale]

🤝 KSERVE FIT ASSESSMENT
[Service — Fit level — Evidence]

📞 CUSTOMER CARE NUMBER
[Number or "Not published"] | Source(s): [URL]

📱 SOCIAL MEDIA FOLLOWERS
LinkedIn: X | Instagram: X | Facebook: X | Twitter/X: X | YouTube: X
Source(s): [URLs]

📊 TRACXN PROFILE
[Score / Not listed / Gated]
Source(s): [URL]

🔀 M&A & FUNDING ACTIVITY
[Summary or "No recent M&A activity found"]
Source(s): [URL]

🧠 BD INTELLIGENCE BRIEFING

Things to Know:
• ...

Conversation Starters:
• ...

Trigger Signals:
• ...

Potential Objections:
• [Objection] → [Suggested response]

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📝 DATA QUALITY
Confidence: High (MCA-verified) / Medium (aggregators + news) / Low (partial data)
Data age: [All within 12 months / Mixed — oldest source: date]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Multi-Agent Architecture

多Agent架构

PARALLEL MODE

PARALLEL模式

User confirms company (Step 1)
         │
         ▼
┌─────────────────────────────────────────┐
│    SPAWN SIMULTANEOUSLY (Steps 2–9, 11–14)   │
│  Worker-2   Worker-3   Worker-4  ...    │
│  Worker-5   Worker-6   Worker-7  ...    │
│  Worker-8   Worker-9   Worker-11 ...    │
│  Worker-12  Worker-13  Worker-14        │
│  (Step 10 spawns last — needs 2–9)      │
└─────────────────────────────────────────┘
         │ (each worker ↔ checker loop)
         ▼
┌─────────────────────────────────────────┐
│      CHECKER (per worker)               │
│  Validates: source · credibility ·      │
│  recency · accuracy · completeness      │
│  Returns to worker if any fail          │
└─────────────────────────────────────────┘
         │ (all approved outputs)
         ▼
┌─────────────────────────────────────────┐
│         ORCHESTRATOR                    │
│  Assembles all 14 approved sections     │
│  Validates completeness of report       │
│  Renders final BD Research Report       │
└─────────────────────────────────────────┘
User confirms company (Step 1)
         │
         ▼
┌─────────────────────────────────────────┐
│    SPAWN SIMULTANEOUSLY (Steps 2–9, 11–14)   │
│  Worker-2   Worker-3   Worker-4  ...    │
│  Worker-5   Worker-6   Worker-7  ...    │
│  Worker-8   Worker-9   Worker-11 ...    │
│  Worker-12  Worker-13  Worker-14        │
│  (Step 10 spawns last — needs 2–9)      │
└─────────────────────────────────────────┘
         │ (each worker ↔ checker loop)
         ▼
┌─────────────────────────────────────────┐
│      CHECKER (per worker)               │
│  Validates: source · credibility ·      │
│  recency · accuracy · completeness      │
│  Returns to worker if any fail          │
└─────────────────────────────────────────┘
         │ (all approved outputs)
         ▼
┌─────────────────────────────────────────┐
│         ORCHESTRATOR                    │
│  Assembles all 14 approved sections     │
│  Validates completeness of report       │
│  Renders final BD Research Report       │
└─────────────────────────────────────────┘

SEQUENTIAL MODE

SEQUENTIAL模式

User confirms company (Step 1)
         │
    ┌────▼────┐
    │ Step 2  │ Worker → Checker validates → approved ✓
    └────┬────┘
    ┌────▼────┐
    │ Step 3  │ Worker → Checker validates → approved ✓
    └────┬────┘
        ...
    ┌────▼─────┐
    │ Step 15  │ Worker → Checker validates → approved ✓
    └────┬─────┘
         │
         ▼
  Orchestrator assembles and presents final report
User confirms company (Step 1)
         │
    ┌────▼────┐
    │ Step 2  │ Worker → Checker validates → approved ✓
    └────┬────┘
    ┌────▼────┐
    │ Step 3  │ Worker → Checker validates → approved ✓
    └────┬────┘
        ...
    ┌────▼─────┐
    │ Step 15  │ Worker → Checker validates → approved ✓
    └────┬─────┘
         │
         ▼
  Orchestrator assembles and presents final report

Checker Instructions

检查器说明

When validating any Worker output, apply all five criteria:
  1. Source present? Every fact must have a URL or named document. No source → send back.
  2. Source credible? Prefer official sources (MCA, company website, major publications) over anonymous forums or low-quality aggregators.
  3. Recency? Is the data from the last 12 months? If older, is it noted in the report with a ⚠️?
  4. Accurate? Does the data make internal sense? (e.g., a 2-year-old company cannot have 50 years of history)
  5. Complete? Did the Worker answer everything the step requires, or are there gaps?
If any criterion fails, return to Worker with specific, actionable feedback: "The turnover figure has no source — find the MCA filing or a news article citing the exact revenue figure."
Max retries: 2. If the Worker cannot satisfy all criteria after 2 attempts, approve with this note in the report: 
⚠️ [Field]: Best available data — [brief reason data is incomplete or unavailable]
Conflicting sources: If sources disagree (e.g., MCA address differs from company website), defer to MCA and report both: 
Registered (MCA): [value] | Current (website): [value]
Only approve when all five criteria are met (or a ⚠️ note is included for genuinely unavailable data).
验证任何Worker的输出时，需应用以下五项标准：
  1. 是否有来源？ 每个事实必须有URL或指定文档参考。无来源→返回给Worker。
  2. 来源是否可信？ 优先使用官方来源（MCA、公司官网、主流出版物），而非匿名论坛或低质量聚合平台。
  3. 是否具有时效性？ 数据是否来自过去12个月？若数据较旧，是否在报告中用⚠️注明？
  4. 是否准确？ 数据在逻辑上是否合理？（例如：成立2年的公司不可能有50年的运营历史）
  5. 是否完整？ Worker是否回答了该步骤要求的所有内容，是否存在遗漏？
若任何一项标准未通过，需向Worker返回具体、可执行的反馈： “营业额数据无来源——请查找MCA申报文件或引用确切收入数据的新闻文章。”
最大重试次数：2次。 若Worker在2次尝试后仍无法满足所有标准，可批准并在报告中注明： 
⚠️ [字段]：最佳可用数据——[数据不完整或不可用的简要原因]
来源冲突： 若来源信息不一致（例如：MCA地址与公司官网地址不同），以MCA数据为准并同时报告两者： 
注册地址（MCA）：[数值] | 当前地址（官网）：[数值]
仅当所有五项标准均满足（或对确实无法获取的数据添加⚠️标注）时，才可批准输出结果。

Orchestrator Instructions

编排器说明

After all 14 Workers complete and each Checker has approved:
  1. Assemble all approved sections into the Output Format template in order
  2. Validate: no field is blank, pending, or "TBD" without a "Not publicly available" statement or a ⚠️ flag
  3. If any section is missing or incomplete, return to that step's Checker with a re-request before rendering
  4. Render the final report for presentation to the user
所有14个Worker完成且每个Checker均批准后：
  1. 按照输出格式模板的顺序组装所有已批准的部分
  2. 验证：无空白字段、待处理字段或“TBD”——未公开的数据需注明“未公开”或添加⚠️标记
  3. 若任何部分缺失或不完整，向该步骤的Checker发送重新请求后再生成最终报告
  4. 生成最终报告并呈现给用户