Pinegrove, InterSystems Pitch Autonomous AI Agents Over Legacy SaaS
Ai4
Autonomous AI agents are replacing legacy sales software by driving 20% to 30% efficiency gains and outperforming human persuasion in early deployments.
## Background
Published on YouTube by the channel Ai4 on August 29, 2025, this video features a presentation by Adam Nahari (Partner, Chief Data and AI Officer at Pinegrove Venture Partners) and Kanishk Mittal (Lead Technical Specialist at InterSystems). According to the page description, the session explores how AI is transforming legacy sales and marketing SaaS platforms into autonomous, goal-driven systems, highlighting real-world efficiency gains at companies like Klarna, Siemens, and Hitachi alongside frameworks for adoption and governance.
## TL;DR
The presentation argues that after 15 years of static, siloed tools, AI is fundamentally shifting go-to-market (GTM) architectures from rigid assembly lines to real-time, adaptive systems like a Formula 1 pit crew. Highlighting research where AI outperformed humans in persuasion, as well as enterprise case studies, the talk demonstrates how autonomous agents streamline customer support, sales orchestration, and marketing flywheels. The speakers outline a three-phase framework—Layer, Automate, and Rewire—to help organizations modernize their tech stacks without operational disruption. During the Q&A, the discussion dives into practical implementation trade-offs, addressing change management challenges with top sales reps, data integration prerequisites, and lightweight automation strategies.
## Atmosphere & style
The session carries an authoritative, pragmatic, and forward-looking tone typical of an enterprise tech conference. The presentation delivers structured insights with clear analogies and real-world enterprise examples, creating an engaging and analytical atmosphere. The transition into the interactive Q&A brings dynamic debate, where audience pushback on implementation friction adds a grounded, realistic sense of operational problem-solving.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=8gYch7PEHww&t=4s) Legacy sales and marketing tech stacks have remained unchanged for over 15 years, operating as disconnected, siloed tools.
- [00:04](https://www.youtube.com/watch?v=8gYch7PEHww&t=4s) AI is transitioning from an add-on support tool to the central operating system for go-to-market workflows.
- [01:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=78s) Autonomous AI agents perform tasks like persuasion and mind-changing 3 to 6 times more effectively than humans due to 24/7 consistency and lack of emotional variance.
- [02:19](https://www.youtube.com/watch?v=8gYch7PEHww&t=139s) Traditional GTM workflows operate like rigid assembly lines, whereas AI-native architectures act like an adaptive Formula 1 pit crew by using real-time data.
- [04:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=258s) Enterprises like Klarna, Siemens, and Hitachi demonstrate 20% to 30% efficiency gains using AI for customer support, call-level guidance, and behavioral marketing triggers.
- [06:10](https://www.youtube.com/watch?v=8gYch7PEHww&t=370s) The GTM model is moving from a delayed linear funnel to a real-time continuous flywheel, shifting roles like BDRs from execution to AI orchestration.
- [07:45](https://www.youtube.com/watch?v=8gYch7PEHww&t=465s) Organizations can adopt AI safely using a three-phase roadmap: Layer (lightweight overlays), Automate (end-to-end workflows), and Rewire (full GTM redesign).
- [10:50](https://www.youtube.com/watch?v=8gYch7PEHww&t=650s) While central data lakes and open APIs are ideal for long-term AI strategy, lightweight orchestration tools allow teams to achieve fast, immediate wins.
## Detailed walkthrough
Here is a detailed, faithful walkthrough of the transcript:
[00:04](https://www.youtube.com/watch?v=8gYch7PEHww&t=4s) Rethinking the Go-To-Market Stack: From Static Tools to Autonomous Agents
The speaker begins by thanking the audience for their patience at the end of a long conference before diving into the core premise of the presentation: what happens when AI stops merely supporting a company's sales and marketing tech stack and starts fundamentally replacing it? For over 15 years, go-to-market (GTM) technology stacks have remained largely unchanged, operating as a collection of siloed tools patched together with extra software to bridge functional gaps. We are now at a pivotal turning point where AI transitions from being just another add-on tool to serving as the central operating system for how sales and marketing workflows are executed.
Rather than relying on legacy systems that merely record information, the emerging paradigm centers on autonomous agents that can analyze data, make decisions, and take direct action without constant human intervention. Crucially, these autonomous systems are proving not only more efficient than traditional tools, but also superior at tasks previously assumed to require a uniquely human touch.
- Legacy sales and marketing tech stacks have gone without fundamental structural changes for over 15 years.
- Autonomous AI agents do not just log data; they make decisions and act independently.
---
[01:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=78s) The Persuasion Shift: AI Outperforming Humans in Influence
To illustrate AI's growing ability to handle complex human interactions, the speaker highlights a research study conducted on Reddit. AI agents were deployed under specific personas to engage in online discussions with the goal of changing people's minds. The results were surprising: the AI agents were three to six times more effective at shifting opinions than actual human Reddit users.
The underlying power of AI in persuasion stems from its absolute consistency. While human performance fluctuates due to bad days, fatigue, or emotion, AI operates continuously 24/7, delivering steady, data-backed arguments. If an AI agent can successfully change minds in heated, deeply personal online forum debates, it holds immense potential for business applications such as sales emails, deal negotiations, and customer contract renewals.
- In a Reddit-based experiment, AI agents were 3 to 6 times (300%–600%) more effective at changing users' minds than human participants.
- AI provides constant 24/7 engagement backed by data without emotional variance.
---
[02:19](https://www.youtube.com/watch?v=8gYch7PEHww&t=139s) Reimagining Workflows: Assembly Lines vs. Formula 1 Pit Crews
The fundamental flaw in modern go-to-market stacks is fragmentation. CRMs rarely communicate seamlessly with marketing platforms, BI reporting lags days behind real-time activity, and systems depend heavily on disconnected SaaS tools and manual data entry. This setup creates severe human bottlenecks, forcing personnel to manually pull reports, analyze metrics, and trigger campaigns—a process compared to "driving a sports car with the parking brakes on."
Traditional SaaS operates like a factory assembly line: linear, rigid, and predictable. Sales reps manually log data into a CRM, execute predetermined workflows based on static lead-scoring rules, run fixed reporting sequences, and adjust strategy only after reviewing past results. Generative AI removes this friction by turning the workflow into an adaptive system analogous to a "Formula 1 pit crew." Instead of waiting for a campaign to end to evaluate performance, AI continuously ingests real-time data from disparate sources, dynamically reprioritizes leads, personalizes outreach, and adjusts strategy on the fly.
- Traditional GTM stacks suffer from disconnected tools, manual data entry, and delayed BI reporting.
- Traditional SaaS follows a rigid sequence: manual data entry $\rightarrow$ fixed rule workflows $\rightarrow$ scheduled reports $\rightarrow$ post-cycle adjustments.
- AI-native architectures dynamically ingest multi-source data, reprioritize leads, personalize outreach, and continuously optimize campaigns in real time.
---
[04:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=258s) Real-World AI Transformation: Klarna, Siemens, and Hitachi
The practical impact of AI-driven GTM workflows is already visible across major global enterprises through three distinct applications:
1. **Klarna:** Previously relied on a massive support staff with variable response times and satisfaction ratings. After adopting AI, autonomous agents handled two-thirds (two out of three) of all customer support chats, driving efficiency gains of 20% to 30% while reducing support headcount and delivering faster, more consistent responses.
2. **Siemens:** Overhauled its sales orchestration. Historically, opportunity prioritization and forecasting depended on static CRM rules, gut feelings, and incomplete data. Siemens introduced predictive AI to guide sales representatives with real-time, data-backed suggestions during live sales calls, resulting in higher pipeline accuracy and smarter allocation of sales reps.
3. **Hitachi:** Replaced traditional, fixed-duration marketing campaigns with adaptive AI systems. Rather than running rigid schedules, AI continuously monitors real-time user behavior to trigger tailored messaging automatically. Prospects enter and exit Hitachi's marketing flywheel at any point in their buyer journey, significantly reducing lead drop-off and increasing sales velocity.
- **Klarna:** AI handles 2/3 of customer support chats, achieving 20%–30% efficiency gains alongside reduced support headcount.
- **Siemens:** Uses real-time predictive AI during live calls to replace gut-feel forecasting, improving pipeline accuracy and rep allocation.
- **Hitachi:** Replaced fixed-date marketing campaigns with real-time behavioral triggers, allowing leads to enter and exit an adaptive GTM flywheel.
---
[06:10](https://www.youtube.com/watch?v=8gYch7PEHww&t=370s) The GTM Flywheel and the Evolution of Sales Roles
Traditional GTM operates as a linear funnel: leads enter through chatbots or cold outreach, undergo manual qualification, suffer handoff delays, and are eventually worked to closure by sales reps. AI replaces this linear progression with a continuous "flywheel" driven by real-time behavior tracking, AI-driven cross-channel outreach, and automated machine learning loops that update strategy instantly without waiting for post-campaign reviews.
This technology transformation is reshaping job definitions across revenue teams. The role of a Business Development Representative (BDR)—traditionally focused on manual high-volume cold outreach and "hammering the phones"—is shifting toward designing, tuning, and orchestrating AI agents. Rather than manually digging through CRMs to qualify leads, reps now instruct AI agents directly with commands like "Prioritize leads by intent." This shift does not eliminate human workers; instead, it elevates their roles by offloading routine execution and freeing up time for high-level strategy and relationship-building.
- The GTM model is shifting from a linear, delayed sales funnel to a continuous, real-time AI flywheel.
- BDR roles are evolving from execution-heavy task work (e.g., cold calling, manual lead entry) to orchestrating and supervising AI agents.
---
[07:45](https://www.youtube.com/watch?v=8gYch7PEHww&t=465s) A Three-Phase Framework for AI Adoption
To implement AI without causing operational disruption, organizations should follow a structured three-phase adoption roadmap:
1. **Layer:** Add lightweight AI features on top of existing tools to secure low-risk, immediate wins (e.g., automated CRM transcript summarization or basic lead prioritization).
2. **Automate:** Once teams are comfortable, build end-to-end automated workflows, such as intelligent prospect routing across systems.
3. **Rewire:** Fully reimagine the entire go-to-market process from scratch, free from the constraints of legacy SaaS architectures.
Leaders must avoid trying to "boil the ocean" all at once. The key to successful rollout is starting small, demonstrating tangible business value, building organizational credibility, and scaling systematically.
- **Phase 1 (Layer):** Integrate low-risk AI overlays (summarization, simple prioritization) into existing tech stacks.
- **Phase 2 (Automate):** Establish end-to-end automated workflows like intelligent lead routing.
- **Phase 3 (Rewire):** Completely redesign GTM processes around an AI-native operational model.
---
[08:44](https://www.youtube.com/watch?v=8gYch7PEHww&t=524s) Key Takeaways and Q&A: Tool Native Agents vs. Platform Native Data
Concluding the formal presentation, the speaker emphasizes that legacy SaaS platforms are becoming obsolete, and companies that adopt AI-native GTM strategies will outperform those that do not. Rather than asking whether AI will replace workflow tasks, leaders should evaluate *which* specific parts of their operations they want AI to take over first: "If AI can run 30% of your GTM process tomorrow, what's the first thing you'd want to hand over?"
During the Q&A session, an audience member asks about the specific orchestration tools being used in practice. The speaker explains that major platforms are embedding native agentic capabilities directly into their software. For instance, their team uses Pega as a CRM, which has introduced native agent features that leverage existing customer data and internal governance controls to save reps time. While legacy vendors are adding AI features, the speaker anticipates the near-future emergence of dedicated, "AI-first" CRM suites.
- Modern CRMs like Pega are building embedded agentic capabilities directly into existing enterprise platforms.
- An AI-first suite of CRM tools is expected to emerge as the market matures.
---
[10:50](https://www.youtube.com/watch?v=8gYch7PEHww&t=650s) Q&A: Human Supervision, Enterprise Data Lakes, and Adoption Barriers
The Q&A continues with a discussion on agent orchestration, organizational data readiness, and change management:
- **Autonomous Agent Learning & Supervision:** An audience member asks at what point human supervision becomes unnecessary and whether agents can learn to orchestrate other agents. The speaker acknowledges that while current models still require human oversight to maintain quality and safety, advances in prompting and model architectures will eventually allow agents to handle multi-agent orchestration autonomously.
- **Building an "AI Brain":** In response to a query about feeding legacy marketing data and APIs into a unified central system to let agents self-correct and consult external LLMs, the speaker agrees that establishing an enterprise data lake is the ultimate goal. However, because GTM stacks are currently fragmented across dozens of siloed systems, consolidating data and exposing robust APIs remains a major prerequisite step.
- **Change Management & Top Performers:** When asked about common implementation hurdles, the speaker identifies change management as the primary obstacle—specifically convincing top-performing sales representatives who consistently hit quota without AI. Convincing successful reps to adhere to enterprise-wide data standards (e.g., logging data inside the CRM rather than personal spreadsheets) requires demonstrating that AI is necessary for long-term organizational competitiveness.
- **Prerequisites for Getting Started:** For companies seeking quick, low-risk AI wins, the speaker jokingly remarks that "before AI, you need BI." Organizations must ensure that any new tool they acquire features open APIs, webhooks, and enterprise data-lake access so that the company retains full ownership of its data for future AI agent deployment.
- **Alternative Viewpoint on Data Integration:** A audience member pushes back on the requirement for a complete data lake first, arguing that consolidating decades of legacy data across large enterprises can turn into a years-long endeavor that delays progress. Instead, they suggest starting small with lightweight automation tools like n8n, which allow teams to deploy functional agents in hours rather than years.
- Enterprise data lakes and open APIs/webhooks are essential for feeding clean, consolidated data to AI agents.
- Change management remains a major barrier, particularly when persuading quota-attaining sales reps to adopt standardized AI workflows.
- An alternative implementation strategy advocates using lightweight orchestration tools (e.g., n8n) to achieve rapid wins without waiting for full enterprise data consolidation.
[16:34](https://www.youtube.com/watch?v=8gYch7PEHww&t=994s) Practical Next Steps: Starting Small vs. Long-Term Strategy
Building on the foundation of bringing data into a single unified place, the discussion focuses on how organizations can balance immediate implementation with long-term vision. While unifying data across systems is essential for a comprehensive long-term strategy, leaders should avoid getting overwhelmed by trying to do everything at once. Instead, the recommended approach is to start small by applying AI to specific, high-friction individual steps within existing sales and marketing workflows.
To demonstrate how to start small, simple yet high-value use cases are highlighted, such as prepping for prospect calls. For example, a sales representative can ask the AI system to provide background details on an organization, check whether the company has tried selling to them in the past, or analyze why a previous deal did not work out. By tackling these immediate, low-risk query tasks first, teams can deliver quick wins while progressively laying the groundwork for full data consolidation and broader GTM strategy integration.
## Notable quotes
> [02:19](https://www.youtube.com/watch?v=8gYch7PEHww&t=139s) driving a sports car with the parking brakes on
> [07:45](https://www.youtube.com/watch?v=8gYch7PEHww&t=465s) boil the ocean
> [08:44](https://www.youtube.com/watch?v=8gYch7PEHww&t=524s) If AI can run 30% of your GTM process tomorrow, what's the first thing you'd want to hand over?
> [13:00](https://www.youtube.com/watch?v=8gYch7PEHww&t=780s) before AI, you need BI
## Video description
Adam Nahari, Partner, Chief Data and AI Officer, Pinegrove Venture Partners
Kanishk Mittal, Lead Technical Specialist, InterSystems
Abstract
What happens when AI stops merely supporting tools - and starts replacing them? This talk explores The AI-Powered Future of Sales and Marketing, where AI is transforming legacy SaaS platforms into dynamic, goal-driven systems of work. The session will outline how this shift is already reshaping roles, collapsing traditional workflows, and enabling faster, smarter, and more adaptive go-to-market strategies.
Examples from companies like Klarna, Siemens, and Hitachi will illustrate how AI-driven workflows are delivering 20–30% efficiency gains by replacing static tools with intelligent agents. The talk will also provide practical frameworks for initial adoption, governance, and organizational alignment - offering a clear roadmap for Sales and Marketing leaders seeking to stay ahead of the curve as legacy tech stacks become increasingly outdated.
Go-to-market technology stacks are shifting from disconnected, human-driven software tools to autonomous AI operating systems that continuously analyze data and execute workflows. Enterprise implementations demonstrate that replacing linear funnels with adaptive systems improves operational speed and consistency, though data readiness and change management remain major hurdles.
From Fragmented Assembly Lines to Real-Time Pit Crews (2:19)
For over 15 years, traditional sales and marketing stacks have operated as siloed tools bound together by manual data entry and lagging business intelligence reports. The speaker compares this setup to an assembly line, where sales reps follow rigid lead-scoring rules and only evaluate performance after campaigns finish. AI transforms this structure into an adaptive model likened to a Formula 1 pit crew, where autonomous agents ingest continuous data to reprioritize leads, tailor messaging, and update tactics in real time.
This functional shift is supported by AI's high consistency and persuasion capabilities. In a cited Reddit experiment, persona-driven AI agents proved three to six times (300% to 600%) more effective at shifting opinions than human users. Operating continuously without emotional fatigue or performance variance, autonomous agents can take on sensitive commercial tasks such as negotiation, contract renewals, and outbound correspondence.
Enterprise Deployments Across Support, Sales, and Marketing (4:18)
Enterprise case studies illustrate the shift across different go-to-market functions. Klarna deployed autonomous agents to manage two-thirds of its customer support conversations, reducing support headcount while delivering 20% to 30% efficiency gains alongside faster response times.
In outbound operations, Siemens integrated real-time predictive AI to deliver live recommendations during sales calls, replacing gut-feel forecasting with data-driven pipeline accuracy and improved representative allocation. Similarly, Hitachi abandoned static marketing campaign schedules in favor of continuous behavioral tracking. In Hitachi's dynamic model, prospects enter and exit marketing flows based on real-time signals, accelerating deal velocity and decreasing drop-off.
Evolving Sales Roles and Phased Adoption (6:10)
Replacing static funnels with continuous AI flywheels alters revenue roles rather than simply eliminating them. Business Development Representatives (BDRs), who previously spent their days cold calling and manually logging data, are moving into supervisory roles where they design, prompt, and guide AI agents to prioritize accounts and initiate outreach. This removes routine friction and allows human reps to concentrate on high-touch strategy and relationship building.
To adopt these systems safely, organizations are advised to follow a three-phase framework: first, 'Layer' lightweight tools such as CRM meeting summaries onto existing software; second, 'Automate' end-to-end workflows like cross-platform lead routing; and third, 'Rewire' the entire go-to-market model around native AI architectures without legacy SaaS constraints.
Data Infrastructure Debates and Adoption Barriers (10:50)
Organizational hurdles complicate deployment. The primary cultural barrier is change management, particularly convincing high-performing sales reps who consistently meet quotas to abandon personal spreadsheets and adopt standardized CRM data practices required by AI systems. Additionally, the speaker stresses that 'before AI, you need BI,' recommending unified enterprise data lakes with open APIs and webhooks to provide agents with complete operational context.
However, this infrastructure requirement is debated. An alternative audience viewpoint warned that centralizing decades of legacy enterprise data can turn into a multi-year project that stalls immediate progress. Instead of waiting for full consolidation, teams can achieve immediate value by using lightweight automation tools like n8n, deploying agents in hours to solve localized bottlenecks such as researching past deals and account history ahead of prospect calls.
While AI-native CRM platforms are expected to mature and gradually replace legacy SaaS, enterprise leaders face a near-term strategic tension: investing in lengthy data unification initiatives versus deploying lightweight, agentic automations to address targeted workflow friction immediately.
## Background
本视频由 Ai4 于 2025 年 8 月 29 日发布在 YouTube 平台上。视频简介指出，本次演讲由 Pinegrove Venture Partners 的合伙人兼首席数据与 AI 官 Adam Nahari 以及 InterSystems 的首席技术专家 Kanishk Mittal 共同带来，旨在探讨人工智能如何将传统销售与营销 SaaS 平台重塑为动态且目标导向的工作系统。
## TL;DR
演讲深入探讨了 AI 智能体如何正在全面替代传统割裂的 SaaS 技术栈，将静态的销售漏斗转变为实时学习与自动优化的“AI 飞轮”。通过 Klarna、西门子和日立等知名企业的落地实例，展示了 AI 工作流在提升效率、精准跟进及自动化沟通方面的显著成效。演讲者提出了“叠加-自动化-重构”的渐进式实施三步法，并在问答环节针对数据湖建设、组织变革阻力及轻量化落地路径进行了深度剖析。
## Atmosphere & style
整场演讲语气专业、务实且充满洞察力，既有对前沿 AI 技术的远大构想，又紧扣企业实操中的痛点与挑战。演讲者的表达条理清晰、比喻生动（如将传统流程比作流水线，将 AI 比作 F1 进站），现场问答环节更展开了关于“重型数据湖建设”与“轻量低代码落地”的精彩观点碰撞。观看过程令人深切感受到企业级 GTM 架构正在发生的深刻范式转移以及从业者对技术变革的务实思考。
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=8gYch7PEHww&t=4s) 市场进入（GTM）技术栈正在经历范式转移，AI 不再是辅助工具，而是逐步替代传统 SaaS 架构成为工作流操作系统。
- [00:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=18s) 实验表明，基于 AI 的智能体在说服与改变观点方面的效果是普通人类用户的 3 到 6 倍，且具备全天候在线和极高的逻辑稳定性。
- [02:11](https://www.youtube.com/watch?v=8gYch7PEHww&t=131s) 传统 GTM 堆栈因数据孤岛和手工流程存在“人类瓶颈”，而 AI 原生系统通过多源实时数据拉取和动态优先级重排实现了实时策略调整。
- [04:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=258s) Klarna 部署 AI 智能体承接了 2/3 的客户咨询，整体效率提升了 20% 至 30%。
- [04:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=258s) 西门子与日立通过预测性 AI 和行为触发机制，分别实现了销售管道预测精准度的提升与线索流失率的降低。
- [06:10](https://www.youtube.com/watch?v=8gYch7PEHww&t=370s) 销售与营销模式正从单向、延迟高昂的“传统漏斗”演变为基于实时行为信号与自适应学习的“AI 飞轮”。
- [06:10](https://www.youtube.com/watch?v=8gYch7PEHww&t=370s) 团队角色从“执行者”提升为“编排者（Orchestrator）”，员工从狂打电话等重复性劳动中解放，专注于战略决策与智能体调优。
- [07:45](https://www.youtube.com/watch?v=8gYch7PEHww&t=465s) AI 落地建议采用“叠加层 -> 自动化 -> 重构”的渐进式三步法框架，切忌一上来就做过于宏大的全面改造。
- [09:56](https://www.youtube.com/watch?v=8gYch7PEHww&t=596s) 构建企业级数据湖并打通各系统 API 是实现全自主多智能体编排与自我学习的前置条件。
- [09:56](https://www.youtube.com/watch?v=8gYch7PEHww&t=596s) 推进 AI 变革最大的阻力常来自依赖传统习惯的顶尖销售人员，必须建立企业级数据规范以保障系统效能。
- [09:56](https://www.youtube.com/watch?v=8gYch7PEHww&t=596s) 采购任何营销与销售软件时，必须确保其具备开放 API、Webhooks 和数据导出所有权，以避免被单一平台锁定。
- [16:34](https://www.youtube.com/watch?v=8gYch7PEHww&t=994s) 在实践中，企业应平衡“远大愿景”与“当下切入点”，短期内可优先从会前客户调查、历史交易分析等微观场景入手验证 AI 价值。
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=8gYch7PEHww&t=4s) 销售与营销技术栈的范式转移：AI 正在全面替代传统 SaaS
演讲者首先指出了当前企业销售与营销技术栈面临的痛点：大多数企业的技术架构在过去 15 年间未发生根本改变，只是不断叠加更多独立工具来填补缝隙和消除数据孤岛。如今技术正在迎来关键转折点，人工智能不再只是堆栈中的又一个辅助工具，而是逐步演变为整个市场进入（Go-To-Market, GTM）架构与工作流的操作系统。
演讲对比了传统工具堆栈与新型自主智能体（Autonomous Agents）架构的差异：旧堆栈依赖孤立工具与人工记录，而智能体不仅能实时分析数据，还能直接做出决策并自动执行任务。更令人吃惊的是，智能体在处理许多传统上被认为极具“人类独特性”的非结构化工作时，展现出了超越人类的效率与表现。
- 传统 SaaS 堆栈多为数据孤岛，依赖大量手工录入与工具桥接。
- AI 智能体兼具决策与执行能力，能够无缝进行自主操作。
---
### [00:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=18s) 说服力转移：AI 在说服与沟通中的惊人表现
为了展示 AI 的强大效能，演讲者分享了今年早些时候研究人员在 Reddit 上进行的一项心理说服实验。在该实验中，AI 智能体被赋予特定虚拟角色，试图改变在线用户的观点。实验结果出乎所有人意料：AI 智能体改变他人观点的有效性是普通 Reddit 用户的 3 到 6 倍（即 300% 至 600% 的提升，而非区区 3% 至 6%）。
AI 的核心优势在于其极高的稳定性和持续性。人类销售或客服人员容易受到情绪影响或状态起伏，而 AI 能够 7x24 小时保持在线，始终输出基于客观数据的严密逻辑。演讲者强调：“如果 AI 能够在互联网上那些情绪激烈的争论中改变别人的想法，想象一下它在销售邮件或续约谈判中能发挥多大的作用。”
- Reddit 实验显示，AI 在改变人类观点上的效果比普通人类用户高出 3 到 6 倍。
- 相比于状态波动的传统人工，AI 具备全天候在线、基于数据输出说服逻辑的稳定性。
---
### [02:11](https://www.youtube.com/watch?v=8gYch7PEHww&t=131s) 传统 GTM 堆栈的局限性与 GenAI 的破局
目前的 GTM 堆栈存在严重的割裂问题：CRM 无法与营销平台无缝对话，BI 报表往往滞后数天，整体运作高度依赖离散的 SaaS 工具和人工数据录入。这种架构导致了严重的“人类瓶颈”，需要专人抽取报告、分析数据并手动启动营销活动。演讲者将这种状态形象地比喻为“踩着手刹驾驶跑车”，底层系统架构严重拖累了整体业务的运行速度。
而生成式 AI（GenAI）正是解开这一手刹的关键。传统的 SaaS 建立在固定、可预测的线性工作流之上（数据录入 -> 预定义规则营销 -> 固定的报表运行 -> 周期性分析调整）。AI 原生系统则颠覆了这一流程，通过持续抓取多源实时数据、自动清洗、动态重排线索优先级并实施个性化触达，实现了无间断的自我学习与实时策略调整。演讲者将其比作“工厂流水线”与“一级方程式赛车进站维修团队”的区别：前者只能死板执行既定步骤，而后者能够根据实时变化的环境条件做出秒级反应。
- 传统流程（流水线模式）：手工数据录入 -> 静态线索打分 -> 固定周期报表 -> 事后人工评估。
- AI 原生流程（F1 进站模式）：多源实时数据拉取 -> 自动清洗与动态优先级重排 -> 实时个性化触达 -> 持续自主学习优化。
---
### [04:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=258s) 企业落地实例：Klarna、西门子与日立的成功实践
AI 原生工作流已经在知名企业中产生了巨大商业价值，演讲者举出了三个典型的真实案例：
1. **Klarna**：此前依赖庞大的客服团队，响应速度与服务质量参差不齐，扩展业务只能依靠增加人力。引入 AI 智能体后，系统成功承接了三分之二（2/3）的客户咨询，实现了 20% 至 30% 的整体效率提升。这不仅大幅降低了客服成本，更提供了更快、更一致的客户体验。
2. **Siemens（西门子）**：传统的销售跟进依赖销售代表的个人直觉及静态 CRM 规则。西门子部署预测性 AI 后，能够在销售人员接听电话时，根据实时数据动态调整商机优先级，提供数据驱动的跟进建议，显著提升了销售管道（Pipeline）预测的准确性与人员分配效率。
3. **Hitachi（日立）**：日立彻底废除了传统固定起止时间的“营销活动（Campaign）”概念，改由 AI 实时监控用户行为数据，并自动触发相应的沟通流。这种敏捷适应机制减少了线索流失，大幅加快了业务推进速度。
- Klarna：智能体承接了 2/3 的客户对话，效率提升 20–30%，实现了成本控制与体验升级的双赢。
- 西门子：实时数据驱动的预测性 AI 取代直觉判断，大大提高了销售管道的精准度。
- 日立：打破固定营销活动周期，采用 AI 行为触发机制，有效降低了线索流失率。
---
### [06:10](https://www.youtube.com/watch?v=8gYch7PEHww&t=370s) 从传统漏斗到 AI 飞轮：岗位角色的重塑与演变
传统的销售营销模式是单向的“GTM 漏斗”：线索通过聊天机器人或冷外呼进入，经人工筛选后转交给销售团队，最终走向成交。这一过程环节冗长且充满延迟。AI 正在将其替换为高效循环的“飞轮（Flywheel）”架构：基于实时行为信号快速迭代定位、跨渠道自动进行 AI 推荐触达，并保持无间断的自适应学习。
随着底层技术的革新，团队内部的角色定位也迎来了根本性改变。以业务开发代表（BDR）为例，过去他们的工作是每天打几百上千个电话；而在未来，他们的核心职责将转变为“设计与调优 AI 智能体”。员工的角色正在从“执行者”提升为“编排者（Orchestrator）”，负责制定策略、监控质量并做最终裁决。例如，以往需要在 CRM 中繁琐筛选数据的操作，现在只需向智能体下达指令：“根据意图自动重排线索优先级。”这并未消灭人类的岗位，而是将人类从重复性劳动中解放出来，专注于高价值决策。
- 模式变革：从单向、延迟高昂的“传统漏斗”转向实时反馈、自适应学习的“AI 飞轮”。
- 角色转型：BDR 等岗位从重复劳动（如狂打电话）转向编排与管理 AI 智能体。
- 核心逻辑：AI 负责执行与优化，人类负责战略决策与质量监控。
---
### [07:45](https://www.youtube.com/watch?v=8gYch7PEHww&t=465s) AI 渐进式实施三步法框架与关键要点
为了避免在销售组织内部造成混乱，演讲者提出了一个循序渐进的“三阶段实施框架”：
1. **叠加层（Layer）**：在现有工具栈上叠加简单的 AI 智能体，尝试低风险、高回报的轻量应用（如 CRM 文本自动摘要、简单的线索优先级排序）。
2. **自动化（Automate）**：在团队建立信任后，构建端到端的 AI 工作流（如基于 AI 的潜在客户自动路由），实现高频重复工作的自动化。
3. **重构（Rewire）**：彻底摒弃传统旧工具的束缚，摆脱固有思维，重构整个市场进入（GTM）流程。
演讲者特别提醒，切忌一上来就尝试“煮干大海（Boil the ocean）”，而是应该从轻量场景切入，先验证价值并建立信任，再逐步扩大规模。演说最后总结道，传统工具正在加速淘汰，未来不是“AI 是否会替换你的工作流”，而是“你希望 AI 先替换哪一部分工作流”。
- 第一阶段（叠加）：引入轻量智能体（摘要、排序），快速取得低风险成效。
- 第二阶段（自动化）：打造端到端工作流（自动路由），解放日常人力。
- 第三阶段（重构）：摒弃旧框架束缚，重新定义整个 GTM 流程。
---
### [09:56](https://www.youtube.com/watch?v=8gYch7PEHww&t=596s) 现场 Q&A：技术落地、智能体演进与数据基座
在随后的问答环节中，观众针对具体落地细节与技术挑战提出了多个深度问题：
* **关于工具选型与编排（）**：
观众询问当前使用什么工具进行编排。演讲者指出，许多主流 CRM 和销售赋能工具（如他们目前使用的 Pega）已经敏锐意识到技术范式的转变，并开始在系统内内置原生智能体能力。这使得企业能直接利用现有 CRM 内的数据与风控流程来帮助销售代表省时。不过演讲者预测，未来市场上必定会孕育出完全“AI 优先（AI-First）”的新一代 CRM 产品。
* **关于智能体自我学习与多智能体编排（ /）**：
观众探讨了智能体何时能够摆脱人类指令，实现自我学习并编排其他智能体，以及如何利用企业旧数据构建“AI 大脑”。演讲者坦言当前模型仍需要人类监督，但实现这一目标的基石在于构建企业级的“数据湖（Data Lake）”。当前最大的瓶颈在于企业数据分散在数十个孤立系统中，把 API、工作流和数据整合在一起是实现全自主智能体的前置条件。
* **关于组织变革管理（Change Management）的挑战（）**：
演讲者指出最大的困难在于销售人员的心态转变。许多业绩顶尖的销售代表（Top Sales）往往拒绝使用 AI 工具，理由是“我不用 AI 也能完成业绩指标，为什么要打乱我的习惯？”演讲者回应称，企业 AI 战略不能依赖个人，必须建立企业级的数据治理标准。所有线索与过程数据都必须规范存入 CRM 而非私人 Excel 中，才能让整体 AI 系统发挥效用。
* **初创或传统企业快速切入 AI 的建议（）**：
针对想要低风险快速获益的企业，演讲者开玩笑说“在 AI 之前你需要先搞定 BI（哪怕 B 不在 A 前面）”。核心在于将分散的数据集中化。此外，演讲者建议企业在购买任何新软件（CRM、外呼工具、数据富化工具）时，必须确保其具备**开放 API（Open APIs）、Webhooks 和企业级数据湖导出权限**，确保企业自身对数据拥有绝对掌控权。
* **现场的不同观点碰撞（）**：
一位现场观众提出了不同见解，认为对于大型企业而言，将数十年积累的庞杂数据集中到一个数据湖可能耗时数年，这会导致企业久久无法迈出第一步。该观众建议采用更轻量的方法，直接基于 N8N 等低代码工作流工具，在数小时内围绕特定小场景快速落地智能体，而非等待漫长的数据集成工程。
- 编排工具：目前优先利用 Pega 等现有 CRM 的原生 AI 智能体，同时关注未来 AI-First 的原生 CRM 发展。
- 数据基座：搭建企业级数据湖并打通 API 是让 AI 智能体发挥全面效能的前提。
- 变革阻力：顶尖销售代表的习惯依赖是主要障碍，必须靠强制的企业级数据与流程规范来推进。
- 选型标准：采购任何工具时必须确保其具备开放 API、Webhooks 和数据导出所有权。
- 落地策略争议：演讲者主张夯实数据湖基础，而现场观众建议利用 N8N 等轻量工具快速实施小场景。
### [16:34](https://www.youtube.com/watch?v=8gYch7PEHww&t=994s) 销售与营销中 AI 落地策略：从具体场景微调到长远数据整合
在讨论人工智能如何彻底变革销售与营销流程时，构建长期的 AI 战略绝不能仅仅停留在高层概念上，而必须兼顾“远大愿景”与“当下切入点”。发言者明确指出，如果从长远策略的角度来看，最核心的根基依然在于“把分散在各处的数据集中整理到一个统一的平台中”。只有将数据打通并统一管理，AI 软件与智能体才能真正具备全局视角，发挥出跨业务流程的巨大潜力。
然而，在实际落地执行时，发言者特别强调了“小步快跑、循序渐进”的重要性，切忌一上来就铺开宏大而繁复的项目。企业应当从具体、微观且高频的业务场景入手。例如，销售代表在准备与某家客户公司开会之前，可以先利用 AI 工具进行快速质询：“请告诉我关于这家客户组织的更多背景信息”、“我们过去有没有尝试向他们销售过产品？”以及“当时交易最终没有成功的原因是什么？”通过这类小巧、具体的问答和辅助场景，员工能够快速体验到 AI 赋能带来的效率提升，进而为后续全方位的数据整合与系统升级奠定坚实的基础。
- **长期战略核心**：实现企业内部数据的统一与整合（Data Consolidation），将其集中托管于统一平台。
- **短期落地路径**：从具体的销售微观场景切入，如会前客户背景调查、历史交易失败原因分析等。
- **关键实践建议**：采用“小步快跑”策略，先在单个工作流步骤中验证 AI 价值，再逐步扩大应用范围。
## Notable quotes
> [00:18](https://www.youtube.com/watch?v=8gYch7PEHww&t=18s) 如果 AI 能够在互联网上那些情绪激烈的争论中改变别人的想法，想象一下它在销售邮件或续约谈判中能发挥多大的作用。
> [02:11](https://www.youtube.com/watch?v=8gYch7PEHww&t=131s) 踩着手刹驾驶跑车
> [02:11](https://www.youtube.com/watch?v=8gYch7PEHww&t=131s) 工厂流水线
> [02:11](https://www.youtube.com/watch?v=8gYch7PEHww&t=131s) 一级方程式赛车进站维修团队
> [06:10](https://www.youtube.com/watch?v=8gYch7PEHww&t=370s) 编排者（Orchestrator）
> [06:10](https://www.youtube.com/watch?v=8gYch7PEHww&t=370s) 根据意图自动重排线索优先级。
> [07:45](https://www.youtube.com/watch?v=8gYch7PEHww&t=465s) 煮干大海（Boil the ocean）
> [07:45](https://www.youtube.com/watch?v=8gYch7PEHww&t=465s) 你希望 AI 先替换哪一部分工作流
> [09:56](https://www.youtube.com/watch?v=8gYch7PEHww&t=596s) 我不用 AI 也能完成业绩指标，为什么要打乱我的习惯？
> [09:56](https://www.youtube.com/watch?v=8gYch7PEHww&t=596s) 在 AI 之前你需要先搞定 BI（哪怕 B 不在 A 前面）
> [16:34](https://www.youtube.com/watch?v=8gYch7PEHww&t=994s) 请告诉我关于这家客户组织的更多背景信息
> [16:34](https://www.youtube.com/watch?v=8gYch7PEHww&t=994s) 我们过去有没有尝试向他们销售过产品？
> [16:34](https://www.youtube.com/watch?v=8gYch7PEHww&t=994s) 当时交易最终没有成功的原因是什么？
## Mentioned
Adam Nahari, Kanishk Mittal, Pinegrove Venture Partners, InterSystems, Reddit, Klarna, Siemens, Hitachi, Pega, N8N
## Video description
Adam Nahari, Partner, Chief Data and AI Officer, Pinegrove Venture Partners
Kanishk Mittal, Lead Technical Specialist, InterSystems
Abstract
What happens when AI stops merely supporting tools - and starts replacing them? This talk explores The AI-Powered Future of Sales and Marketing, where AI is transforming legacy SaaS platforms into dynamic, goal-driven systems of work. The session will outline how this shift is already reshaping roles, collapsing traditional workflows, and enabling faster, smarter, and more adaptive go-to-market strategies.
Examples from companies like Klarna, Siemens, and Hitachi will illustrate how AI-driven workflows are delivering 20–30% efficiency gains by replacing static tools with intelligent agents. The talk will also provide practical frameworks for initial adoption, governance, and organizational alignment - offering a clear roadmap for Sales and Marketing leaders seeking to stay ahead of the curve as legacy tech stacks become increasingly outdated.
[Source](https://www.youtube.com/watch?v=8gYch7PEHww)
Source quotes
[0:18] 如果 AI 能够在互联网上那些情绪激烈的争论中改变别人的想法，想象一下它在销售邮件或续约谈判中能发挥多大的作用。
[2:11] 踩着手刹驾驶跑车
[2:11] 工厂流水线
[2:11] 一级方程式赛车进站维修团队
[6:10] 编排者（Orchestrator）
[6:10] 根据意图自动重排线索优先级。
[7:45] 煮干大海（Boil the ocean）
[7:45] 你希望 AI 先替换哪一部分工作流
[9:56] 我不用 AI 也能完成业绩指标，为什么要打乱我的习惯？
[9:56] 在 AI 之前你需要先搞定 BI（哪怕 B 不在 A 前面）
[16:34] 请告诉我关于这家客户组织的更多背景信息
[16:34] 我们过去有没有尝试向他们销售过产品？
[16:34] 当时交易最终没有成功的原因是什么？
Source quotes
[2:19] driving a sports car with the parking brakes on
[7:45] boil the ocean
[8:44] If AI can run 30% of your GTM process tomorrow, what's the first thing you'd want to hand over?
[13:00] before AI, you need BI
Original transcript
[0:04] Okay, thanks for bearing with us. I know
[0:06] this has been a very long conference. So
[0:08] I'm here to talk about the AI powered
[0:10] future of sales and marketing. So I'd
[0:13] like to start off with a high level
[0:15] overview of what I'm going to be
[0:16] discussing. And the idea here is or
[0:19] rather the question is what happens when
[0:21] AI stops supporting your sales and
[0:23] martekch stack but starts fundamentally
[0:26] replacing it. Right? And so think about
[0:28] your sales and marketing stack today,
[0:30] right? Most of it hasn't fundamentally
[0:32] changed in over 15 years. We just keep
[0:35] adding more and more tools to fill in
[0:36] the gaps and bridge the silos that we
[0:39] have. But we're at a turning point. AI
[0:42] isn't just another tool that you're
[0:43] adding to the stack. It's becoming the
[0:45] operating system for how your go to
[0:47] market stack actually works and how work
[0:49] gets done. So on the left you see the
[0:52] old stack where you have individual
[0:53] tools that are siloed. But on the right,
[0:56] and this is what we're envisioning, are
[0:57] autonomous agents. And these aren't
[0:59] systems that are logging information for
[1:01] you. They're acting on it without you as
[1:04] well. They decide on it, and they
[1:05] decision on it. And so, as it turns out,
[1:08] these tools are not just more efficient,
[1:11] but they're actually better at doing
[1:12] some of the things that we thought were
[1:15] very uniquely human.
[1:18] So, the persuasion shift, what happens
[1:20] when AI outperforms humans? Earlier this
[1:23] year, researchers ran an experiment on
[1:25] Reddit. And the goal of the experiment
[1:27] was to have AI agents act as personas
[1:30] and try and change people's minds
[1:32] online. And the results were actually
[1:34] very surprising. AI agents were three to
[1:37] six times more effective, not 3 to 6%, 3
[1:40] to six times more effective than Reddit
[1:42] users at changing other users' mind. And
[1:45] so, Whoops. I think we
[1:49] Yeah. And so what's really powerful here
[1:51] is the consistency. Humans can have a
[1:53] bad day, but AI doesn't. AI is always on
[1:56] 24/7 delivering a databacked argument.
[2:00] And if an AI can change someone's mind
[2:02] in a heated conversation online where
[2:04] things can get deeply personal, just
[2:06] imagine what it can do for you in a
[2:08] sales email or renewal conversation. But
[2:11] before we talk about the future, let's
[2:13] talk about what our current go to market
[2:15] stack looks like today.
[2:19] So here is the problem. Our current
[2:22] stack is fundamentally fragmented. Your
[2:25] CRM is not naturally talking to your
[2:27] marketing platform and your BI tool is
[2:29] almost always a few days behind. And so
[2:32] these systems are relying on a lot of
[2:34] disconnected SAS tools and manual data
[2:37] entry. And what that gives us is human
[2:40] bottlenecks. Someone has to pull
[2:42] reports, analyze the data, and then run
[2:44] the campaigns. It's kind of like driving
[2:46] a sports car, but with the parking
[2:48] brakes on. The way it's wired is holding
[2:51] it back.
[2:54] And so, Gen AI is essentially removing
[2:56] that parking brake for us.
[3:00] Traditional SAS is built around
[3:03] fixed predictable workflows. So, you
[3:06] start with data entry and setup. For
[3:07] example, you have sales reps manually
[3:09] entering data into your CRM or marketing
[3:11] tools. And then you run some sort of
[3:13] predefined workflow, typically a
[3:15] campaign based on your lead scoring
[3:16] rules. And then you have reports that
[3:18] execute in a fixed sequence. And then
[3:20] you review it and you adjust your cycle
[3:22] based on that. And so you feed in the
[3:24] data, you run the process and you review
[3:26] the results. What what an AI native
[3:29] system does is it flips that on its
[3:32] head, right? So the AI is continuously
[3:35] pulling real-time data from multiple
[3:37] sources. It's cleaning it up. The models
[3:39] are dynamically rep prioritizing the
[3:41] leads for you. You have personalized
[3:43] outreach and you can forecast outcomes.
[3:45] And then these agents are actually
[3:47] learning continuously. So there's no
[3:49] close date for your campaign after which
[3:51] someone sits down and analyzes the
[3:52] results and and then adapts. These
[3:55] agents are automatically monitoring and
[3:57] adjusting the strategy on the fly. And
[3:59] the difference between these two is is
[4:01] very stark, right? It's kind of like the
[4:03] difference between a factory assembly
[4:05] line and a Formula 1 pit crew, right?
[4:08] One is following a set predecided fixed
[4:11] recipe whereas the other is looking at
[4:13] what's happening and adapting on the fly
[4:15] reacting to changing conditions.
[4:18] And this isn't just theoretical
[4:19] companies have actually already started
[4:21] to see value from these. So three real
[4:24] world examples we'll be talking about
[4:25] here. The first one is Clarum. Clumo's
[4:28] customer support was primarily handled
[4:30] by a very large team of agents. Average
[4:33] response times varied, satisfaction
[4:34] varied, outcomes varied, and scaling
[4:36] meant hiring more support staff. Once
[4:39] they adopted AI, now you have agents
[4:41] handling two out of three customer
[4:42] chats. They've achieved 20 to 30%
[4:45] efficiency gains, reduce support
[4:47] headcount, but more than that, they have
[4:49] faster and more consistent responses.
[4:51] And this is what I'm talking about here.
[4:52] AI is not just about cost cutting. It's
[4:55] also about improving the value you're
[4:57] delivering to your customers. Seammens
[4:59] did something different. Sales
[5:01] orchestration at Seammens relied on
[5:03] manual prioritization by reps and static
[5:06] CRM rules and forecasting and OPC
[5:08] scoring was often based on gut feeling
[5:10] and incomplete data as it always is in
[5:11] most sales orgs. But now predictive AI
[5:14] can drive OP prioritization and what
[5:17] that means is sales teams are actually
[5:18] guided by data backed suggestions that
[5:21] update in real time when they're on a
[5:22] sales call and that value cannot be
[5:24] substituted. What this gives us overall
[5:27] is increased pipeline accuracy and
[5:29] smarter rep allocation. You're no longer
[5:31] trying to figure out what to do on the
[5:32] fly, but you have data back guidance
[5:34] which you can base those decisions on.
[5:36] And Hitachi did something completely
[5:38] different. They did away with campaigns
[5:40] altogether. So instead of having a fixed
[5:43] campaign which starts at a certain time
[5:45] and ends at a certain time, what they
[5:47] did was they had AI monitor real time
[5:49] behavioral data and trigger messaging
[5:52] and campaign workflows based on that.
[5:53] And so marketing at Hitachi is now
[5:55] adaptive, right? You're no longer in a
[5:58] certain stage or in a certain lead
[5:59] stage. you're just you can enter and
[6:01] exit their their flywheel which we'll
[6:03] talk about in a second at any given
[6:05] point in time and this reduces lead drop
[6:07] off and increases velocity.
[6:10] So traditionally the way we've thought
[6:12] about sales and marketing in general is
[6:15] your traditional GDM funnel right you
[6:18] have your fixed process where leads
[6:20] enter in through either chat bots or
[6:22] marketing campaign some sort of
[6:24] typically cold outreach they then get
[6:26] qualified and handed off to sales which
[6:28] itself has a lot of delays because it's
[6:29] a complicated process and then the sales
[6:32] teams work these ops to closure AI is
[6:36] replacing this with what I like to call
[6:38] a flywheel right so you have updates and
[6:41] targeting based on real-time behaviors
[6:42] and signals. You have AIdriven outreach
[6:45] and recommendations across channels and
[6:47] then there's continuous learning. So you
[6:49] don't need to wait for a campaign to
[6:50] finish, right? AI is looking at the data
[6:53] and adapting in real time. And when the
[6:56] tech is changing like this, it's
[6:57] inevitable that the roles across teams
[6:59] will change too. So until recently being
[7:03] a BDR meant making hundreds or thousands
[7:05] of calls and just hammering the phones,
[7:08] what it might mean going forward is
[7:10] designing and tuning AI agents, right?
[7:12] The tasks are shifting from execution to
[7:16] orchestration, deciding what AI should
[7:18] do, monitoring quality, and making
[7:21] strategic calls. The three boxes here
[7:23] are great examples, right? Instead of
[7:25] manually digging through your CRM and
[7:27] prioritizing leads and then sort of
[7:29] offloading them into some sort of
[7:30] outreach tool, you can just go to your
[7:32] CRM, go to an agent and say, "Prioritize
[7:34] leads by intent." And this doesn't
[7:37] eliminate the human role in any way. It
[7:39] just elevates it. It gives us all time
[7:41] back and allows us to focus on what we
[7:43] need to do uniquely.
[7:45] So, how do you get started without
[7:47] actually creating chaos within your
[7:49] sales organization?
[7:50] So there's a few different ways, but I
[7:53] like a three three-phase framework. The
[7:55] first one is to layer AI onto your
[7:57] existing tools, right? You take your
[8:00] existing stack, you add a couple of
[8:02] simple agents, lowrisk, easy wins.
[8:04] Summarization in your CRM or lead
[8:06] prioritization could be some of these.
[8:08] And then once your teams are comfortable
[8:10] with this, you start to automate. You
[8:12] build end-to-end AI workflows like
[8:14] AIdriven prospect routing. And you make
[8:16] sure your team can start to automate
[8:18] some of the work they're doing. And then
[8:20] once you've done this and and you've
[8:22] proven the value so to say, that's when
[8:24] you rewire. You fundamentally imagine
[8:26] reimagine what your go to market process
[8:28] looks like without the constraints of
[8:30] any of your old tools. And the trick
[8:32] here is to avoid boiling the ocean. You
[8:35] need to start small, build credibility,
[8:38] show value, and then you scale.
[8:44] So what are the takeaways and how you
[8:46] can how how can you sort of get started
[8:47] on this process? I think the one thing
[8:50] we all need to realize is that legacy
[8:52] tools are on their way out. And the
[8:54] companies that are able to figure out an
[8:56] AI native go to market strategy will be
[8:59] the ones that survive. And AI isn't just
[9:02] supporting, it's persuading, adapting,
[9:05] and optimizing in ways humans just
[9:07] cannot match at scale. So the question
[9:09] is not will AI replace parts of your
[9:12] workflow. It's which parts of your
[9:14] workflow do you want AI to replace? And
[9:17] I'd like to end with an open question.
[9:19] If AI can run 30% of your GTM process
[9:22] tomorrow, what's the first thing you'd
[9:25] want to hand over?
[9:27] Thank you.
[9:29] [Applause]
[9:37] Do we have any questions?
[9:47] Okay, perfect.
[9:56] >> So, sorry, can someone get the gentleman
[9:58] a mic, please? Um,
[10:00] >> whatolation
[10:01] are you using? Mhm.
[10:05] >> Are you seeing other tools or
[10:08] >> Well, so for some of the orchestration
[10:10] tasks, what's interesting is that the
[10:11] CRM out there or the sales enablement
[10:13] tools themselves have realized that this
[10:14] is a paradigm shift that's coming and so
[10:16] they've started offering agents natively
[10:19] that do that. So we use Pega as a CRM
[10:21] and they've come out with some pretty
[10:23] interesting agent capabilities that
[10:25] embed into the CRM itself. And so what
[10:27] that allows us to do is leverage all the
[10:30] data that's in there, leverage our
[10:31] process and our controls to help reps
[10:34] kind of save time. Uh but I think what's
[10:36] going to be interesting is that I I
[10:38] think in the near future there will
[10:40] emerge an AI first suite of CRM tools
[10:43] and I think we're just sort of waiting
[10:44] to see how that space evolves before we
[10:46] sort of commit to one.
[10:50] >> Yes. So if a human's orchestrating the
[10:53] agent given instruction
[10:56] workflow at what point does the human uh
[11:00] you know not need the human agent can
[11:02] actually learn and also orchestrate
[11:04] other agents within that workflow how do
[11:06] you otherwise agents know it just takes
[11:09] instructions right so how do how do you
[11:10] create a learning agent that can
[11:12] orchestrate other agents
[11:13] >> I think that's a very good question and
[11:15] I don't think we have an answer for that
[11:16] right now I think right now what we're
[11:18] trying to do is just give reps back time
[11:20] improve value. And as I said, you know,
[11:22] that Reddit study really hit home for me
[11:24] because it it told me that it's not just
[11:26] about automating tasks or saving time,
[11:28] it's about doing them better. And I
[11:30] think I think as models evolve and as
[11:32] best practices evolve, as people figure
[11:34] out what prompts are really working, I
[11:36] think we can get to a point where agents
[11:37] can kind of figure out what they need to
[11:38] do automatically. But right now, if
[11:40] you're trying to use the current
[11:41] existing models and frameworks, I do
[11:43] think there's a need for supervision and
[11:44] there likely will be for the near
[11:46] future.
[11:47] >> So I'll follow up on that. So what if
[11:49] the HP
[11:51] a new AI brain multi loaded the company
[11:53] loads in all their legacy data right
[11:56] all their APIs and everything. So now
[11:58] you have this new AI brain and you're
[12:00] given a instructions. Can it bounce off
[12:03] that new brain and also access you know
[12:06] multiple lms to figure out what the next
[12:08] step process might be?
[12:10] >> Sorry. So what was the question? So uh
[12:13] if you given nation instruction
[12:16] >> and what if you took uh all your legacy
[12:19] marketing
[12:20] >> and all the workflows within market you
[12:22] loaded that into a custom
[12:24] >> right
[12:24] >> kind of a new center fority
[12:27] >> and the agent then if it gets stuck it
[12:30] can go to the operating system well
[12:31] here's what happened last time and can
[12:33] it actually learn and if it doesn't
[12:34] learn access outside LLMs to bring those
[12:39] uh get those questions actually start
[12:40] learning on it
[12:42] Yeah. So I think that's that's very
[12:44] interesting and I think the first thing
[12:45] you need to do to even start to leverage
[12:47] something like that is to have an
[12:49] enterprise level data lake and that's
[12:51] the effort we're focusing on right now.
[12:52] Right. As we said our our GTM stack has
[12:55] been very fragmented and data is siloed
[12:58] in about a dozen or two dozen different
[12:59] systems. So pulling in all that data as
[13:01] you said building out the APIs the
[13:03] workflows consolidating that is the
[13:05] process we're in right now. And we do
[13:06] hope to kind of have a model sit on top
[13:08] of that to do exactly what you're
[13:10] saying. That's the end goal.
[13:12] >> Yeah.
[13:14] >> Yes.
[13:15] >> I just wanted to ask what are what is
[13:17] the most common problem that you're
[13:19] seeing when
[13:21] models?
[13:23] >> Are you are you asking what in change
[13:26] management is the problem? Because
[13:27] change management is the problem that I
[13:29] think most dogs are facing. But I think
[13:31] I I I think it's about I I think this is
[13:34] I'll put it in sales people language,
[13:36] right? um what's good for the gander may
[13:40] not always be good for the goose, right?
[13:42] And so if you talk about individual
[13:44] sales reps, we we have a lot of
[13:46] phenomenal people who always hit their
[13:47] coder, right? And when we try to get
[13:50] them to sort of use our process and use
[13:52] these AI tools, what they come back and
[13:53] say is, "Hey, I'm hitting my coder,
[13:55] right? Why are you trying to disrupt my
[13:56] workflow?" Right? And there's no easy
[13:59] answer to that. But the overwhelming
[14:01] point we're trying to make is that it's
[14:02] not about an individual. If you want
[14:04] good good AI in your organization, you
[14:07] have to enforce enterprise level
[14:08] standards and you have to make sure all
[14:10] the sales reps are following the
[14:11] process, putting the data in the CRM,
[14:13] not in a separate Excel sheet. And I
[14:15] think that's kind of where we're
[14:16] figuring out how we need to pitch this
[14:18] change because we have a lot of people
[14:19] who've sold successfully and they
[14:21] personally don't feel the need for AI,
[14:23] but I think seeing what we've seen at
[14:25] this conference, they will eventually
[14:27] need to start using AI if they want to
[14:29] be competitive in the space. And so
[14:30] convincing them of that personally has
[14:32] been a pretty hard challenge.
[14:35] Yes.
[14:36] >> I'm just curious for some of the
[14:37] companies that are sort of starting out
[14:39] that may not be as advancing AI, they
[14:43] want some really basic integration but
[14:47] quick win, lowriskation.
[14:49] >> What are some of recommendations? Cuz
[14:51] there's so much out there.
[14:52] >> Yeah.
[14:53] >> And uh you know, you really don't know
[14:54] how to start especially when it comes to
[14:56] agents. Mhm.
[14:57] >> What are some of the entrylevel
[15:00] um uh
[15:01] >> prerequisites
[15:03] >> actually enter?
[15:05] >> So I think before AI you need BI even
[15:07] though B doesn't come before A. I'm
[15:09] sorry about that joke. Um and so what
[15:11] that means is you need to make sure all
[15:12] the data that you have within an org is
[15:15] consolidated in one place and that would
[15:17] be the starting point for AI. And so
[15:20] this is something I'm working on
[15:21] separately as an article and this is not
[15:23] something a lot of people think of but
[15:24] when you're buying a tool what you need
[15:26] to make sure is that that tool has open
[15:28] APIs that tool has web hooks and that
[15:30] tool has data lake level enterprise
[15:32] access because you need to pull that
[15:34] data in from their system to yours if
[15:37] you ever want AI to be able to leverage
[15:39] that data. And so whatever tool you
[15:40] adopt, whether it's a CRM or an
[15:42] enrichment tool or an outreach tool or
[15:44] an ESB, you need to make sure that you
[15:47] are owning the data that you are putting
[15:48] into their system because eventually
[15:50] when you're trying to build agents,
[15:51] that's the data they're going to need to
[15:53] use.
[15:59] >> Yes.
[15:59] >> If I may, I'm not I'm not sure that I
[16:01] agree that the absolutely to put all of
[16:04] your data into one place.
[16:06] >> I think depending on the size of the or
[16:08] I mean There could be decades worth of
[16:11] dis desperate data in different places
[16:12] and the amount of effort required to get
[16:14] the data into one place could be such
[16:16] could be a yearslong undertaking such
[16:19] that you never really truly get started.
[16:22] >> What I might offer is
[16:23] >> start small smart start with a small use
[16:26] case
[16:27] >> and tools like an N8 be up and like
[16:31] running with an agent in hours as
[16:33] opposed to a year'slong effort of
[16:34] bringing data into one place on it.
[16:37] >> I I agree with that. I mean, if you're
[16:38] asking about a long-term strategy, I
[16:40] think this is what I would recommend.
[16:41] But as I said, you do want to start
[16:42] small, right? You want to start with,
[16:44] oh, you know, tell me more about this
[16:45] org, right? I'm taking a call with
[16:47] someone from this org. Have we have we
[16:49] tried to sell to them in the past,
[16:50] right? Why has that deal not worked out,
[16:52] right? And so, you can start small with
[16:53] individual steps, but if you're talking
[16:55] about long-term strategy, I do think you
[16:56] have to start with consolidating data.
