Cognizant Finds $200M in Trapped Sales Pipeline With AI Account Twins

From Scattered Signals to Qualified Pipeline With Cognizant — Ai4

Ai4

Standard CRMs miss $30 of qualified pipeline for every $100 logged, leaving enterprise revenue buried across delivery and finance records until autonomous agents synthesize the signals.

AI-assisted summary of From Scattered Signals to Qualified Pipeline With Cognizant — original source. Check the original for full context.

Source published: 2026-08-20

Detailed notes

## Background Published by Ai4 on YouTube on August 20, 2026, this presentation features Cognizant executives Sushant Warikoo and Sirish Davuluri detailing the firm's internal sales transformation. According to the description, the session explores Cognizant's 'Customer Zero' journey of deploying autonomous AI agents across delivery, finance, and legal systems to uncover hidden, unrecorded sales pipeline and turn scattered enterprise signals into self-sourced demand.

## TL;DR Cognizant leaders explain how their 1,500-person global sales force deployed autonomous AI 'account twins' to extract commercial value trapped outside traditional CRM systems. By streaming edge signals across client delivery, finance records, and support interactions, the platform correlates disparate breadcrumbs into scored opportunities and automates proposal drafting, objection hardening, and multi-twin legal and finance reviews. The system also supports cross-account agent debates, pricing simulations, and autonomous risk interventions while keeping sellers in the loop. Scaled from an initial 25-person pilot under strict multi-tenant data governance, the production system surfaced $200 million in qualified pipeline within its first five weeks.

## Atmosphere & style The presentation carries a polished, pragmatic enterprise conference tone that pairs sweeping corporate strategy with detailed live software demonstrations. Warikoo opens with an assertive, grounded posture, emphatically pitching production-grade AI value realization over academic hype with remarks like 'this is not for the faint of heart.' The mood transitions smoothly as Davuluri provides a crisp, engaging product tour laced with dry humor about stock performance, making complex multi-agent redlining and cross-account debates feel immediate and tangible. It gives viewers the impression of looking directly into the operational engine room of a massive services firm aggressively dogfooding its own frontier AI systems.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=4s) Enterprise sales reps spend upwards of 60% of their time on administrative and operational overhead instead of active client engagement. - [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) For every $100 of pipeline recorded in a standard CRM, an estimated $30 of qualified pipeline sits hidden across non-sales enterprise systems. - [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) Cognizant deployed autonomous 'account twins' to synthesize cross-functional data across delivery, IT support, finance, and customer success records. - [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) Deploying the account twin architecture surfaced $200 million in net new hidden pipeline within the first four to five weeks of production. - [05:27](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=327s) The inferencing engine correlates isolated data points across delivery, engineering notes, and budget pressures to score high-probability deal opportunities. - [08:17](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=497s) Proposals are drafted, hardened against historical client pushback, and redlined in real time by autonomous legal, finance, and functional twins. - [11:08](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=668s) Account twins collaborate peer-to-peer across accounts to debate solution architectures and reuse institutional knowledge across vertical practices. - [12:14](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=734s) Enterprises can run predictive account simulations to evaluate commercial impacts—such as price increases—based on individual client health and tech priorities. - [13:37](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=817s) Autonomous account missions monitor global delivery health to identify churn risks and suggest staffing interventions while keeping sellers in full control. - [13:37](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=817s) Executive twins automatically assemble contextual briefings for leadership meetings by synthesizing real-time operational delivery signals. - [15:33](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=933s) Successful enterprise adoption requires rigorous cross-account data isolation guardrails to prevent proprietary client information leakage. - [17:00](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=1020s) Cognizant scaled the Customer Zero deployment from a targeted 25-seller pilot to the broader global sales force to refine change management and governance.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=4s) Transforming Enterprise Sales: Hyper-Intelligence and Hyper-Productivity

Sushant Warikoo, Chief Business Officer for AI at Cognizant, opens the presentation alongside Sirish Davuluri, Head of AI Strategy & Business Design. Setting the scale, Warikoo highlights that Cognizant is a $20 billion global consulting firm with 350,000 employees. Within this massive enterprise, Warikoo leads a specialized unit comprising AI designers, industry strategists, and frontier engineers united under a single mission: driving tangible business value realization from AI and bringing it into production. The session focuses on Cognizant acting as its own "Customer Zero" by overhauling its 1,500-person sales organization with autonomous AI agents.

The core motivation behind this sales transformation rests on two pillars: hyper-intelligence and hyper-productivity. Today's sales teams miss critical demand signals because the underlying data is trapped in disconnected enterprise silos outside traditional systems. Simultaneously, sellers face severe productivity drains: enterprise sales reps typically spend upwards of 60% of their working hours bogged down by operational overhead—conducting market research, analyzing internal data, navigating red tape, and building proposals and contracts—rather than meeting with clients. The primary objective was to invert this ratio entirely, enabling sellers to focus almost exclusively on high-value client engagement while AI takes over the operational grind.

[01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) Unearthing $200M in Trapped Pipeline via Account Twins and Context Fabric

Warikoo shares a foundational benchmark discovered during Cognizant's internal deployment: for every $100 of pipeline recorded in a standard CRM, another $30 of qualified pipeline sits completely hidden across non-sales systems. To capture this unseen value, Cognizant partnered with Context Fabric to build and deploy "account twins"—autonomous AI agents that operate continuously as virtual account executives for major enterprise accounts. These account twins pull and synthesize data across sales, delivery, IT support, finance, and customer success records.

```mermaid flowchart TD accTitle: Cognizant Context Fabric and Account Twin Architecture

subgraph SENSORS["Distributed Context Capture"] E["Edge & User Device Sensors"] D["Delivery & Support Records"] F["Finance & Legal Systems"] end

subgraph ENGINE["Context Layer & Agent Fabric"] CL["Context Fabric Layer"] AT["Account Twins"] OT["Cross-Account / Functional Twins\n(Legal, Finance, SAP Practice)"] end

subgraph OUTPUT["Seller Action & Opportunity"] DASH["Qualified Pipeline Dashboard\n(Scored Opportunities)"] PROP["Autonomous Proposal Drafting\n& Multi-Twin Redlining"] end

E --> CL D --> CL F --> CL CL --> AT AT <--> OT AT --> DASH DASH --> PROP ```

The enterprise-wide roll-out relies on context sensors running on every edge device across the 1,500-person sales organization, continuously feeding signals into the context layer. In just the first four weeks of operation, the account twins surfaced $200 million in hidden pipeline that was completely invisible to conventional CRM workflows.

Beyond lead identification, these account twins interact autonomously with specialized "legal twins" and "finance twins" to eliminate the endless cross-functional pricing and contract review calls that typically bottleneck deals. Furthermore, account twins communicate peer-to-peer across accounts to share proven problem-solving patterns or proactively uncover targeted cost-saving opportunities when clients demand budget reductions. Warikoo emphasizes that this is a live, production-grade deployment rather than an experimental demo: "This is running at Cognizant. This is not academic. This is not theoretical. This is not for the faint of heart."

- **Hidden Pipeline Ratio:** $30 of unrecorded pipeline exists outside the CRM for every $100 logged inside it. - **Initial Impact:** $200 million in newly discovered pipeline identified within the first 4 weeks of deployment. - **Deployment Scale:** Active across Cognizant's 1,500-strong global sales force.

[05:27](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=327s) Live Demo: Apex Auto and the Anatomy of a 10/10 Opportunity

Sirish Davuluri takes over the stage to demonstrate the system in action, opening with a lighthearted remark linking Cognizant’s aggressive practical AI adoption to its recent stock performance as the top performer in its market category over the preceding month. Davuluri demonstrates how a Cognizant account executive receives automatically generated, scored opportunities directly on their dashboard and via email notifications, powered by multi-source context aggregation rather than manual entry.

To show how the engine builds an opportunity from disconnected data points, Davuluri explores a high-priority lead generated for a mock enterprise account, "Apex Auto." The platform flagged the deal by correlating disparate signals across six different team members: 1. **Quality Assurance Benchmarking:** Katherine in client delivery was asked by the customer to perform a QA benchmarking report—an isolated request that seemed routine on its own. 2. **Delivery Engineering Notes:** Rupesh, working in delivery in India, logged meeting notes with engineering that indicated friction and issues inside the client's QA organization. 3. **Budgetary Pressures:** Mark from sales captured notes indicating the client was actively seeking ways to lower overall engineering and operational costs.

When the inferencing engine connected these separate breadcrumbs, it synthesized a high-probability value proposition and scored the opportunity as a "10 out of 10" priority. Davuluri stresses that the system preserves human agency: the seller evaluates the AI's evidence, accepts or rejects the opportunity, and sets a directional deal value at Stage 0 without having had to spend hours researching the account's internal dynamics.

[08:17](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=497s) Autonomous Proposal Drafting, Hardening, and Multi-Twin Redlining

Once the seller accepts the Stage 0 opportunity, the platform immediately accelerates the sales cycle by generating a bespoke proposal. Rather than simply producing a generic AI text draft, the Apex Auto account twin actively reaches out to delivery, sales, and finance systems in real time, querying other account twins across the company to see if similar engineering cost-reduction problems have been resolved elsewhere. It also automatically pulls in relevant Subject Matter Experts (SMEs) and finance parameters to construct a structured proposal complete with the appropriate technology stack recommendations.

To prepare the proposal for executive scrutiny, the seller uses an automated "hardening" workflow. The platform analyzes the proposal against historical client pushback, displaying potential buyer objections on the left side of the screen alongside concrete recommendations on the right. The seller selects which counter-arguments to incorporate, and the system dynamically regenerates the proposal.

From there, the platform automates the traditionally painful internal review cycle. Legal twins, finance twins, and functional team twins (such as a twin representing a sales leader named Lynn) access the live document simultaneously and execute real-time redlines and structural edits. The seller skips the standard multi-week internal alignment cycle and receives a fully vetted, redlined proposal ready for final review before presenting it to the client.

[11:08](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=668s) Cross-Account Brainstorming and Vertical Agent Debates

Davuluri concludes the walkthrough by demonstrating cross-account intelligence sharing across Cognizant’s vertical industry units. When an account executive on a different account—such as "Orion Motors"—needs to tackle a similar client objective, the Orion Motors account twin directly queries the Apex Auto twin to learn what strategies and solutions succeeded.

On screen, Davuluri reveals a real-time agentic collaboration feed where the twins debate approaches with one another at the bottom of the interface. In this interaction, the Apex Auto twin suggests pulling in the specialized "SAP practice twin" to evaluate how technical enterprise resource planning issues were previously resolved. The agents independently debate the nuances of the solution architecture and frame the opportunity's core pain points, enabling account teams across the entire firm to instantly reuse institutional knowledge without manual coordination.

[12:14](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=734s) Simulating Account Scenarios and Pricing Impacts

Moving beyond simple collaborative brainstorming between digital twins, Cognizant explores proactive simulation workflows. In one tangential use case, Cognizant tests product pricing strategies, such as simulating the outcome of a 15% price increase on a specific product offering across all clients within a particular vertical. Rather than relying on guesswork, the account twins across that vertical automatically model and evaluate the potential fallout and reception of the price hike.

The simulation reveals distinct account behaviors alongside detailed rationales. For instance, two accounts are flagged as unaffected by the increase because they are actively pursuing cutting-edge technology and perceive the value as worth the premium. Conversely, an account at the bottom of the list is flagged as currently experiencing heavy cost pressures, alerting the account team to tread carefully. This automated intelligence equips sellers with nuanced context before any commercial change is formally communicated to customers.

[13:37](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=817s) Autonomous Account Missions and Delivery Intervention

The most powerful capability within the platform is giving account twins their own autonomous missions. In an illustrated scenario for client "Orion Motors," the account twin is assigned a mission on May 11th to identify churn risks and prevent the customer from leaving Cognizant's services. Upon accepting the mission, the twin continuously inspects ongoing delivery signals across multiple operational locations.

```mermaid flowchart TD accTitle: Autonomous Account Mission Flow subgraph SignalMonitoring["Signal Monitoring"] A["Delivery Operations"] --> B{"Location Status"} B -->|"Dallas / Others"| C["Status: Healthy"] B -->|"São Paulo"| D["Alert: Delivery Disruption"] end

subgraph AgentIntervention["Twin Autonomous Actions"] D --> E["Propose Staffing Resolution: Deploy Field Engineers"] E --> F["Seller Dashboard: Accept / Adjust / Reject"] D --> G["CEO Schedule Match: Generate Executive Briefing"] end ```

While rollouts in cities like Dallas proceed smoothly, the twin detects anomalies and delivery friction at the São Paulo center. It immediately flags the operational issue and recommends a concrete resolution: deploying two engineers immediately to stabilize the engagement. Human agency remains central to this architecture; the system serves as a co-pilot where the seller retains full control to accept, reject, or modify the proposed intervention.

Furthermore, Cognizant's CEO, Ravi Kumar S, has his own digital twin integrated into the system. When the twin detects an upcoming executive meeting on Ravi's calendar with the CEO of Orion Motors, it automatically generates a tailored executive briefing synthesized from delivery signals and account health metrics, enabling a productive high-level customer conversation.

- **Account Mission Example**: Orion Motors assigned a mission to prevent churn and monitor delivery health across global delivery centers. - **Human-in-the-Loop Agency**: Sellers maintain final decision-making power over twin recommendations. - **Executive Integration**: CEO Ravi Kumar S's twin automatically drafts real-time executive briefings before client leadership meetings.

[15:33](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=933s) Enterprise Lessons: Business Process Reimagination Over Hype

Cognizant emphasizes that enterprises should focus less on the underlying AI hype or buzzwords and concentrate instead on context engineering and business process redesign. With their partner WorkFabric AI providing the underlying platform, Cognizant prioritized high-impact use cases directly aligned with frontline sales pain points, securing executive sponsorship, and implementing enterprise-grade security, scalability, observability, and data controls.

A foundational architectural requirement involves cross-account data isolation. When account twins interact, strict guardrails and partition rules guarantee that proprietary client data is never leaked or shared across boundaries. The speakers advise against attempting to "boil the ocean," urging teams to set attainable milestones, start with high-confidence use cases, and scale iteratively.

[17:00](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=1020s) Realized Outcomes: $200 Million in Pipeline and Scaling Customer Zero

Serving as "Customer Zero" for their internal deployment produced rapid, measurable commercial results. Within just five weeks of rolling out the platform across their environment, Cognizant generated $200 million in net new qualified pipeline by ingesting scattered organizational data, extracting latent signals, and operationalizing twin-driven recommendations.

```mermaid flowchart LR accTitle: Pilot to Enterprise Rollout Progression subgraph PilotStage["Phase 1: Pilot"] A["25 Sellers"] --> B["Context & Guardrail Testing"] B --> C["Change Management Validation"] end

subgraph ScaleStage["Phase 2: Scale"] C --> D["Enterprise Rollout across Sales Force"] D --> E["$200M Net New Pipeline in 5 Weeks"] end ```

This success required addressing internal challenges around sales force change management, user adoption, legal compliance, and multi-tenant data governance. The deployment originated with a targeted pilot of 25 sellers, allowing the team to iterate on friction points before rolling the solution out to the broader global sales organization. Cognizant concluded the session on time, offering their lessons and framework to other enterprises looking to turn siloed enterprise data into self-sourced commercial pipeline.

- **Pipeline Generation**: $200 million in net new pipeline produced in the first 5 weeks of production. - **Pilot Scope**: Started with an initial cohort of 25 sellers before expanding enterprise-wide. - **Core Governance Pillars**: Adoption management, cross-twin data boundaries, privacy controls, and legal compliance.

## Notable quotes > Every company is sitting on pipeline it can't see. The signals that would surface it — a renewal at risk in a finance record, a redline in a legal contract — are scattered across systems that rarely talk to each other, most of them beyond the CRM.

> [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) This is running at Cognizant. This is not academic. This is not theoretical. This is not for the faint of heart. — Sushant Warikoo

> [00:04](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=4s) Customer Zero — Sushant Warikoo

> [00:04](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=4s) hyper-intelligence and hyper-productivity — Sushant Warikoo

> [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) account twins — Sushant Warikoo

> [05:27](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=327s) Apex Auto — Sirish Davuluri

> [05:27](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=327s) 10 out of 10 — Sirish Davuluri

> [08:17](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=497s) hardening — Sirish Davuluri

> [11:08](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=668s) Orion Motors — Sirish Davuluri

> [15:33](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=933s) boil the ocean

## Mentioned Cognizant, Ai4, Sirish Davuluri, Sushant Warikoo, Context Fabric, WorkFabric AI, Ravi Kumar S, Apex Auto, Orion Motors, SAP

## Video description Sirish Davuluri - Vice President, AI Strategy & Business Design, Cognizant Sushant Warikoo - Chief Business Officer AI, Cognizant

Every company is sitting on pipeline it can't see. The signals that would surface it — a renewal at risk in a finance record, a redline in a legal contract — are scattered across systems that rarely talk to each other, most of them beyond the CRM. Individually they're easy to miss; together they point to a deal. But no seller has the time to connect them, so the pipeline stays buried.

In this session, Cognizant shares its own "Customer Zero" story: how agents working across sales, delivery, finance, and legal read these signals as they appear and turn them into qualified pipeline the business wasn't tracking —in new opportunities, in weeks rather than quarters.

You'll leave with a practical model for turning the signals you already own into self-sourced demand: how to instrument the enterprise, route the right signal to the right seller, and act before the window closes.

[Source](https://www.youtube.com/watch?v=cOYIRy5Bpmo)

中文简报

Cognizant利用AI“客户孪生”发掘2亿美元沉淀销售商机标准CRM每记录100美元商机便会遗漏30美元合格销售管线，企业收入往往埋没于交付和财务记录中，直至自主智能体将这些信号综合提炼。 Cognizant管理层阐述了其拥有1,500名员工的全球销售团队如何部署自主AI“客户孪生”，以挖掘滞留在传统CRM系统之外的商业价值。该平台通过实时汇聚客户交付、财务记录和技术支持互动等边缘信号，将零散线索关联为打分商机，并实现提案起草、异议应对强化以及多孪生法务与财务审查的自动化。该系统还在确保销售人员全流程参与的同时，支持跨账户智能体辩论、定价模拟和自主风险干预。在严格的多租户数据治理下，该生产系统从最初25人的试点逐步扩展，投产前五周内便挖掘出2亿美元的合格销售管线。企业销售代表将高达60%以上的时间耗费在行政和日常运营开销上，而非积极拓展客户。 (0:04) 标准CRM中每记录100美元的商机，据估计就有30美元的合格销售管线隐藏在非销售类企业系统中。 (1:51) Cognizant部署了自主“客户孪生”，以综合整理交付、IT支持、财务及客户成功记录等跨职能数据。 (1:51) 部署客户孪生架构在系统投入使用的前四至五周内，便挖掘出2亿美元净新增的隐性商机。 (1:51) 推理引擎将交付状况、工程记录和预算压力等孤立数据点进行关联，对高概率成交机会进行评分。 (5:27)

Reading summary

Cognizant overhauled its 1,500-person sales force by deploying autonomous AI "account twins" to surface demand signals trapped outside traditional CRM systems. Leaders report that the system generated $200 million in qualified pipeline within five weeks by automating opportunity scoring, proposal generation, and cross-functional reviews. Context Sensors and the Account Twin Architecture (1:51) Enterprise sales representatives typically spend more than 60% of their working hours on operational overhead—including research, proposal generation, and administrative coordination—rather than client engagement. Sushant Warikoo, Chief Business Officer for AI at Cognizant, noted an internal benchmark indicating that for every $100 in pipeline logged in a standard CRM, another $30 in qualified pipeline sits undetected across delivery, support, finance, and IT systems. Synthesizing Cross-Signal Opportunities and Proposal Hardening (5:27) Sirish Davuluri demonstrated how account twins construct scored leads from disconnected operational breadcrumbs. In a walkthrough for mock client Apex Auto, the engine synthesized a top-priority opportunity by correlating a delivery QA benchmark request, internal notes on engineering friction in India, and sales notes on client budget reductions. Human agency remains deliberate: sellers retain final decision-making authority to accept or reject the Stage 0 opportunity and set directional deal values. Once an opportunity is accepted, the account twin queries relevant delivery, finance, and peer twins to assemble a tailored proposal and technology stack. The platform conducts an automated "hardening" process, testing the document against historical client objections and incorporating seller-approved rebuttals. Specialized legal, finance, and functional twins then conduct simultaneous redlines on the live document, compressing what typically takes weeks of cross-departmental alignment calls into a streamlined review. Collaborative Twin Networks, Simulations, and Autonomous Missions (13:37) Account twins also collaborate directly across Cognizant's industry verticals to share institutional knowledge. During demonstrations, an account twin for "Orion Motors" directly queried the Apex Auto twin and a specialized SAP practice twin to debate solution architectures and reuse proven problem-solving patterns. The system can also simulate commercial adjustments, such as modeling how a 15% price increase across a vertical would impact individual accounts based on their specific technology priorities and budget pressures. Beyond reactive sales assistance, twins can be assigned autonomous missions to monitor account health and mitigate churn. In an illustrated deployment, a twin identified delivery disruptions in São Paulo, recommended an immediate two-engineer field intervention, and notified the account seller for approval. The platform also interfaces with leadership: a digital twin representing CEO Ravi Kumar S automatically drafts synthesized executive briefings when client meetings appear on his calendar. Governance, Pilot Rollout, and Commercial Validation (17:00) Cognizant stressed that executing agentic architectures requires prioritizing business process redesign and data governance over technological novelty. The deployment enforced strict data isolation partitions so that proprietary client data cannot leak across boundary lines during cross-twin interactions. To manage organizational change and validate compliance, Cognizant initiated the project as "Customer Zero" with a 25-seller pilot before scaling across its 1,500-person global sales force. Across the first five weeks of broader enterprise deployment, the twin architecture surfaced $200 million in net new qualified pipeline by operationalizing latent internal signals. Cognizant presents its internal deployment as proof that multi-agent systems can convert latent enterprise data into measurable revenue. However, the organization emphasized that long-term success requires strict multi-tenant data boundaries, disciplined change management, and continuous human oversight to validate agent recommendations.

精炼阅读摘要

IT 咨询企业 Cognizant 将自身 1,500 人的销售团队作为“零号客户”，通过部署基于上下文织网（Context Fabric）的自主 AI 客户数字孪生网络，重塑原本耗费销售人员超过 60% 工时的繁琐运营流程；据其高管介绍，该系统通过跨系统抓取并合成孤立业务信号，在全面上线数周内便挖掘出 2 亿美元的隐性合格商机。销售痛点与隐性商机挖掘机制 (1:51) Cognizant AI 首席商务官 Sushant Warikoo 指出，传统企业销售团队面临严重生产力流失，销售代表往往将超过 60% 的工时消耗在市场调研、内部数据分析、合规流程以及方案与合同起草等运营杂务上，严重挤压了与客户面对面沟通的高价值时间。与此同时，关键客户需求线索常常分散并封锁在传统 CRM 之外的交付和后台系统中。Cognizant 内部测试得出一项基准数据：在传统 CRM 系统每记录 100 美元销售管道的同时，各类非销售系统中通常沉睡着 30 美元完全未被记录的合格商机。为捕获这些隐性价值，Cognizant 在 1,500 名销售人员的终端设备上部署上下文传感器，构建底层上下文织网（Context Fabric），并为重点大客户运行自主运作的“账户孪生”（Account Twins）智能体。这些虚拟客户总监持续抓取并合成来自交付、IT 支持、财务和客户成功等部门的数据记录。Warikoo 表示，这一系统并非概念验证，而是在真实生产环境中运行，在全员上线后的前四周内便挖掘出 2 亿美元传统 CRM 流程未能识别的潜在新管道。离散信号聚合与提案自动化硬化 (5:27) Cognizant AI 战略与业务设计主管 Sirish Davuluri 通过模拟大客户 Apex Auto 演示了智能体如何构建高确定性销售机会。该系统综合了来自 6 名不同团队成员在非销售场景下的离散线索：交付人员 Katherine 接收到的客户例行 QA 基准评测请求、印度交付工程师 Rupesh 记录的关于客户 QA 部门内部摩擦的会议纪要，以及销售人员 Mark 记录的客户削减工程与运营成本诉求。推理引擎将这三项看似孤立的数据串联推演，生成了评分达“10/10”的高优先级商机，并在 0 阶段交由销售代表评估证据、设定初步交易规模，保留了以人为核心的决策权。在销售代表确认商机后，平台进入提案自动化编制与“硬化”环节。客户孪生智能体实时调用跨账户交付与财务经验，调取技术栈与专家配置生成定制化方案，并在屏幕一侧罗列潜在买方异议及应对推荐供销售选择。随后，代表法务、财务和销售主管职能的智能体（如法务孪生、财务孪生）直接在方案上实时执行红线批注与结构性修改。这替代了传统的跨部门拉会与数周沟通周期，使销售代表能够直接获取经过内部各职能严格校审的最终提案草案。跨账户智能体协作与商业模拟 (11:08) 除了单个账户内部的线索挖掘，系统还支持跨垂直行业部门的知识复用。当服务另一客户 Orion Motors 的账户团队面临类似降本挑战时，Orion Motors 客户孪生可以直接向 Apex Auto 孪生发起查询，学习其成功解决路径。在后台交互中，智能体之间展开自主协作辩论，并主动引入“SAP 实践孪生”评估过往技术方案与痛点框架。这种点对点智能体通信机制打破了企业内部的知识壁垒，无需跨团队人工繁琐协调即可实现组织知识的即时复用。在此协作网络基础上，系统进一步演化出商业模拟能力。Cognizant 演示了针对某垂直行业所有客户模拟特定产品提价 15% 的测试场景。各客户孪生综合其对客户当前战略、预算与技术的理解进行推演：两家正积极引入前沿技术的账户被判定不受涨价影响，因其认可溢价价值；而另一家承受沉重成本压力的账户则被标记为存在高度流失风险，从而提前向销售团队发出预警。这种模拟让销售代表在正式对外商务沟通前即可掌握精准的客户接受度全貌。交付异常主动干预与高管孪生 (13:37) 平台的一项核心能力在于为客户孪生指派长期的自主任务。以客户 Orion Motors 为例，孪生智能体被赋予“识别流失风险并维护客户留存”的持续监控任务，不间断审查全球交付网络信号。在监测中，智能体发现达拉斯等运营地交付平稳，但圣保罗交付中心出现交付阻碍与异常波动。智能体随即主动提出应对方案，建议立即调配两名现场工程师以稳定交付质量；销售人员拥有是否采纳该干预措施的完全决定权，确保人机协同控制。同时，系统还将企业最高管理层纳入数字孪生生态中。Cognizant 首席执行官 Ravi Kumar S 拥有系统集成的个人数字孪生。当系统识别到首席执行官日程中排定了与 Orion Motors 首席执行官的会面时，其数字孪生会自动抓取并整合底层交付健康指标与关键账户数据，在会前自动生成针对性的高管简报。这一机制将底层运营事实与高层商务会晤无缝对接，消除了传统层层汇总汇报的信息损耗与滞后。规模化落地路径与数据隔离治理 (15:33) Cognizant 强调，推动此类企业级 AI 变革的关键在于业务流程再造与上下文工程，而非追求模型炒作。项目最初仅由 25 名销售代表组成的小规模团队进行概念与阻力测试，重点验证员工采纳度、变革管理以及工作流顺畅度，随后才逐步扩展至全公司 1,500 人的全球销售组织。最终在全面上线仅 5 周时间内，系统实现了挖掘 2 亿美元净新增合格商机管道的商业回报。在此扩展过程中，Cognizant 与技术平台伙伴 WorkFabric AI 确立了严格的企业级安全与治理框架。最基础的架构边界在于跨账户数据的严格隔离：当不同客户的孪生智能体跨边界交流解决方案与结构性模式时，底层系统通过分区规则和防护机制确保专有的客户私密数据绝不外泄。演讲者建议企业在实施自主智能体项目时，应避免试图一步到位，而应从高置信度痛点用例起步，结合自上而下的高管支持与严格的数据权限控制稳步演进。 Cognizant 通过内部实践验证了智能体网络在挖掘分散信号与压缩企业销售周期上的潜力，但该系统的推广成效高度依赖于前期 25 人试点沉淀的组织变革经验，以及在多账户协作时防范私有数据泄露的隔离机制；如何长期维持人机协同决策的边界并在不同业务线中平衡自动化红线校审与合规审查，仍是此类自主智能体架构需要持续应对的核心课题。

Detailed notes — zh

## Background 本内容由 Ai4 于 2026 年 8 月 20 日发布在 YouTube 平台上。视频记录了 Cognizant 人工智能首席商务官 Sushant Warikoo 与 AI 战略与业务设计副总裁 Sirish Davuluri 的演讲，重点分享了 Cognizant 如何作为“零号客户”（Customer Zero），利用 AI 智能体跨部门抓取 CRM 之外散落的微弱信号，并在数周内将其转化为高价值销售商机。 ## TL;DR Cognizant 针对旗下销售体系推进 AI 变革，旨在实现“超级智能”与“超级生产力”，将销售人员从繁重的案头工作中解放出来以专注于客户交流。通过构建基于上下文工程的“账户数字孪生”（Account Twins），系统能够自动采集并串联跨部门碎片化信号、自主生成并加固商业提案、实现跨智能体实时审阅与定价模拟，甚至自动生成高管简报。从 25 人小规模试点起步，Cognizant 在全量上线仅 5 周内便成功挖掘出 2 亿美元的净新增商机管线，展现了多智能体协同在企业级销售落地的巨大价值。 ## Atmosphere & style 整场演讲呈现出务实、自信且极具实战导向的科技商业风格。两位讲者配合默契，从战略架构的高屋建瓴迅速切换到真实系统的现场实操演示，节奏紧凑明快。台风既有大型咨询巨头推动重大变革的严谨与坚定，又洋溢着前沿技术落地见效带来的兴奋感，给观众一种真实、硬核且非纸上谈兵的沉浸式体验。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=4s) Cognizant 作为“零号客户”（Customer Zero）推进销售 AI 变革，确立了构建“超级智能”挖掘隐藏数据与释放“超级生产力”减少案头事务的两大核心目标。 - [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) 企业销售普遍存在“CRM 每记录 100 美元商机，外部就隐藏 30 美元不可见潜在商机”的痛点，商机往往散落在非结构化数据与跨部门系统中。 - [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) 通过部署 Context Fabric 上下文层与全天候运行的“账户数字孪生”（Account Twins），可在边缘端收集全域信号并自动推理高价值商机。 - [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) 多智能体网络（Agent-to-Agent）支持跨部门与跨账户自主协同，法务、财务和销售孪生可直接协作拟定提案，且不同客户孪生间能直接共享解决方案经验。 - [05:27](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=327s) 上下文引擎能够跨地域、跨角色碰撞碎片化信息（如交付例会痛点、测试基准需求与降本压力），自动聚合出满分（10/10）高优先级商机。 - [05:27](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=327s) 人机协同中始终保留销售人员的完全主观能动性（Agency），销售拥有商机采纳、定级以及执行行动建议的最终决策权。 - [09:49](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=589s) “提案加固”功能可自动预判客户异议并生成应对策略，配合法务与财务孪生的毫秒级实时红线批注（Redlining），大幅压缩内部审校周期。 - [12:14](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=734s) 账户数字孪生可用于模拟商业决策影响（如产品提价 15%），根据客户的技术追求或成本敏感度自动输出针对性行动情报。 - [13:37](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=817s) 支持为分身分配预防客户流失等自主任务（Missions），可跨地域交付中心排查风险异常，并自动为高管会晤生成结构化 CEO 简报。 - [15:26](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=926s) 企业落地上下文工程（Context Engineering）需专注于业务流程重塑并设立严格的多分身通信安全护栏（Guardrails），切忌好高骛远。 - [17:00](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=1020s) 项目采取从 25 人最小可行性试点（Pilot）到全量推广的策略，在系统上线 5 周内即创造了 2 亿美元的净新增销售管线（Net New Pipeline）。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=4s) Cognizant 的 AI 愿景：赋能销售团队实现“超级智能”与“超级生产力” Cognizant 是一家拥有 35 万名员工、年营收约 200 亿美元的全球技术咨询巨头。在本次大会上，Cognizant 人工智能首席商务官 Sushant Warikoo 携手 AI 战略与业务设计副总裁 Sirish Davuluri，共同分享了他们作为“零号客户”（Customer Zero）在企业内部推进 AI 变革的真实历程。Sushant 带领着一支由 AI 设计师、行业策略专家以及前沿工程技术人员组成的精英团队，其核心使命只有一个：通过 AI 将商业价值落到实处。当天分享的主题，正是如何将这一理念应用于 Cognizant 自身庞大的销售体系。 Cognizant 针对旗下 1500 名销售人员发起这场 AI 变革，主要确立了两个核心目标： 1. **构建“超级智能”（Hyper-intelligence）**：传统的商机往往被困在散落于各个角落的非结构化数据中。通过 AI 从这些隐秘位置提取关键信号，为销售团队源源不断地输送高潜线索。 2. **释放“超级生产力”（Hyper-productivity）**：统计显示，销售人员有超过 60% 的时间并未花在与客户交流上，而是被市场调研、起草合同提案、走内部审批流程等繁杂事务占据。Cognizant 致力于颠覆这种现状，把销售从日常运营琐事中解脱出来，将主要精力重新投入到客户面对面的深度沟通中。 ```mermaid flowchart TD accTitle: Cognizant 销售 AI 变革双轮驱动模型 subgraph Drivers["变革双核心"] HI["超级智能 Hyper-intelligence 从分散数据中挖掘隐形信号"] HP["超级生产力 Hyper-productivity 削减60%+的繁琐流程与案头工作"] end subgraph Outcome["最终成效"] INV["颠覆时间分配：专注客户对话"] REV["激活隐藏的潜在商机与营收"] end Drivers --> Outcome ``` --- ### [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) Context Fabric 与“账户数字孪生”：4 周挖掘 2 亿美元隐形商机 Sushant Warikoo 指出一个在企业销售中普遍存在的严峻现实：“在 CRM 系统中每记录 100 美元的商机管线，系统外往往还隐藏着 30 美元完全不可见的潜在商机。”为了把这部分隐藏的价值挖掘出来，Cognizant 与技术合作伙伴 Context Fabric 携手，先在小范围内进行试点验证，随后全面推广至 1500 名销售人员的实际业务中。这项方案的核心在于为公司所有重要客户构建**“账户数字孪生”（Account Twins）**。这些 AI 智能体充当着常驻的自动化客户总监（AE），全天候在后台运转。部署在销售组织所有边缘设备和终端上的上下文传感器，会持续收集来自销售、交付、客户支持、财务和客户成功等部门的信号，并将数据汇聚到底层的上下文层（Context Layer）中。账户数字孪生对这些信号进行实时综合与推理，自动浮现出高价值的商机线索。在方案上线的前四周内，系统便成功挖掘出了高达 2 亿美元在原有 CRM 及传统渠道中完全无法察觉的增量商机。 ```mermaid flowchart LR accTitle: 基于上下文层与账户孪生的商机挖掘架构 subgraph Inputs["全域端点信号采集"] S1["销售与售前终端"] S2["交付与工程会议记录"] S3["财务与合同账目"] S4["售后支持与客户成功"] end subgraph Core["上下文引擎与智能体"] CL["Context Fabric 上下文层"] AT["账户数字孪生 Account Twin"] end subgraph Outputs["业务落地价值"] PL["4周识别 2 亿美元隐形商机"] PROP["多智能体协同提案生成"] end Inputs --> CL --> AT --> Outputs ``` Sushant 强调，这套系统不仅仅是提示线索，更能够驱动多智能体协同办公： - **跨部门智能体协同**：账户孪生可以与法务孪生（Legal Twin）、财务孪生（Finance Twin）自主对接，代替人工拉长途会议，直接协作起草提案与合同初稿，人工只需进行最后的审核与把关。 - **跨账户孪生经验共享**：当遇到特定业务痛点时，一个账户孪生可以直接向另一个账户孪生提问：“你以前解决过这个问题吗？能否分享你的解决路径？”整个经验调用与信息交换过程无需任何人类介入。 - **目标驱动的主动探索**：销售可以直接向智能体下达指令，例如“本季度需要冲刺业绩，请帮我找寻 2 亿美元的潜在机会”或“客户希望降低 2 亿美元成本，请综合内外部数据提出具体建议”。 - Sushant 总结道：“这已经在 Cognizant 全面运行，绝非学术研究，不是纸上谈兵，更不是给胆小者玩的玩具（This is not academic. This is not theoretical. This is not for the faint of heart）。” --- ### [05:27](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=327s) 现场演示：多源信号碰撞与 10/10 高优先级商机的诞生随后，Sirish Davuluri 接过控制权进行现场实操演示。他指出，上下文（Context）是当下 AI 领域的关键，而 Cognizant 已经让每位一线销售切实体会到了它的生产力价值。打开系统仪表板或点击邮件通知后，销售人员会看到一系列由系统根据 CRM 之外的多渠道数据自动整理出来的商机。 Sirish 以虚构客户“Apex Auto”为例展示了线索聚合的推理全过程。系统之所以将该机会标记为 **10/10 满分的高优先级**，是因为它抓取并串联了来自 6 个不同角色的隐性碎片信息： 1. **测试基准需求**：团队成员 Katherine 接到客户要求提供一份质量保证（QA）基准评估报告。单独看这项需求可能很普通，未引起传统销售警觉； 2. **工程交付线索**：身在印度的交付工程师 Rupesh 在与客户工程团队的技术例会纪要中，记录了客户在 QA 环节遇到的工程瓶颈； 3. **商业紧迫性信号**：销售人员 Mark 在业务往来中捕捉到该客户正在承受削减整体工程研发成本的压力。当这些来自不同国家、不同系统的碎片化信息经由上下文引擎推理碰撞后，系统立即识别出这是一个向 Apex Auto 推荐全面 QA 工程降本方案的绝佳商机。系统保留了销售人员的主观能动性（Agency），销售可以自主选择接受或拒绝该商机。一旦点击接受，商机以“阶段 0”（Stage 0）录入，并由销售指定预期价值。紧接着，账户孪生不仅能草拟触达邮件，还会自动联动交付、销售、财务等部门的数据，并检索其他类似账户的成功案例，瞬间生成包含技术栈和业务方案的提案初稿。 --- ### [09:49](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=589s) 提案加固、跨职能智能体实时红线标注与跨行业协同为了避免方案粗糙，系统提供了“提案加固”（Harden）功能。当销售选择加固提案时，AI 会自动推演客户在面对该方案时可能提出的各项质疑与顾虑，在左侧列出所有潜在抗辩理由，并在右侧给出对应的应对建议。销售确认采纳后，提案会实时更新。随后，针对咨询行业耗时最长的内部审查环节，系统展现了跨部门智能体协同能力： - 法务孪生、财务孪生和销售孪生在后台以毫秒级的速度对提案文档进行实时的审阅与红线修改（Redlining），原本需要跨部门反复开会扯皮数周的审校工作，在系统内自动完成，并由各职能孪生代表（如法务 Lynn 的孪生体）完成批注后交付销售。 - 系统还支持跨垂直行业（Vertical）的知识借鉴。在演示的第二部分中，来自汽车行业另一家客户“Orion Motors”的账户孪生主动与“Apex Auto”的账户孪生展开讨论，询问对方是如何成功解决类似痛点的。屏幕下方甚至展示了智能体之间的自主辩论：Apex 孪生建议调入“SAP 实践孪生”（SAP Practice Twin）以评估底层架构兼容性。这种智能体网络（Agent-to-Agent）的自主互动，真正将销售团队从繁杂的信息检索和方案打磨中解放了出来。 ### [12:14](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=734s) 数字分身模拟定价策略对不同客户的影响 AI 分身不仅能用来头脑风暴，还能进行复杂的商业决策模拟。以 Cognizant 自身的产品业务为例，团队模拟了“如果将某款产品价格提高 15%”这一定价策略对特定垂直行业客户带来的连锁反应。系统中各个客户的数字分身（Account Twins）会自动模拟其客户组织的反应：部分分身评估后显示涨价影响不大，并给出了明确的逻辑解释——这些客户正在追求前沿技术（cutting-edge technology），对价格相对不敏感；而对于另一些处于成本压力下的客户分身（如展示在底部的客户），系统则发出警告，提示销售团队在此类账户上必须谨慎行事。这种全自动化的分身模拟将分散的客户背景转化为销售人员可直接参考的行动情报。 ```mermaid flowchart TD accTitle: 模拟定价策略影响流程图 subgraph Input["决策模拟输入"] P["产品提价 15% 策略"] end subgraph Simulation["客户分身 (Account Twins) 模拟分析"] A["高新技术诉求客户"] B["成本敏感型客户"] end subgraph Output["行动情报输出"] O1["评估影响较小：可推进提价并提供理由"] O2["发出风险预警：提示销售谨慎应对成本压力"] end P --> A P --> B A --> O1 B --> O2 ``` --- ### [13:37](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=817s) 自动化账户任务与高管简报生成平台最具威力的功能是为客户分身分配自主任务（Missions）。例如，针对 Orion Motors 这个客户，团队给分身下达了“制定方案防止客户流失（prevent churn）”的任务。分身在 5 月 11 日接受任务后，开始全面审视该账户在全球各地的交付状态：达拉斯（Dallas）等区域进展顺利，但在分析到圣保罗（Sao Paulo）交付中心时发现了异常迹象。系统立即发出警报，并向销售及交付团队给出明确建议：“立即增派两名工程师前往现场解决问题”。整个过程中，销售人员拥有完全的决策自主权（Agency），可以接受、拒绝或调整系统建议。此外，Cognizant 的 CEO Ravi 作为行业前瞻者，也将其专属分身接入了系统；当系统获知 Ravi 即将与 Orion Motors 的 CEO 会面时，便自动生成了一份结构完整的 CEO 简报（CEO Briefing），使高层沟通更加高效精准。 - **自主任务分配**：系统分身自主接单并追踪交付全流程，跨越区域数据排查风险。 - **人机协同控制**：系统仅提供可解释的行动建议，最终行动决策权始终保留在销售代表手中。 - **高管级上下文联动**：自动感知 CEO 日程并提前准备客户背景与问题简报。 --- ### [15:26](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=926s) 企业落地 AI 的核心原则与安全护栏在 Context Engineering（上下文工程）层面，市场上不乏各种底层技术，但核心关键在于“如何让技术真正产生业务价值”。Cognizant 携手合作伙伴 WorkFabric AI 构建平台，将重点放在业务流程重塑（Business Process Reimagination）、推动投产所需的高管赞助（Sponsorship），以及全面管控安全、可扩展性和可观测性（Observability）等风险上。尤为关键的是，当不同账户的数字分身相互通信与协作时，必须设置严格的安全护栏（Guardrails），防止不同客户之间发生非预期的数据泄露。演讲者建议企业设定切实可行的 AI 目标，切忌好高骛远（“Do not try to boil the ocean”），不要盲目追逐新流行词，而是从眼前的真实用例切入并小步迭代。 --- ### [17:00](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=1020s) “Customer Zero”成果：5 周生成 2 亿美元新增商机作为该方案的“零号客户”（Customer Zero），Cognizant 取得了显著的业务成效：在上线实施仅仅 5 周之内，便在内部环境中挖掘并生成了 2 亿美元的净新增商机管线（Net New Pipeline）。这一成果源于对全企业分散数据的全面摄取、智能提炼以及敏捷运营化。回顾整个落地历程，团队成功克服了销售团队的使用习惯与变革管理（Change Management）、分身通信过程中的数据控制与权限隔离，以及法务和隐私合规等核心挑战。项目最初仅以 25 名销售人员的小型核心群组作为试点，在验证成功后逐步推广至 Cognizant 全球销售网络。两位主讲人最后对现场观众表示感谢，强调这套基于上下文工程的实践经验能够为其他正在探索销售 AI 转型的企业提供宝贵借鉴。 - **核心业绩产出**：系统上线仅 5 周，净新增销售管线规模达 2 亿美元（$200M net new pipeline）。 - **试点推广策略**：从 25 名销售的最小可行性试点（Pilot）起步，验证闭环后再全量推广至整个销售团队。 - **落地关键支柱**：涵盖变革管理、数据访问控制（Data Control）、跨分身通信护栏以及法务合规（Privacy & Legal）。 ## Notable quotes > [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) 在 CRM 系统中每记录 100 美元的商机管线，系统外往往还隐藏着 30 美元完全不可见的潜在商机。 — Sushant Warikoo > [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) This is not academic. This is not theoretical. This is not for the faint of heart — Sushant Warikoo > [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) 本季度需要冲刺业绩，请帮我找寻 2 亿美元的潜在机会 > [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) 客户希望降低 2 亿美元成本，请综合内外部数据提出具体建议 > [01:51](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=111s) 你以前解决过这个问题吗？能否分享你的解决路径？ > [12:14](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=734s) 如果将某款产品价格提高 15% > [13:37](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=817s) 制定方案防止客户流失 > [13:37](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=817s) 立即增派两名工程师前往现场解决问题 > [15:26](https://www.youtube.com/watch?v=cOYIRy5Bpmo&t=926s) Do not try to boil the ocean ## Mentioned Cognizant, Ai4, Sushant Warikoo, Sirish Davuluri, Context Fabric, WorkFabric AI, Apex Auto, Orion Motors, Katherine, Rupesh, Mark, Lynn, Ravi, SAP ## Video description Sirish Davuluri - Vice President, AI Strategy & Business Design, Cognizant Sushant Warikoo - Chief Business Officer AI, Cognizant Every company is sitting on pipeline it can't see. The signals that would surface it — a renewal at risk in a finance record, a redline in a legal contract — are scattered across systems that rarely talk to each other, most of them beyond the CRM. Individually they're easy to miss; together they point to a deal. But no seller has the time to connect them, so the pipeline stays buried. In this session, Cognizant shares its own "Customer Zero" story: how agents working across sales, delivery, finance, and legal read these signals as they appear and turn them into qualified pipeline the business wasn't tracking —in new opportunities, in weeks rather than quarters. You'll leave with a practical model for turning the signals you already own into self-sourced demand: how to instrument the enterprise, route the right signal to the right seller, and act before the window closes. [Source](https://www.youtube.com/watch?v=cOYIRy5Bpmo)

Source quotes

[1:51] Sushant Warikoo: 在 CRM 系统中每记录 100 美元的商机管线，系统外往往还隐藏着 30 美元完全不可见的潜在商机。 [1:51] Sushant Warikoo: This is not academic. This is not theoretical. This is not for the faint of heart [1:51] 本季度需要冲刺业绩，请帮我找寻 2 亿美元的潜在机会 [1:51] 客户希望降低 2 亿美元成本，请综合内外部数据提出具体建议 [1:51] 你以前解决过这个问题吗？能否分享你的解决路径？ [12:14] 如果将某款产品价格提高 15% [13:37] 制定方案防止客户流失 [13:37] 立即增派两名工程师前往现场解决问题 [15:26] Do not try to boil the ocean

Source quotes

Every company is sitting on pipeline it can't see. The signals that would surface it — a renewal at risk in a finance record, a redline in a legal contract — are scattered across systems that rarely talk to each other, most of them beyond the CRM. [1:51] Sushant Warikoo: This is running at Cognizant. This is not academic. This is not theoretical. This is not for the faint of heart. [0:04] Sushant Warikoo: Customer Zero [0:04] Sushant Warikoo: hyper-intelligence and hyper-productivity [1:51] Sushant Warikoo: account twins [5:27] Sirish Davuluri: Apex Auto [5:27] Sirish Davuluri: 10 out of 10 [8:17] Sirish Davuluri: hardening [11:08] Sirish Davuluri: Orion Motors [15:33] boil the ocean

Original transcript

[0:04] Good afternoon everyone. My name is Susant Wariku. I'm the chief business officer for AI [0:11] for Cognizant and I'm joined today by my colleague Sesh Dealuri who heads AI strategy and business [0:18] design. To tell you a little bit about Cognizant, we are a 350,000 people consulting company. You [0:26] probably heard of us. about $20 billion in size and I am responsible for an elite organization [0:34] within the firm that consists of AI designers, industry strategists and frontier engineers who [0:43] are together in one unit and they have only one goal which is to drive business value realization [0:49] from AI make it come to life. Uh today's story is all about how we brought AI to life for our [0:58] own company within the sales organization. Our objective of transforming sales with AI was [1:05] twofold. One was drive hyper intelligence for our,500 strong sales organization and [1:13] give them signals that they are missing from data that's trapped in undisclosed locations. [1:21] So drive hyper intelligence for them, give them the leads they need. And the other one was drive [1:25] hyperproductivity. Sales organizations spend 60% plus time not talking to clients but doing market [1:33] analysis, research, building contracts, proposals, following internal processes, [1:38] etc., etc., etc. while they should be spending more time with clients. We wanted to free up our [1:44] sales organization so that they could invert this and spend more time with our clients and less time [1:51] on actually the operational stuff. Let me move forward. For every $100 in CRM pipeline which is [2:01] in your CRM system, there's $30 hidden outside is what our finding is which is not visible to [2:09] you. If you look at what we did, we implemented this for a small pilot first at our company and [2:18] then scaled it out with,000 1500 people. This is running right now. The technology we used [2:24] is called context fabric. It's a company that's a partner with Cognizant. And what we did is we [2:31] basically created account twins for all of our major accounts out there. These are AI agents [2:37] that are running continuously and acting as account executives if you may. And what these [2:44] account twins do is they gather information from sales, from delivery, from support, from finance, [2:52] from customer success and extract all of that information and unearth signals that can lead [2:59] to new opportunities. In the first four weeks of implementation, we were able to unear $200 million [3:06] of pipeline that was not visible to us within our CRM system or through conventional channels. [3:13] How did that happen? What we really do is once we implemented the context layer at Cognizant [3:19] and we onboarded all of our sales organization. Now what this really really means is there are [3:25] context sensors sitting on every device on every edge device on every user device the [3:32] entire sales organization which is collecting all of these signals continuously. Those signals are [3:38] fed into the context layer and then the account twin synthesizes this information and bubbles [3:44] up relevant signals or relevant leads and then we'll talk about the use cases that we put them [3:50] to work with. The scenarios that my colleague Sish will talk about will range from how did [3:56] we unear pipeline that was not visible otherwise. What you will also see is how an an accounts twin [4:05] worked with a legal twin and a finance twin to put together an actual proposal. How much time [4:11] do we spend on pi calls where all we're doing is putting a proposal or putting a contract together [4:17] with finance, with legal, with sales teams. The agents will do the work and we will review and [4:23] validate the final iteration. He would talk about a scenario where we're trying to solve a problem [4:28] for a client and one account twin is talking to another account twin and saying, "Have you solved [4:34] this problem before? Can I learn from what you the way you solved it?" So these are AI agents talking [4:39] to each other, exchanging information when they're trying to solve similar problems across the client [4:44] base. No human involvement. You'll also see a scenario where you said, "Listen, I need to make [4:49] my quota for this quarter account when can you go find me $200 million of pipeline or my client is [4:58] looking to reduce cost by $200 million. Can you go figure out a way? Look at external sources, [5:03] look at internal sources. Can you come up with ideas around how can we deliver that extra saving [5:08] for our client?" These are the scenarios you will see in action. This is running at Cognizant. This [5:14] is not academic. This is not theoretical. This is not for the faint of heart. So sir, [5:19] I will with that move this to the demo. Can I grab that clicker? Thank you. Thanks. Hello everyone. [5:27] So context, context, context. Probably you'll hear that a lot during this conference. But what you'll [5:32] see here today is how we've actually implemented it and made it practical so that every seller at [5:40] Cognizant and every person at Cognizant can do more. And we'd like to think something like this, [5:47] innovations like this has a lot to do with our recent stock performance if you're following [5:53] Cognizant. Uh we were the best performer by the way in the last month in uh uh in the in in the [6:00] stock market. So here you have a number of uh opportunities that showed up automatically for [6:07] a seller. And how did they show up? not just from what's sitting in CRM but also from all the other [6:13] channels that you heard Susan talk about like delivery, sales and so on and so forth. Here when [6:19] a seller a cognizant seller opens the dashboard or through clicks through an email notification [6:26] they'll see all these opportunities that are there that are automatically populated for the [6:31] seller to action on. So how does that work? Like I said, it senses all the context from a [6:38] number of different sources and combines all these signals and says that hey, Apex Auto, for example, [6:44] which we will double click into uh in a second, has a high probability of a new opportunity that [6:52] you would want to proactively reach out to the customer. And how did we know that? Well, number [6:57] one, as I was telling you about, like it picks up context from a number of different sources. Like [7:01] you can see all the signals like six different people that the system the platform picked up [7:07] from and then look if you look at the pain points it addresses like Katherine for example here has [7:14] found out has been asked to do a quality assurance benchmarking report. Now that that in itself might [7:21] not be like super okay the client may be asking for a benchmarking but Katherine here goes through [7:27] and then another piece of evidence shows up from uh Rupes all the way in India looking at [7:33] uh delivery meeting notes with engineering that flags that there's something going on with QA [7:39] then the what's the urgency then we see a person from a sales mark not notices that the client may [7:47] be looking to lower overall engineering cost. So all these signals coming together makes a [7:53] pretty high compelling value proposition for the system to flag that hey this is a very important [8:00] opportunity for you to pay attention to. So that's what happens uh when all this inferencing [8:06] comes together that this opportunity becomes a high priority 10 out of 10. If you notice at [8:12] the top right there will be a 10 out of 10 kind of rating for this one. Now the agency is still [8:17] with the seller. it is they've already only been informed there something like this they can either [8:22] reject the opportunity or accept it if they accept still still it's state zero opportunity they pick [8:30] a directional uh value and then they say okay here's the interesting part they've identified [8:36] the opportunity now how do you accelerate it now this is where most of the time that goes without [8:41] being noticed from a seller perspective they want to be engaged in the clients with clients all the [8:48] So building an automated email draft is not a big thing. But what you're seeing here is that [8:54] account twins this particular Apex auto account twin talking to a number of people from whom [9:02] information has been pulled here whether it's delivery sales finance etc. and other accounts [9:08] too to see if anything similar has been solved elsewhere. Now all of this is happening in real [9:14] time without the seller actually doing anything. And it'll also pull in SME because it's a uh cost [9:20] reduction thing. The seller has clicked here like uh uh bring in finance as well and it has [9:26] to build the proposal and then it automatically builds the proposal for the seller. Now this is [9:30] not by no means uh a completely done uh proposal. It's a starting point. That in itself is a huge [9:37] step forward for a seller who spent so much time pulling things together. Now this has a variety [9:42] of things like technology stack and so on and so forth that uh seller can start from anywhere. The [9:49] next step the seller would choose is to harden. They might want to harden the proposal. So then it [9:55] automatically goes and picks up like what could be possible objections for this particular proposal [10:01] and then on the left you see all the different objections that might have come up and on the [10:06] right are the recommendations. Now once again the agency is with the seller. the seller can [10:11] pick which ones to incorporate into the proposal and there you go it regenerates that proposal for [10:17] the seller to action on. Now this is one part. Another part is usually we do like uh word [10:23] documents kind of uh uh responses and that's what Susant was sharing earlier where we added legal, [10:30] we added sales, we added finance all of them in real time are actually redlinining the proposal. [10:37] Then this is again the seller is still not yet involved in shaping this. It's the system that's [10:43] actually pulled this together. And uh you see these uh twins of other um functions like Lynn and [10:51] other folks they're automatically editing their twins are editing the proposal and when it's done [10:56] it's again presented back to the seller to see okay here's here's what you do go take it to the [11:02] customer kind of thing. Now that's the first part. Now there are three other elements that in itself [11:08] closes that workflow but there are three other elements that we'd love to show use cases of what [11:13] we've implemented. Again every seller at Cognizant has access to this. What what it's showing right [11:19] now is if a seller asked okay brainstorm with other accounts in your vertical because we are [11:25] structured by vertical they might want to look at what's happening elsewhere. In this case, the [11:30] Orion Motors one we did just did the Apex one, but the Orion Motors account twin is now interested [11:36] to see, okay, what did you do? And then the twins are talked to each other and then they they frame [11:42] up what is the opportunity, what pain points it addressed. Now, what's interesting is what you're [11:47] seeing right now on the screen at the bottom where the twins themselves are debating with each other. [11:53] Apex twin is saying okay I uh bring in the SAP twin uh SAP practice twin to see what's going on [12:01] uh how they have addressed some of the issues and so on and so forth. So once again, it's all agents [12:08] talking to each other to help the seller take an action and to get them to spend majority of their [12:14] time in the field talking to customers uh and and drive more uh revenue and pipeline and so on. So [12:22] that's the uh first step around like brainstorming with another uh twin. Now we'll look at a couple [12:28] of other use cases where what if we want to simulate this is a tangential use case cognizant [12:34] we also have products that we sell uh what if like one of the products we'd want to increase [12:41] uh the price by 15%. So what would be the impact of uh the clients that are in that particular [12:48] vertical of uh that said pricing strategy. Now here it's automatically the twins are of all [12:54] these accounts are trying to simulate well this price increase might have a not might not have a [13:03] significant impact on some but there are others that you would need to pay attention to how that [13:09] might play out as an outcome. So as you can see two of the accounts uh the twins say that okay [13:15] they might be fine and it also gives the rationale okay this is why they uh they're looking for this [13:20] cutting edge technology so they might be okay with the price increase and others for example active [13:26] one at the very bottom they might say okay they're a little bit going through under cost pressures [13:31] right now so you got to tread that a little bit carefully again all intelligence for the seller to [13:37] do better now the final view here I want to share is the best part of the whole platform here. We're [13:46] deploying we're giving these accounts missions of their own to go figure out like for example for [13:51] Orion Motors we're asking it to do go hey figure out a way to prevent churn of the customer from [13:58] our services. So we gave that mission. You see that on May 11th it accepted the mission and [14:04] then it starts to see well what's going on with this account. There are apparently some rollout [14:09] happening that our delivery teams are managing. So Dallas is fine. Some other areas are fine. But the [14:16] first flag that it sees that oh something could be off is when it hits Sao Paulo. So all of a sudden [14:22] there's some issue going on at the Sao Paulo uh center that uh there might be some issues. [14:28] So then it flags that and it automatically helps the seller and the entire team involved in making [14:34] this happen that hey to resolve this what I would recommend is deploying a couple of engineers right [14:43] away to address the problem and again the agencies with the seller here is the system [14:48] is only helping the seller do things better. Now the seller has the choice to either accept it or [14:55] reject it or adjust however possible. Now here's another interesting thing. Our CEO is very tech [15:01] forward is is one of the best forward thinkers in the industry. He has also given access to his own [15:08] twin to the system. Now the system automatically learned that he has a meeting with the CEO at [15:13] Orion Motors. So the system automatically built a CEO briefing so that it's now readily available [15:20] for Ravi to take a look at it and have a very productive conversation. So that's the uh missions [15:26] that we can give our accounts here. We have five twins on a mission. Now this is our starting point [15:33] around early stage opportunities for the seller to act on. So the important takeaway here is this [15:40] focus less on the technology. You'll see that there are a lot of technologies out there to [15:44] implement the context engineering layer. Our focus was more on how do we put the technology to work. [15:52] We had a good platform in our partners' work fabric but we thought about the right use [15:56] cases that we had to focus on which were really relevant to our sales organization. So focus on [16:03] the business process reimagination focus on the sponsorship needed to get this in production so [16:08] people can start adopting it leveraging it. Focus on mitigating the risk and uh the concerns around [16:15] security and scaling and observability etc etc. Remember when we have accountants talking to [16:20] each other, we have to make sure that we're not sharing information between clients that we're not [16:25] supposed to share. So guardrails like that had to be implemented to make sure this technology [16:30] can work as expected. But what you see here is something that we are using and is driving [16:37] tangible value for us. Create attainable goals with AI. Do not try to boil the ocean. Do not pick [16:43] up the new buzzword. you take what you have in front of you imple implement it for the right use [16:49] cases and move forward from there in an iterative fashion that's kind of the learnings that we had [16:55] and if you are trying to transform your own sales organization as well with AI we definitely have [17:00] lessons to share yeah great uh thank you so in the pudding as you can see here we've generated $200 [17:08] million in net new pipeline within just 5 weeks of implementing in our environment ment. So, [17:16] and how we did it, we ingested everything. We extracted the intelligence and operationalized [17:21] the solution in a way that you just saw. Now, um it wasn't without problems, without issues [17:27] that we needed to solve for ourselves, which in a way gives those who would like to uh get our [17:34] help in doing something like this a head start. We had to solve for adoption, change management [17:40] uh within our um you know the the sales community within cognizant and all the account twins [17:47] etc. As Susant was saying, we had to implement the data control aspect so that as twins talk to each [17:54] other, there are rules that they follow and that nothing is overshared um you know inadvertently. [18:01] And then privacy and legal from a complaint perspective are also core elements or issues [18:07] that we've addressed to for ourselves. We started with just 25 sellers as a pilot like any pilots [18:13] around AI happen a small core group. We learned from that and we saw that it was working and we [18:20] slowly started rolling out to entire cognizant sales force. So we are happy to help if anybody [18:26] here is interested in talking uh to us about what we did, how we did um and how like context [18:34] is actually working for us uh through the our partner partner implementation with work fabric [18:40] AI. Uh so together we've been able to accomplish something phenomenal and we're really proud of [18:47] uh how we've been in customer zero for this uh implementation and solution. I can't believe [18:54] we finishing on time s. Yeah. Thank you very much for listening to us after a lovely lunch.