Cox Automotive Cuts Vehicle Repair Estimates to 30 Minutes With AI
Ai4
Enterprise AI shifts from experimentation to bottom-line ROI when organizations invert their strategy to become AI-first and deploy autonomous agents to eliminate frontline administrative overhead.
## Background
Published on YouTube by Ai4 on August 27, 2025, this presentation features Jodi Blomberg, VP of AI and Machine Learning at Cox Automotive, alongside technology leader MJ. According to the video description, the session explores how Cox Automotive takes a pragmatically aggressive approach to generative and agentic AI, leveraging the industry's largest automotive dataset to reshape consumer experiences and dealership operations.
## TL;DR
Cox Automotive leaders outline the enterprise's transformation from being 'data first, AI enabled' to an 'AI first, data differentiated' organization capitalizing on the emerging wave of agentic systems. Following AWS re:Invent, the team established twin accelerators that shipped 27 proofs of concept in six months, drastically reducing frontline vehicle repair estimation times from up to 48 hours to just 30 minutes. The strategy balances everyday enterprise productivity gains with high-impact, top-down bets, supported by agile governance scaffolding and an approved 'yes list' to empower decentralized innovation. Through hands-on executive immersion using Claude for Enterprise and proprietary data integrations like DRIVEQ, the company demonstrates how rapid, pragmatic experimentation translates directly into operational efficiency and recurring revenue.
## Atmosphere & style
The session is energetic, pragmatic, and paced with the urgency of seasoned operators who view agentic AI as an epochal shift comparable to the Industrial Revolution. Jodi and MJ share an effortless, conversational rapport, pairing high-octane automotive metaphors with grounded stories of riding along in field service trucks to witness technician friction firsthand. The tone feels like an authentic insider playbook for enterprise leadership—unflinchingly candid about token economics, corporate inertia, and security governance, yet deeply invigorating and bias-to-action driven.
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=ITPb0DipNqg&t=4s) The emergence of agentic AI represents a tectonic shift of historical magnitude, broader even than prior technological cycles like the Industrial Revolution.
- [02:50](https://www.youtube.com/watch?v=ITPb0DipNqg&t=170s) Cox Automotive established two dedicated agentic accelerators following AWS re:Invent, delivering 27 completed proofs of concept within six months.
- [02:50](https://www.youtube.com/watch?v=ITPb0DipNqg&t=170s) Deploying agentic workflows to automate administrative lookups and invoicing for field technicians reduced vehicle inspection and estimation turnaround times from 8–48 hours down to 30 minutes.
- [06:39](https://www.youtube.com/watch?v=ITPb0DipNqg&t=399s) Cox Automotive strategically inverted its organizational identity from being 'data first, AI enabled' to 'AI first, data differentiated,' placing AI architecture at the core of value delivery.
- [09:27](https://www.youtube.com/watch?v=ITPb0DipNqg&t=567s) Successful adoption balances broad, bottom-up everyday AI productivity gains across tools like GitHub Copilot with targeted, top-down bets on high-value customer workflows.
- [09:27](https://www.youtube.com/watch?v=ITPb0DipNqg&t=567s) Dedicated, focused AI teams dramatically accelerate development velocity, shipping working builds in days or weeks compared to traditional multi-month scrum timelines.
- [11:41](https://www.youtube.com/watch?v=ITPb0DipNqg&t=701s) Central AI teams must transition from isolated pioneers into organizational enablers, providing lightweight scaffolding and expanding a vetted 'yes list' of pre-approved tools.
- [11:41](https://www.youtube.com/watch?v=ITPb0DipNqg&t=701s) Managing token unit economics and operating cost models through targeted proofs-of-concept is critical to ensure generative AI solutions remain commercially viable.
- [14:35](https://www.youtube.com/watch?v=ITPb0DipNqg&t=875s) Driving enterprise buy-in requires hands-on executive enablement, exemplified by four-hour immersive workshops in Claude for Enterprise for the C-suite.
- [18:42](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1122s) Organizations should adopt a 'pragmatically aggressive' posture, driving innovation proactively from any level without waiting for centralized permission.
- [19:48](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1188s) Connecting generative LLMs to proprietary consumer intelligence platforms like DRIVEQ automates hyper-personalized dealership communications, lifting response rates and vehicle sales conversions.
- [20:51](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1251s) Achieving sustainable AI velocity requires starting immediately with low-risk quick wins and implementing human-in-the-loop validation to responsibly manage risk.
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=ITPb0DipNqg&t=4s) The AI Highway and the Tectonic Shift to Agentic AI
Jodi Blomberg introduces herself as the Vice President of AI and Machine Learning at Cox Automotive—a roughly $10 billion enterprise spanning software products and tech-enabled physical services. Reflecting on the fast-paced nature of their work under MJ's leadership, Jodi frames the moment in a fitting automotive metaphor: when the AI highway opened, leadership handed the team the keys to a Porsche and said, "You better step on it." Jodi turns to MJ, noting her extensive tenure leading data and software engineering teams, to ask what fundamentally distinguishes the current AI wave from prior technological cycles.
```mermaid
flowchart TD
accTitle: Evolution of Cox Automotive's Tech and AI Strategy
subgraph S1["Foundation"]
A["70+ Acquisitions & Complex Stacks"] --> B["7-Year Cloud Migration to AWS"]
end
subgraph S2["Generative Wave"]
B --> C["Generative AI (~3 Years Ago)<br/>Internal Efficiency & Early Value"]
end
subgraph S3["The Agentic Paradigm"]
C --> D["AWS re:Invent Spark"]
D --> E["Tectonic Plate Shift<br/>Agentic AI & Autonomy"]
end
```
MJ reflects on Cox Automotive's deliberate journey, explaining that over the previous seven years, the company methodically migrated to the cloud across more than 70 distinct corporate acquisitions, each bringing its own bespoke tech stack and operational complexity. While the arrival of generative AI roughly three years ago allowed them to lean in and deliver early value, MJ emphasizes that the emergence of agentic AI represents a completely different magnitude of change. Coming back from AWS re:Invent in November, the leadership team realized they were witnessing a "tectonic plate shift" on par with the invention of the wheel. MJ underscores that having worked through eras before email and through the advent of the commercial internet, the agentic wave is orders of magnitude broader even than the Industrial Revolution, requiring leaders across all industries to pay immediate attention.
[02:50](https://www.youtube.com/watch?v=ITPb0DipNqg&t=170s) Launching the Twin Accelerators and Field-Worker Agentic Solutions
On the four-hour flight home from re:Invent, the Cox Automotive team mapped out an aggressive plan to discover what was real and what was hype by establishing two dedicated agentic accelerators:
```mermaid
flowchart LR
accTitle: Cox Automotive Twin Agentic Accelerators
subgraph Accelerators["Twin Agentic Accelerators"]
direction TB
A1["Accelerator 1: Business Functional Areas<br/>Efficiency, Effectiveness & Field Operations"]
A2["Accelerator 2: Next-Gen Platform<br/>Future of Code Creation & Software Architecture"]
end
Accelerators --> B["27 Completed POCs in First 6 Months"]
```
Jodi Blomberg’s accelerator team was immediately flooded with internal intake requests and had to winnow down projects to those with a viable business case where AI was a genuine solution rather than an imagined "magic wand." Operating with a lean team, they completed 27 proof-of-concept (POC) implementations in the first six months. While standard use cases included document comprehension, content generation, and business process automation, the most significant breakthroughs occurred in Cox's tech-enabled field operations. The company employs thousands of frontline workers who repair commercial trucks on the highway, perform preventative maintenance, paint cars, repair dents, and manage auction lanes.
To solve real frontline friction, the accelerator team rode along in service trucks with field technicians. They discovered technicians were burdened with complex administrative overhead: cross-referencing regional labor rates, looking up obscure part numbers, sourcing backup vendors when inventory ran out, and assembling multi-line estimates for customer approval before any physical repair could begin. By handing this administrative lookup and invoicing burden over to an agentic workflow, technicians could focus on the physical repairs that AI cannot perform.
MJ adds dramatic context regarding the measurable business impact: for a major fleet customer that requires frequent residential visits, vehicle inspection and estimation turnaround times that previously took anywhere from eight hours to two full days per vehicle were slashed down to just 30 minutes per vehicle, sending technician job satisfaction off the charts.
- **Accelerator Output:** Completed 27 proofs of concept within the first six months of operation.
- **Field Estimation Time Reduction:** Cut vehicle inspection and estimation turnaround from 8–48 hours down to 30 minutes per vehicle.
- **Frontline Focus:** Automated part matching, regional labor pricing, and invoice approval workflows for field technicians repairing heavy trucks and auction vehicles.
[06:39](https://www.youtube.com/watch?v=ITPb0DipNqg&t=399s) Strategic Transformation: Shifting from Data-First to AI-First
Jodi and MJ address the dual mandate faced by enterprise leaders: optimizing internal business operations for cost efficiency while building net-new commercial AI products that drive top-line revenue. MJ highlights that Cox Automotive has fundamentally inverted its long-standing corporate identity from being a "data first, AI enabled" organization to becoming "AI first, data differentiated." In a $10 billion automotive software and services company, proprietary data remains the core competitive differentiator, but AI architecture now dictates how that value is delivered.
MJ, whose leadership purview unifies product, technology, data, corporate IT, and security, emphasizes that executing this shift requires full executive alignment. The CEO and executive peers recognize that AI adoption cannot be treated as a segregated technology project or an IT-led experiment; it represents a comprehensive, whole-business transformation that reshapes every corporate role, go-to-market strategy, margin structure, and top-line revenue stream. To support this across the ranks, Cox has instituted a bottoms-up enablement strategy—deploying tools like GitHub Copilot and cloud coding assistants to every software engineer, while embedding AI capabilities directly into everyday workplace productivity platforms like Microsoft Teams to ensure human creativity remains central to the loop.
[09:27](https://www.youtube.com/watch?v=ITPb0DipNqg&t=567s) Balancing Everyday AI with Hyper-Focused Transformation
Organizations must cultivate AI adoption on two distinct levels: broad everyday productivity gains across existing ecosystems and targeted, top-down bets that radically disrupt how the company goes to market. Within common software stacks—whether ServiceNow, Salesforce, or other enterprise platforms—everyday AI features should be actively turned on to lift baseline productivity across the workforce. Using a baseball analogy, these widespread tools generate a continuous stream of "singles, doubles, and hopefully some triples," steadily elevating collective organizational capability.
At the same time, executive leadership must identify one or two needle-moving priorities per core customer segment to genuinely differentiate their offerings and "change the conversation." In the automotive sector, Cox Automotive works with nearly every dealership nationally and extensively abroad. A prime target for disruption is the dealership service drive: reimagining how repair technicians work and transforming vehicle maintenance through real-time data. By setting up dedicated AI accelerators focused on these top two or three initiatives, development velocity accelerates by orders of magnitude. Teams are shipping functional builds in four days rather than months; in one notable instance, a two-person team delivered a complete project in five weeks that previously would have required multiple scrum teams working for six months. This bifurcated approach pairs bottom-up capability lifting with top-down hyper-focus across both customer-facing solutions and core internal functions like HR, finance, and contact/operations centers.
```mermaid
flowchart TD
accTitle: Dual-Track AI Strategy at Cox Automotive
subgraph BottomUp["Bottom-Up: Everyday AI"]
A["Enterprise Ecosystems<br/>(ServiceNow, Salesforce, Copilots)"]
B["Baseline Productivity Gains<br/>('Singles, Doubles, Triples')"]
A --> B
end
subgraph TopDown["Top-Down: AI Accelerator"]
C["High-Value Core Segments<br/>(e.g., Dealership Service Drive)"]
D["Rapid POCs & Delivery<br/>(Weeks instead of 6 Months)"]
C --> D
end
subgraph Foundation["Enterprise Scaffolding & Governance"]
E["Safe Guardrails ('The Yes List')"]
F["Platform Scale & Enablement"]
E --> F
end
B --> Foundation
D --> Foundation
```
- **Speed of execution**: Dedicated accelerator teams build working solutions in as little as 4 days, or achieve in 5 weeks with 2 people what previously required multiple scrum teams over 6 months.
- **Strategic focus**: Target 1–2 transformational differentiators per customer segment (such as technician workflows in dealership service drives) and functional department.
---
[11:41](https://www.youtube.com/watch?v=ITPb0DipNqg&t=701s) Moving from Pioneer to Enabler: Scaffolding and the "Yes List"
As initial quick wins—such as distributing AI code-assist copilots to every software engineer across the enterprise—take hold, the primary operational challenge shifts toward scaling safely without bottlenecking the central AI team. To prevent the dedicated AI accelerator from becoming an organizational constraint, the AI leadership team must evolve from being isolated "pioneers" into broad enterprise "enablers." This requires constructing lightweight, agile structural support—termed "scaffolding" rather than rigid "bricks," given how fast underlying models evolve—that allows team members across the company to build independently within pre-approved parameters.
```
"The yes list is like: what are the places where I can let you run with scissors, except I took the scissors away."
```
This governance framework expands what Jodi Blomberg terms the "yes list"—clearing safe zones pre-vetted by the Chief Information Security Officer (CISO) and AI leadership so distributed teams can experiment without bureaucratic friction. While this deliberate decentralization may introduce temporary "AI sprawl," it is actively encouraged to build momentum. To operationalize this shift, Cox brought together more than 70 Assistant Vice Presidents (AVPs) and senior leaders for an intensive alignment summit. The company currently has five large-scale proofs-of-concept running to iron out the ambiguous operating economics of generative AI, particularly managing token unit economics so applications do not become more expensive to run than what customers will pay. Rather than forcing hesitant teams to adopt AI, leadership focuses on creating high-value scaffolding that tempts stragglers into the ecosystem while empowering eager builders.
- **Expanding the 'Yes List'**: Establishing pre-approved architecture, tooling, and CISO-compliant environments so employees can build safely without centralized gatekeeping.
- **POC economics**: Five major active proofs-of-concept are actively testing cost sustainability, token usage swings, support models, and commercial go-to-market strategies.
- **Leadership alignment**: Gathered 70+ AVPs to establish enterprise scaffolding and shift from central development to widespread enablement.
---
[14:35](https://www.youtube.com/watch?v=ITPb0DipNqg&t=875s) Driving Executive Buy-in, Workforce Transformation, and Anthropic Partnerships
For organizations starting earlier on the maturity curve or operating in highly regulated, risk-averse environments (such as financial services), driving adoption requires proactive technical experimentation that delivers undeniable value before asking for formal permission. AI transformation cannot succeed as an isolated functional initiative; it requires full CEO-level sponsorship. Technology leaders must take initiative—such as standing up accelerators immediately after returning from major industry conferences like AWS re:Invent—and use the tangible prototypes to demonstrate revenue growth, margin defense, and disruption risks directly to executive peers.
```
"It is every role in the company should change... We've set expectations of what we think productivity changes should look like, and we are bringing folks on that journey."
```
This journey represents a fundamental workforce transformation where tool adoption shifts from an optional pull to an active baseline expectation across engineering and operational roles. To educate leadership, Cox developed executive "starter kits" to demystify generative AI for leaders who were hesitant to admit their lack of familiarity. Building on their work across leading foundation models, Cox leaned heavily into Anthropic and Claude for Enterprise. Leadership hosted intensive, four-hour interactive workshops for the C-suite:
- **Revenue leaders session**: Focused on direct monetization strategies and agentic workflows alongside Anthropic's leadership team.
- **Functional C-suite session**: Hands-on onboarding where executives received active user accounts and spent four hours directly building and testing use cases in Claude for Enterprise, completely transforming executive demand overnight.
- **Proactive initiation**: Start experimenting and proving value quickly without waiting for full corporate mandates to generate C-suite buy-in.
- **Executive enablement**: Executed 4-hour hands-on Claude for Enterprise bootcamps for C-suite and top functional leaders to demonstrate monetization and agentic capabilities firsthand.
- **Workforce expectations**: Established concrete productivity benchmarks and tool requirements across engineering organizations.
[18:42](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1122s) Leading from Anywhere and Being Pragmatically Aggressive
Taking full advantage of artificial intelligence requires an organizational posture described as being "pragmatically aggressive"—a deliberate balance between moving quickly and exercising discipline, avoiding a recklessness likened to "running with scissors." By remaining assertive, intentional, and selective about high-priority areas, teams accelerate their organizational learning loops. Crucially, driving AI adoption does not require formal executive authority or a sprawling team; individuals across all corporate levels possess the ability to influence peers, demonstrate what is technically feasible, and spearhead meaningful business outcomes from their current positions.
```mermaid
flowchart TD
accTitle: High-Impact AI Adoption Strategy
subgraph Mindset["Leadership Mindset"]
A["Lead from Anywhere"] --> B["Pragmatically Aggressive Action"]
end
subgraph Implementation["High-Impact / Low-Risk Entry Points"]
B --> C["Customer Touchpoints"]
B --> D["LLM Access Layer on Internal Data"]
B --> E["Automated Content Creation"]
end
```
For organizations wondering where to begin, the most effective roadmap starts with high-impact, low-risk opportunities. The most immediate, accessible applications center on streamlining customer contact points, layering large language models over existing datasets to make information readily searchable and intuitive, and automating content generation workflows.
[19:48](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1188s) High-Impact Generative AI: Personalizing Dealer Communications with DRIVEQ
A prime demonstration of pairing generative AI with enterprise data is Cox Automotive's work optimizing dealership outreach. Dealerships routinely struggle with writing timely, effective communications for prospective buyers. Cox Automotive bridged this gap by connecting large language models to their proprietary consumer intelligence platform, DRIVEQ.
```mermaid
flowchart LR
accTitle: DRIVEQ-Powered Hyper-Personalized Dealer Outreach
subgraph Data["Data Layer"]
A["DRIVEQ Platform<br>Proprietary Consumer Data"]
end
subgraph Engine["AI Generation"]
B["Generative LLM Engine"]
end
subgraph Distribution["Outreach & Conversion"]
C["Hyper-Personalized Content<br>Multi-Channel Modalities"]
D["Increased Response Rates"]
E["Vehicle Appointments & Sales"]
end
A --> B
B --> C
C --> D
D --> E
```
The system automatically drafts hyper-personalized messages on behalf of the dealer across whatever communication channel the customer prefers. Because the copy directly incorporates proprietary data regarding shopper behaviors and intent, the outreach is exceptionally relevant and timely without burdening dealership staff with copywriting. This straightforward application of generative AI has driven dramatic gains in both response rates and downstream sales conversions—turning inquiries into scheduled showroom appointments and completed vehicle purchases—while generating substantial monthly recurring revenue (MRR).
- **Data-Driven Foundations**: Success hinges entirely on clean, structured proprietary data ("have your data house in order").
- **Commercial Impact**: Translates basic generative AI capabilities into demonstrable monthly recurring revenue and higher vehicle sell rates.
[20:51](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1251s) Key Takeaways: Data, Quick Wins, and Balancing Risk with Speed
Wrapping up the session, the presentation concludes with actionable guidance for organizations embarking on their AI journeys. The foundational imperative is putting the organization in motion immediately rather than waiting for ideal conditions. Finding early, low-friction quick wins generates tangible proof points, which in turn unlock project funding and organizational buy-in.
Teams must plan for imperfection and treat initial challenges as necessary steps in the learning curve. When balancing innovation speed against risk—a topic echoing recent discussions in fintech and financial services—incorporating human-in-the-loop validation provides the safety net needed to deploy models responsibly while enabling continuous refinement.
- **Put Yourself in Motion**: Start small, act immediately, and empower teams to experiment.
- **Secure Buy-in via Quick Wins**: Delivering early business value creates momentum and secures sustained initiative funding.
- **Manage Risk Dynamically**: Use human-in-the-loop mechanisms in sensitive domains to maintain quality control without sacrificing velocity.
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=ITPb0DipNqg&t=4s) You better step on it. — MJ
> [06:39](https://www.youtube.com/watch?v=ITPb0DipNqg&t=399s) data first, AI enabled — MJ
> [06:39](https://www.youtube.com/watch?v=ITPb0DipNqg&t=399s) AI first, data differentiated — MJ
> [09:27](https://www.youtube.com/watch?v=ITPb0DipNqg&t=567s) singles, doubles, and hopefully some triples
> [09:27](https://www.youtube.com/watch?v=ITPb0DipNqg&t=567s) change the conversation
> [11:41](https://www.youtube.com/watch?v=ITPb0DipNqg&t=701s) The yes list is like: what are the places where I can let you run with scissors, except I took the scissors away. — Jodi Blomberg
> [14:35](https://www.youtube.com/watch?v=ITPb0DipNqg&t=875s) It is every role in the company should change... We've set expectations of what we think productivity changes should look like, and we are bringing folks on that journey. — MJ
> [18:42](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1122s) running with scissors
> [19:48](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1188s) have your data house in order
## Mentioned
Jodi Blomberg, MJ, Cox Automotive, AWS, AWS re:Invent, GitHub Copilot, Microsoft Teams, ServiceNow, Salesforce, Anthropic, Claude for Enterprise, DRIVEQ
## Video description
Jodi Blomberg, VP, AI and Machine Learning, Cox Automotive
Abstracts
Artificial Intelligence is no longer a futuristic concept – it is changing the way we live and work, and the automotive industry is no exception. The global automotive AI market is projected to reach $48.9B by 2034 and AI is already shaping consumer experiences and dealership operations. Cox Automotive is leading the way by taking a pragmatically aggressive approach and delivering value through generative and agentic solutions. A leader in making powerful technology accessible to all dealership roles, Cox Automotive is harnessing the power of its unmatched data to create AI applications are dynamic and adaptive, built upon the industry's largest automotive dataset, allowing for solutions that get smarter with each interaction. Join us for a future-focused conversation about existing and emerging AI technologies and their impact on the automotive industry of tomorrow.
Cox Automotive leaders Jodi Blomberg and MJ argue that enterprise artificial intelligence has transitioned from exploratory generative software to autonomous agentic systems requiring an 'AI-first, data-differentiated' corporate posture. By pairing enterprise-wide productivity enablement with focused operational accelerators, the company aims to eliminate frontline friction and generate commercial revenue while managing deployment risks.
Inverting Enterprise Strategy to AI-First (6:39)
Cox Automotive, an enterprise spanning roughly $10 billion in software and tech-enabled physical services across more than 70 acquisitions, spent seven years methodically migrating its infrastructure to the cloud. Following an AWS re:Invent conference, executive MJ characterized the emergence of agentic AI as an epochal disruption comparable to the Industrial Revolution, prompting leadership to invert the company's operating posture from 'data first, AI enabled' to 'AI first, data differentiated.'
Automating Frontline Friction via Agentic Accelerators (2:50)
To separate tangible value from industry hype, Cox established twin accelerators focusing on business operational functions and next-generation code platforms, completing 27 proofs of concept within six months. The most immediate efficiency gains targeted frontline technicians repairing commercial trucks and servicing auction lanes. Accelerator teams shadowed field workers to automate administrative overhead, including sourcing replacement parts, cross-referencing regional labor rates, and drafting invoices. For a major fleet customer, this automated estimation workflow reduced turnaround times from 8–48 hours down to 30 minutes per vehicle.
Balancing Everyday Adoption with Guardrailed Scaffolding (9:27)
The organization pursues a bifurcated implementation model: driving broad baseline productivity through embedded ecosystem tools like GitHub Copilot and Microsoft Teams, while deploying dedicated accelerators to high-value areas like dealership service drives. Dedicated two-person teams delivered functional solutions in five weeks that previously required multiple scrum teams over six months.
To prevent the core AI unit from becoming an operational bottleneck, leadership transitioned from isolated builders into enterprise enablers. VP Jodi Blomberg instituted lightweight governance called 'scaffolding,' underpinned by a CISO-approved 'yes list' that permits distributed teams to experiment safely without central gatekeeping. Simultaneously, Cox is running five proofs of concept to analyze operational token economics and ensure inference costs do not outstrip commercial pricing.
C-Suite Buy-in and Commercial Deployment (14:35)
Rather than waiting for executive mandates, technology leadership accelerated buy-in through four-hour interactive bootcamps using Anthropic's Claude for Enterprise. Executives directly built prototypes to assess agentic capabilities and direct monetization strategies, establishing baseline productivity expectations across corporate roles.
Commercially, Cox paired large language models with DRIVEQ, its proprietary consumer intelligence platform. The system generates hyper-personalized communications for dealership prospects based on behavioral intent data. According to leadership, automating this outreach drove higher response rates and showroom conversion into vehicle sales while generating ongoing monthly recurring revenue.
Cox Automotive concludes that enterprise deployment requires pragmatic aggression, early quick wins, and human-in-the-loop oversight to mitigate risk. Long-term commercial success remains dependent on disciplined token unit economics and maintaining structured proprietary data.
## Background
该视频由 YouTube 频道 Ai4 于 2025 年 8 月 27 日发布，记录了 Cox Automotive 人工智能与机器学习副总裁 Jodi Blomberg 的主题演讲。内容聚焦汽车行业正在经历的智能化颠覆，详细阐述了 Cox Automotive 如何依托海量行业专有数据，通过生成式与 Agentic（智能体）解决方案重塑经销商生态与企业运营体系。
## TL;DR
演讲系统复盘了 Cox Automotive 将战略重心升级为“AI 第一、数据构筑护城河”的转型实践。技术团队通过双轨 Agentic 加速器敏捷攻坚，不仅将现场技师的车辆检测评估耗时从数天骤降至 30 分钟，还通过普及日常 AI 与组织高管实操工作坊打破了全员认知壁垒。借助安全“脚手架”机制与 DriveQ 专有数据平台的超个性化赋能，企业成功将技术突破变现为显著的经常性收入（MRR），为后来者提供了小步快跑、人机协同与速赢驱动的落地行动指南。
## Atmosphere & style
整体基调务实进取且极富感染力，开场以“踩下保时捷油门”的生动隐喻瞬间拉满紧迫感与变革张力。叙述在万米高空的宏观战略决策与深入一线维修卡车的现场实态之间自如穿梭，既有冷静理性的商业经济账推演，又有激发去中心化领导力的实战鼓励。整场分享充满实干家的热情与笃定，为身处 AI 浪潮中的企业管理者和技术人员带来了清晰可落地的行动底气。
## Key takeaways
- [00:04](https://www.youtube.com/watch?v=ITPb0DipNqg&t=4s) Cox Automotive 领导层将 Agentic 智能体的崛起视为堪比“地质板块构造剧变”或“发明车轮”的技术拐点，其深度超越互联网甚至工业革命。
- [02:50](https://www.youtube.com/watch?v=ITPb0DipNqg&t=170s) 从 AWS re:Invent 归程中，团队迅速拍板启动了分别聚焦“业务职能与效能重塑”和“下一代核心平台架构”的两大 Agentic 加速器。
- [03:48](https://www.youtube.com/watch?v=ITPb0DipNqg&t=228s) 在 6 个月内交付 27 个 PoC 的过程中，加速器深入现场维修卡车，将技师繁重的查配件、算工时、开票等文书流程交给智能体，单车评估耗时从 8 小时~2 天骤降至 30 分钟。
- [06:39](https://www.youtube.com/watch?v=ITPb0DipNqg&t=399s) 企业信条从“数据第一、AI 赋能”升级为“AI 第一、数据构筑护城河”，推动全员重塑岗位职责与商业模式，实现自底向上的技术拉动。
- [09:27](https://www.youtube.com/watch?v=ITPb0DipNqg&t=567s) 采取“双轨推进”策略：通过 ServiceNow 等现有生态推行“日常 AI”提升全员基线，同时利用敏捷加速器将核心业务攻坚周期从 6 个月大幅压缩至 2 人 5 周。
- [11:37](https://www.youtube.com/watch?v=ITPb0DipNqg&t=697s) 集中团队从“先驱开拓者”转型为“平台赋能者”，通过搭建合规安全“脚手架”与扩大“Yes 清单”，让业务部门在安全护栏内大胆自主构建 AI 应用。
- [13:15](https://www.youtube.com/watch?v=ITPb0DipNqg&t=795s) 推进规模化 PoC 必须严谨精算 Token 消耗与运行成本，在算力开销与业务收益之间找到平衡，并通过轻巧灵活的架构诱导全员参与而非生硬强制。
- [15:06](https://www.youtube.com/watch?v=ITPb0DipNqg&t=906s) 建议后来者采取“先斩后奏”快速做小步实验，对高管提供启动包并与 Anthropic 合作举办 4 小时沉浸式实操工作坊，快速打破认知壁垒。
- [18:42](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1122s) 秉持“务实且进取”的领导力原则，鼓励任何层级的员工在“高影响力、低风险”领域（如客户交互触点、数据聚合与内容生成）主动破局推动变革。
- [19:48](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1188s) 依托专有 DriveQ 数据平台部署大语言模型，代表经销商自动生成超个性化沟通文案，显著提升到店预约与成交转化率并创造可观的 MRR 经常性收入。
- [20:50](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1250s) 总结核心行动法则：理顺数据家底、快速交付速赢成果以锁定预算、拥抱系统不完美，并在高敏感业务中通过“人在回路”兼顾安全与速度。
## Detailed walkthrough
### [00:04](https://www.youtube.com/watch?v=ITPb0DipNqg&t=4s) “踩下保时捷油门”：Agentic 智能体带来的板块构造级巨变
Cox Automotive 人工智能与机器学习副总裁 Jodi Blomberg 在开场时用了一个生动的汽车行业隐喻：当人工智能的高速公路正式通车时，领导团队的 MJ 直接把一辆保时捷的钥匙交到了技术团队手中，并叮嘱道：“你最好一脚油门踩到底（You better step on it）。”这一过程虽充满混乱，却是一场“混乱而极致的狂欢”。长期统领数据与软件团队的 MJ 坦言，自己作为经历过多次技术浪潮的“老将”，曾见证过没有电子邮件的时代，也亲历了互联网的爆发，但眼前的变革与以往截然不同。
Cox Automotive 在过去七年中通过 70 多次收购整合了极其庞杂的技术栈，并坚决完成了向云端的迁移。三年前生成式 AI 爆发时，公司迅速跟进并赋能内外部业务；然而，2023 年 11 月在 AWS re:Invent 大会接触到 Agentic（智能体）概念后，团队意识到范式发生了本质颠覆。MJ 将其形容为一场“地质板块构造般的剧变（tectonic plate shift）”，甚至堪比人类“发明车轮”的历史节点。这种变革的深度与广度不仅超越了互联网的诞生，甚至比工业革命更为深远。对于那些领导层尚未意识到技术拐点、仍处于观望状态的企业，MJ 强调当前必须全神贯注应对这一不可逆的行业洗牌。
### [02:50](https://www.youtube.com/watch?v=ITPb0DipNqg&t=170s) 四小时飞行中的决策：启动两大 Agentic 加速器
从 AWS re:Invent 归程的 4 小时航班上，技术领导团队在万米高空迅速敲定了行动方案。为了在纷繁复杂的 AI 概念中辨别真伪、探索技术潜能，他们决定立即启动两个专门的“Agentic 加速器（Agentic Accelerators）”，各自承担明确的战略方向：
```mermaid
flowchart TD
accTitle: Cox Automotive 双轨 Agentic 加速器架构
subgraph S1["加速器 1：业务职能与效能重塑"]
A1["业务功能痛点梳理"] --> A2["效率与效益落地评估"]
A2 --> A3["未来代码生成模式演进"]
end
subgraph S2["加速器 2：下一代核心平台架构"]
B1["未来平台架构设计"] --> B2["Agentic 原生工作流验证"]
B2 --> B3["探索 AI 驱动的技术底座"]
end
```
- **业务职能加速器**：聚焦公司内部各业务条线，探索如何通过智能体大幅提升运营效率与业务成效，并重新定义未来的代码生成与软件研发模式。
- **下一代平台加速器**：面向未来构建核心平台底座，深入推演在智能体全面落地的现实下，下一代企业级软件与平台架构应如何演化。
### [03:48](https://www.youtube.com/watch?v=ITPb0DipNqg&t=228s) 深入一线：从 27 个 PoC 到现场技师工作流重塑
在 Jodi 带领业务加速器推进的过程中，各业务部门提交的 AI 需求呈井喷之势。团队建立了严格的筛选机制，坚决过滤掉那些把 AI 幻想成“万能魔法棒”的不切实际诉求，在短短六个月内高标准交付了 27 个具备明确商业价值的 PoC。除了常规的文案生成、文档解析和业务流程自动化外，团队将 Agentic 的核心重心放在了平时鲜少被科技圈关注的广大一线员工身上——数千名在公路旁抢修卡车、在汽车拍卖行钣金喷漆、做预防性维护以及调度车辆的现场技术人员。
```mermaid
flowchart LR
accTitle: 现场维修技师 Agentic 智能体工作流对比
subgraph Trad["传统手动模式（耗时 8小时~2天）"]
T1["查零件号与报价"] --> T2["核算区域工时费率"]
T2 --> T3["寻找替代配件与开票"]
end
subgraph Agentic["Agentic 智能体模式（缩短至 30分钟）"]
A1["智能体自动检索聚合数据"] --> A2["一键生成合规报价单"]
A2 --> A3["技师专注物理维修与检测"]
end
```
加速器团队甚至直接坐上了现场维修卡车，跟拍技术工人的实际维修过程。他们发现，技师在修车前需要耗费大量时间翻查适配零件编号、核实价格、计算当地工时费率、寻找替代配件并录入审批发票。这些琐碎繁重的案头文书工作正是 Agentic 智能体的用武之地：智能体在后台自动聚合所有信息生成待审账单，而将“修理卡车”这一 AI 无法替代的物理操作留给专业技师。
MJ 补充了一个极具震撼力的真实业务案例：公司的一家大型客户（涉及高频上门服务的物流巨头）要求在极短时间内对大批车队进行检测与估损。原本每台车人工检测、核对与评估需要耗费 **8 小时到 2 天**，而在部署了智能体工具后，单车评估时间直接被骤降至 **30 分钟**。这不仅带来了巨大的经济效益，更让一线技师从繁琐厌烦的文书工作中解脱出来，员工满意度“彻底爆表”。
### [06:39](https://www.youtube.com/watch?v=ITPb0DipNqg&t=399s) 从“数据第一”到“AI 第一”：全企业的全面转型战略
针对企业如何在“对内提升运营效率”与“对外打造创收产品”双重诉求间取得平衡，MJ 阐述了 Cox Automotive 从战略口号到顶层设计的全面进化。作为一家年营收近百亿美元的汽车软件与技术服务企业，Cox Automotive 将延续多年的企业信条从“数据第一、AI 赋能（Data first, AI enabled）”正式转变为“**AI 第一、数据构筑护城河（AI first, Data differentiated）**”。
MJ 强调，这一波智能体浪潮绝非单纯的技术团队项目，而是一场由 CEO 及全体高管层深度参与的**全业务转型（Whole business transformation）**：
- **重塑所有岗位职责**：从顶层战略到底层代码，每一个岗位的工作流都在被 AI 重构，重新定义人类创造力在业务闭环中的核心价值。
- **重构商业与营收模式**：全面审视 AI 如何改变产品的市场推广路径（Go-to-Market），如何直接提升利润率（Margin）并带动顶线收入（Topline）增长。
- **自底向上的全员赋能**：在底层，为每一位软件工程师全量配置 GitHub Copilot、Cloud Code 等 AI 编码工具；在协作端，将智能能力深度嵌入 Teams 等日常办公与业务系统中，推动企业从被动的技术“推动（Push）”转变为业务部门主动索求 AI 的“拉动（Pull）”。
### [09:27](https://www.youtube.com/watch?v=ITPb0DipNqg&t=567s) 双轨并行战略：日常 AI 提升全员基线与加速器攻关核心业务
在企业落地 AI 的路径选择上，团队采取了“双轨并行”的策略。一方面是推行嵌入在现有企业生态中的“日常 AI”（Everyday AI），例如 ServiceNow、Salesforce 等平台自带的智能化功能，旨在普遍提升全员的生产力基线。这就像打棒球时的击球策略，“你不断击出一垒打、二垒打，偶尔还能打出三垒打”，让整个组织的业务能力普遍水涨船高。另一方面，执行管理团队则专注于寻找能够带来“颠覆性差异化优势”的关键抓手——明确在面向市场的每一个核心客户领域中，到底哪一两项举措能够从根本上改变与客户的对话方式。
以汽车经销商业务为例，Cox Automotive 几乎与全美及部分海外市场的所有汽车经销商都有业务往来。在售后服务与维修车间（Service Drive）这一传统高价值场景中，团队通过引入数据智能与 AI 工具直接赋能现场维修技师，从而彻底重塑整个服务接待与维修流程。为了实现这种突破，团队设立了敏捷加速器（AI Accelerator），以周甚至以天为单位极速推进创新：原本需要多个 Scrum 敏捷团队耗时 6 个月的大型项目，现在加速器团队仅需 2 个人历时 5 周即可交付，部分轻量原型甚至在 4 天内就能成型。这种自下而上的日常提效与自上而下的核心攻坚（涵盖外部业务以及 HR、财务、运营中心、技术呼叫中心等内部职能）相互汇聚，构成了企业转型的核心动力。
```mermaid
flowchart TD
accTitle: AI 双轨推进战略架构
subgraph S1["日常普及层 (Bottoms-up)"]
A["现有生态工具<br/>(Salesforce / ServiceNow)"] --> B["全员生产力普惠<br/>(一垒打 / 二垒打效益)"]
end
subgraph S2["核心攻坚层 (Tops-down)"]
C["管理层战略差异化场景<br/>(如经销商维修车间)"] --> D["AI 加速器极速交付<br/>(2人5周 vs 6个月)"]
end
B & D --> E["企业全面智能化转型"]
```
### [11:37](https://www.youtube.com/watch?v=ITPb0DipNqg&t=697s) 搭建安全“脚手架”与扩大“Yes 清单”：从先锋到赋能者的转变
随着 AI 项目数量呈爆发式增长，单纯依靠集中式的加速器团队已经无法满足全公司的需求，团队的职能必须从早期的“先驱开拓者”（Pioneer）转变为“平台赋能者”（Enabler）。为了避免集中团队自身沦为交付瓶颈，首要任务是建立健全的“脚手架”（Scaffolding）以及一份不断扩大的“Yes 清单”（Yes List）。
Jodi Blomberg 形象地将这个机制比喻为：“就像允许你在房间里拿着剪刀奔跑，但前提是我已经提前把剪刀的安全隐患拿掉了。”在前期，团队已经抓住了如 GitHub Copilot 这样显而易见的机会，为公司所有软件工程师全面配发代码辅助工具并强烈鼓励使用；而接下来的核心在于建立起标准化的平台架构与安全护栏，使业务部门能够在符合首席信息安全官（CISO）合规与安全标准的前提下自主构建 AI 应用。为了统一思想，管理层召集了 70 多位助理副总裁（AVP）及以上的高管举行专题闭门会，即便短期内会出现一定程度的“AI 蔓延（AI Sprawl）”，也必须主动鼓励全员在合规边界内大胆尝试。
### [13:15](https://www.youtube.com/watch?v=ITPb0DipNqg&t=795s) 规模化 PoC 的商业经济账与全员文化诱导
AI 的落地不仅涉及技术开发，更深刻颠覆了企业的组织分工与商业化模式。目前团队内部正在推进 5 个更大规模的概念验证（PoC），其核心目的在于摸索商业化路径、明确运营归属以及厘清背后的经济账。在实际构建中，如果缺乏精细化核算，很容易设计出一套“Token 消耗与运行成本高于客户承受上限”的亏本方案；因此在推动采纳的过程中，需要在算力成本与业务收益之间找到动态平衡。
在组织文化层面，变革不能仅局限于坐在会议室里的管理层，而必须带动全员前行。针对那些持观望态度甚至掉队的员工，团队主张采取“诱导与吸引（Tempt）”而非生硬强制的策略。正因为底层技术更迭极其迅猛，团队选择快速搭设轻巧灵活的“脚手架”而非构筑僵硬笨重的传统架构，以此迅速为那些渴望动手创造的业务骨干提供施展拳脚的工具与土壤。
### [15:06](https://www.youtube.com/watch?v=ITPb0DipNqg&t=906s) 给后来者的启动建议：先斩后奏、防守反击与高管实战工作坊
对于那些起步较晚、风险偏好较低（例如金融服务背景企业）或面临诸多准入门槛的组织，发言人给出了极具实战性的建议：
- **立即行动与证明价值**：技术或产品团队应当立刻着手小步实验，用跑出来的实际业务价值说服最高管理层；这本质上是一场必须自上而下推动的“CEO 级转型”，若想从职能部门破局，就必须主动承担起对领导层的认知教育。
- **敢于“先斩后奏”**：在从 AWS re:Invent 大会归来后，团队没有层层申请审批，而是直接调动现有资源组建加速器团队跑出原型，用真切展示的增收与利润改善前景打动管理层。
- **攻防兼备的战略视角**：既要利用 AI 拓宽营收上限、优化利润率，也要清醒审视自身业务可能被 AI 颠覆的薄弱环节，实现以攻代守。
- **明确工程师与管理层的转型预期**：对工程师直接确立使用智能化工具的标准期望与生产力指标；对高管团队则提供专门的“新手启动包”（Starter Kit），化解部分领导因不懂工具而产生的顾虑与羞涩心理。
- **引入前沿大模型进行沉浸式实操**：深度携手 Anthropic 团队开展实战赋能。在近期的密集推进中，团队组织全体业务线 C-suite 及其营收负责人进行了长达 4 小时的研讨，深入探讨如何利用智能体（Agentic）重构商业变现逻辑；随后又组织各职能部门高管及其核心副手开展了 4 小时的 Claude for Enterprise 动手工作坊。高管们人手一个账号现场实操，极大地打破了认知边界，瞬间激发了全公司对 AI 落地的海量需求。
### [18:42](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1122s) 务实而进取的领导力：在任何层级推动变革与低风险切入点
在推进组织内的 AI 战略时，必须在各级领导层中树立一种“务实且进取”（pragmatically aggressive）的态度。这种策略的核心在于既不能“拿着剪刀盲目狂奔”去冒不必要的险，又必须保持果断与主动出击的姿态。正是通过这种目标极其明确且进取的方式，团队在聚焦的核心领域中能够以更快的节奏汲取经验。
演讲者向会场各个层级的从业者强调，领导力并不受限于职级头衔——无论你身处组织架构的哪个位置，哪怕手下没有庞大的团队，任何人都可以成为领导者。通过展示技术落地的可能性、主动施加影响力，普通员工同样能驱动重大的业务变革。若要寻找切入点，最明智的做法就是瞄准那些“高影响力、低风险”（high impact, low risk）的领域。如果团队一时间不知从何着手，可以优先考虑最直接的典型场景：客户的所有触达交互节点、归集底层数据并架设大语言模型以简化访问与理解，以及自动化的内容生成。
- **务实进取原则**：在把控风险底线的同时保持强烈的行动意愿，通过明确的重点聚焦加快试错与迭代周期。
- **去中心化领导力**：无需依赖科层制赋予的职权，通过输出可行性方案与实际成果自下而上地施加影响。
- **首选落地场景**：涵盖全渠道客户交互触点、基于 LLM 的企业数据聚合查询系统，以及营销与运营内容的自动化生成。
```mermaid
flowchart LR
accTitle: DriveQ 驱动的汽车营销赋能流程
subgraph DataLayer["底层资产"]
D["DriveQ 数据平台<br/>(专有车市与消费者数据)"]
end
subgraph AIEngine["AI 核心"]
LLM["大语言模型引擎<br/>(代销商撰写超个性化内容)"]
end
subgraph Conversion["渠道与业务回报"]
C["多模态全渠道触达"]
R["到店预约与成交流水激增<br/>(转化为显著 MRR)"]
end
D --> LLM
LLM --> C
C --> R
```
### [19:48](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1188s) 基于 DriveQ 数据平台的超个性化落地案例与商业回报
为了具体展示 AI 赋能的巨大威力，讲者分享了一个来自 Cox Automotive 的实战案例。汽车经销商日常需要与海量的潜在购车客户沟通，而 Cox Automotive 旗下拥有汽车行业独一无二的专有消费者数据，并统一沉淀在其 DriveQ 数据平台中。通过在 DriveQ 之上部署大语言模型，系统能够直接代表汽车经销商自动生成高度个性化的沟通文案，并通过消费者偏好的任意渠道进行推送。
这一方案的关键在于“超个性化”（hyper-personalized）——AI 基于平台所掌握的深层用户偏好与行为洞察，精准把握沟通时机与话术，经销商业务人员甚至完全不需要费心构思如何写信。实际业务数据显示，从初步的客户响应率，到最终预约到店试驾、购买车辆的成交转化率，整套链路的指标均迎来了大幅飙升。这项在技术构架上看似并不复杂的生成式 AI 应用，已经为公司带来了极其可观的每月经常性收入（Monthly Recurring Revenue, MRR），再次印证了“AI 落地的根基在于把自己的数据家底整理好”（have your data house in order）。
- **数据底座赋能**：依托 DriveQ 平台汇聚的行业专有数据，为大模型提供精准的上下文与用户画像输入。
- **全链路效果提升**：不仅拉动了营销文案的即时打开与回复，更实质性地提升了线下预约率与整车销售转化率。
- **商业价值兑现**：将基础生成式 AI 与高价值数据资产结合，成功孵化出具备高额 MRR 的稳定商业模式。
### [20:50](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1250s) 核心行动法则与总结：从小步快跑、人机协同到持续领先
在演讲尾声，讲者展示了一张总结核心要点的收官幻灯片，供参会者拍照记录，并凝练了启动 AI 项目的关键准则：
首先，“数据至关重要”，一切的前提都是理顺数据基础；其次是“立即行动起来”，不必在特定主题上过度纠结，关键在于打破停滞状态，赋能团队并建立起前进的惯性。团队应当采取小步快跑的策略，尽早锁定并交付“速赢”（quick wins）成果，因为及时的正向业务反馈是持续获取项目预算与资源支持的最有效途径。
在实施过程中，必须对不完美有所预期（plan for imperfection），在直面挑战与解决问题中持续学习。面对金融或汽车金融（FinTech）等对合规和安全性极敏感的业务场景，可以通过引入“人机协同 / 人在回路”（human-in-the-loop）机制来平衡风险与交付速度，让系统在与人类的协同互动中不断学习进化，从而在行业竞争中保持领先身位。
- **快速行动与快速见效**：从小处切入积累动能，用立竿见影的“速赢”项目证明价值，以此锁定后续资金与资源。
- **拥抱试错与动态学习**：接受初期系统的不完美，在实践中暴露问题并持续优化算法与业务流。
- **平衡速度与安全边际**：在高风险环节引入人在回路机制，确保合规与精准度的同时不牺牲推进效率。
## Notable quotes
> [00:04](https://www.youtube.com/watch?v=ITPb0DipNqg&t=4s) 你最好一脚油门踩到底（You better step on it）。 — MJ
> [00:04](https://www.youtube.com/watch?v=ITPb0DipNqg&t=4s) 地质板块构造般的剧变（tectonic plate shift） — MJ
> [06:39](https://www.youtube.com/watch?v=ITPb0DipNqg&t=399s) AI 第一、数据构筑护城河（AI first, Data differentiated） — MJ
> [11:37](https://www.youtube.com/watch?v=ITPb0DipNqg&t=697s) 就像允许你在房间里拿着剪刀奔跑，但前提是我已经提前把剪刀的安全隐患拿掉了。 — Jodi Blomberg
> [09:27](https://www.youtube.com/watch?v=ITPb0DipNqg&t=567s) 你不断击出一垒打、二垒打，偶尔还能打出三垒打
> [18:42](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1122s) 务实且进取（pragmatically aggressive） — Jodi Blomberg
> [19:48](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1188s) AI 落地的根基在于把自己的数据家底整理好（have your data house in order） — Jodi Blomberg
> [20:50](https://www.youtube.com/watch?v=ITPb0DipNqg&t=1250s) 对不完美有所预期（plan for imperfection） — Jodi Blomberg
## Mentioned
Cox Automotive, Jodi Blomberg, MJ, AWS re:Invent, GitHub Copilot, Cloud Code, Microsoft Teams, ServiceNow, Salesforce, Anthropic, Claude for Enterprise, DriveQ
## Video description
Jodi Blomberg, VP, AI and Machine Learning, Cox Automotive
Abstracts
Artificial Intelligence is no longer a futuristic concept – it is changing the way we live and work, and the automotive industry is no exception. The global automotive AI market is projected to reach $48.9B by 2034 and AI is already shaping consumer experiences and dealership operations. Cox Automotive is leading the way by taking a pragmatically aggressive approach and delivering value through generative and agentic solutions. A leader in making powerful technology accessible to all dealership roles, Cox Automotive is harnessing the power of its unmatched data to create AI applications are dynamic and adaptive, built upon the industry's largest automotive dataset, allowing for solutions that get smarter with each interaction. Join us for a future-focused conversation about existing and emerging AI technologies and their impact on the automotive industry of tomorrow.
[Source](https://www.youtube.com/watch?v=ITPb0DipNqg)
Source quotes
[0:04] MJ: 你最好一脚油门踩到底（You better step on it）。
[0:04] MJ: 地质板块构造般的剧变（tectonic plate shift）
[6:39] MJ: AI 第一、数据构筑护城河（AI first, Data differentiated）
[11:37] Jodi Blomberg: 就像允许你在房间里拿着剪刀奔跑，但前提是我已经提前把剪刀的安全隐患拿掉了。
[9:27] 你不断击出一垒打、二垒打，偶尔还能打出三垒打
[18:42] Jodi Blomberg: 务实且进取（pragmatically aggressive）
[19:48] Jodi Blomberg: AI 落地的根基在于把自己的数据家底整理好（have your data house in order）
[20:50] Jodi Blomberg: 对不完美有所预期（plan for imperfection）
Source quotes
[0:04] MJ: You better step on it.
[0:04] MJ: tectonic plate shift
[6:39] MJ: data first, AI enabled
[6:39] MJ: AI first, data differentiated
[9:27] singles, doubles, and hopefully some triples
[9:27] change the conversation
[11:41] Jodi Blomberg: The yes list is like: what are the places where I can let you run with scissors, except I took the scissors away.
[14:35] MJ: It is every role in the company should change... We've set expectations of what we think productivity changes should look like, and we are bringing folks on that journey.
[18:42] pragmatically aggressive
[18:42] running with scissors
[19:48] have your data house in order
Original transcript
[0:04] Uh, I'm Jody. I'm the VP of AI and ML
[0:06] for Cox Auto. We're a software services
[0:08] company as well as tech enabled services
[0:11] that we'll talk about in a little bit.
[0:12] Um, working in MJ's or uh is my best
[0:17] metaphor is automotive. Um, when they
[0:20] opened the AI highway, she handed us the
[0:21] keys to a Porsche and said, "You better
[0:23] step on it." And so that's true.
[0:25] >> It has been um
[0:27] >> uh chaotic but chaotically fun. um all
[0:30] the stuff that we've been able to do.
[0:31] And so um I'm going to start out by
[0:33] asking MJ, she's been leading data teams
[0:36] and software teams for a long time. Um
[0:39] what feels different this time around
[0:41] last few years?
[0:42] >> Yeah. Long time. She said my former CEO
[0:44] said we Susan D was salty. I kind of
[0:47] feel that way. Um yeah, I think what's
[0:50] interesting for us is we have been on an
[0:52] AI journey for a long time. Um
[0:55] connecting our data and the ways that we
[0:57] go to market. Um we have been super
[0:59] intentional about that and we went to
[1:02] the cloud over the past uh seven years.
[1:04] We were uh 70 plus acquisitions and
[1:07] every tech stack you could imagine and
[1:10] every complexity you could imagine. Um
[1:12] and our journey to the cloud put us in a
[1:15] great posture. But you know when
[1:17] generative hit I guess around three
[1:20] years ago it was interesting. Um we saw
[1:22] value in it. we leaned into it to start
[1:25] to deliver value for our customers and
[1:26] internal. But I will tell you what's
[1:29] fundamentally different for me right now
[1:31] is um Agentic. So when we saw that Jody
[1:35] in November, we're on we're in AWS shop
[1:38] and we came back from reinvent and said
[1:40] okay something is fundamentally
[1:42] different here and we need to go check
[1:45] it out and see what's really possible.
[1:47] And for me, what I'm seeing, and we're
[1:48] going to talk about what we've done and
[1:50] a little bit of the details of that, but
[1:51] what I have seen is a fundamental shift.
[1:54] I've been calling it a tectonic plate
[1:57] shift. Um, and of course, we're an
[1:59] automotive, so we always talk about an
[2:00] automotive um term. So, it's like when
[2:03] the wheel was created. That's how
[2:05] fundamentally different this is. And
[2:07] since I'm so seasoned, I'm salty. Uh, I
[2:10] remember when we didn't have email. I
[2:11] remember when the internet hit. And this
[2:14] is nothing like that. This is orders of
[2:16] magnitude different to me. It's even
[2:18] broader than the industrial revolution
[2:20] when you start to think about what's
[2:21] really happening. So, um, for those of
[2:24] you listening and for those of you and
[2:25] then if you're here, you obviously
[2:27] figured out that something is happening
[2:28] or you're neck deep in it. Um, but you
[2:31] may be in an organization that may not
[2:33] have had that same realization. You may
[2:35] have leadership that's not quite there
[2:37] yet and are trying to figure it out. Um,
[2:39] but this is so fundamentally different.
[2:42] you you have to pay attention. So that
[2:45] you know to me we've had a plate shift
[2:46] and we got to figure it out. Yeah. So I
[2:50] want to um move into the next part of
[2:52] the conversation and Jody when we came
[2:55] back what I just talked about we came
[2:56] back and we saw Agentic. We really
[2:59] wanted to figure out what was real, what
[3:01] was not, what was possible. Um we were a
[3:04] bunch of us were on a plane ride back
[3:06] and we had four hours to figure out like
[3:07] what we were going to do next. And I
[3:09] basically yes it was a Porsche. I said,
[3:11] "Okay, we're going to go buy a Porsche.
[3:13] We're going to spin up two Agentic
[3:14] accelerators and we're going to figure
[3:16] out what's real." And one was focused uh
[3:19] on our business functional areas to
[3:21] figure out what we could do to make an
[3:22] impact both from an efficiency and
[3:25] effectiveness perspective. Um and then
[3:27] how code was going to be created in the
[3:30] future. So our second accelerator uh was
[3:33] around the next generation platform of
[3:35] the future and what was how was an
[3:38] agentic reality going to change that. So
[3:40] Jody, you took on one of those
[3:42] accelerators. Talk about what that was
[3:44] like and what you've seen and maybe a
[3:46] few examples of what you've done so far.
[3:48] >> Well, we're all here, so AI is popular.
[3:51] You already know that. Um, we had we
[3:53] were flooded with intake at first. Um,
[3:56] and actually still have many, many more
[3:58] coming in. Um, and we had to do a
[4:00] winnowing process. We couldn't even the
[4:01] team's pretty small. Uh, for the AI
[4:03] accelerator, we completed 27 PC's in the
[4:06] first six months of the year. um those
[4:09] that could claim a business value
[4:10] proposition um and frankly had AI as a
[4:13] solution. Certainly some people had
[4:16] delusions of a magic wand that we don't
[4:18] actually own. Um a bunch of those cases
[4:20] are the typical ones that you already
[4:22] know about content generation, making
[4:24] sense of documents, business process
[4:26] automation. Um where we've decided to
[4:28] use Agentic is mainly actually when we
[4:31] think about people in the field. So, we
[4:32] employ thousands of people um who do
[4:35] jobs probably very different than
[4:36] everyone in this room. They fix trucks
[4:39] um on the side of the highway. They do
[4:41] preventative maintenance. They work at
[4:42] car auctions. They pound out dents. They
[4:44] paint cars. They drive them around. Um
[4:46] and so for that population, for example,
[4:48] for a technician, paperwork is part of
[4:51] their job. And our accelerator went and
[4:53] rode on that truck with them while they
[4:55] were doing their repairs. And one of the
[4:56] things that came up was, well, I have to
[4:59] look up which part number to use, what's
[5:00] the cost, what's the labor rate in this
[5:02] area of the country, who should I use
[5:04] for that part, what's if they're run
[5:05] out, what's the next one? That's a good
[5:07] job for an agent, right? To sort of pull
[5:09] all that paperwork together into
[5:11] something that they can get to an
[5:12] invoice that can get approved so they
[5:14] can start doing their job. Um, one that
[5:17] AI cannot do, uh, which is to fix that
[5:19] truck. And so that was one good example
[5:23] where we've got an agentic workflow
[5:25] doing something that both kind of
[5:27] delighted our user base. It's like
[5:28] something they don't really want to do.
[5:31] um and was able to let us sort of test
[5:33] and learn because I think the thing
[5:34] about Agentic
[5:36] um especially if you're at a conference
[5:38] like this and you feel like you're on
[5:39] some sort of AI rat race to make sure
[5:42] you're not falling behind um is that
[5:45] there just are a lot of unknowns and for
[5:46] us just building it you turn a lot of
[5:48] unknowns into known and that has been
[5:50] such a good learning curve because we're
[5:52] going to go to market with a Gentic
[5:53] products and learning the lessons sort
[5:55] of on ourselves um has been a good way
[5:57] there's still lots of unknowns but
[5:59] >> that use case that Jody just talked
[6:01] about. Just let me tell you about the
[6:02] business impact. An inspector or
[6:05] estimator who would have to go through a
[6:06] vehicle. We had a very large customer
[6:08] that maybe visits you pretty frequently
[6:10] at your home that we had to go u change
[6:13] the number of inspections on a vehicle
[6:15] within a very short period to meet their
[6:16] need. That activity would normally take
[6:19] anywhere from eight hours to two days
[6:21] per vehicle. The agentic um tool that we
[6:24] put in the hands of the estimators and
[6:25] technicians took it down to 30 minutes
[6:26] per vehicle. and and and then just also
[6:30] the person who's doing it, their
[6:31] satisfaction was just off the charts.
[6:33] So, they got to focus on what they do
[6:35] best, not on stuff they really didn't
[6:37] want to do.
[6:39] So, yeah. Well, like MJ said, we are
[6:41] here to make money. Um, and that's why
[6:43] we know how much time we've saved. But
[6:45] my question back to MJ is, uh, her and
[6:48] the executive team, we have sort of two
[6:49] opportunities to unlock potential with
[6:51] AI, like any company. What you do for
[6:53] yourself,
[6:54] >> how do you make yourselves more
[6:55] efficient? And then if you've got go to
[6:57] market products, what are you offering
[6:58] out that will actually what can you
[7:00] charge for basically? Um, and I'm going
[7:02] to ask you how you guys are thinking
[7:04] about those dual imperatives.
[7:06] >> Well, it's interesting too. So, because
[7:08] we've been AI forward, we were always
[7:11] talking about ourselves as a data first
[7:13] company, AI enabled. We've really
[7:15] changed that mantra to be AI first, data
[7:18] differentiated. I think depending on
[7:20] what you're doing and what type of
[7:22] business you're in, how you got
[7:24] differentiation in the past may not be
[7:26] where your differentiation is in the
[7:28] future. And because we do software, we
[7:30] sell software, we also have tech enabled
[7:32] services across, you know, we're roughly
[7:34] a $10 billion um automotive uh software
[7:38] and tech enabled services company. We
[7:40] have a variety of use cases. So we
[7:43] started on generative and teaching the
[7:46] organization and building some momentum
[7:48] and then when agentic hit I personally
[7:50] as I run product tech data the back
[7:53] office IT and security I have all that
[7:55] blessed to have all that together and I
[7:58] really had to show the rest of the
[8:00] organization
[8:01] what it meant and and really kind of
[8:04] blow their mind and the reality of it so
[8:06] they would come along the journey with
[8:07] me. So, um, we that's why we did the
[8:10] PC's to be able to do really how do we
[8:12] go fast and how do we show value and how
[8:15] do we prove out to ourselves what we
[8:17] need to learn. Um, and so that has been
[8:20] pretty mind-blowing and now the it's
[8:22] it's definitely not a push. It's it's
[8:24] more of a pull and now we got to figure
[8:26] out how to do this at scale. But one of
[8:28] the things I would say to you is that
[8:30] we've come up with a strategy on the
[8:31] executive team around really and I'm
[8:34] personally spent a lot of time helping
[8:37] to get there. But our CEO and all of my
[8:40] peers are realizing this is a whole
[8:41] business transformation. This is not a
[8:44] tech enabled project. This is not
[8:46] something that the technology team is
[8:48] going to do. This has to be a whole
[8:50] company transformation. And so it's
[8:52] really not just software development.
[8:54] It's not just services to help how when
[8:56] we do tech enabled services. It's how
[8:58] every role in your company will be
[9:01] impacted. It's how you go to market.
[9:04] It's how you drive margin. It's how you
[9:05] drive topline and it's also how you
[9:08] create the human creativity still in the
[9:10] loop and figuring out where your
[9:12] differentiation is going to come from.
[9:13] So we've done bottoms up focus where
[9:16] every engineer has GitHub copilot um
[9:18] cloud code um everybody in service
[9:21] different products and services that we
[9:23] use if it's teams you know you've got
[9:25] capabilities built into teams if it's
[9:27] service now it's Salesforce whatever is
[9:29] in your ecosystem there's everyday AI
[9:31] that you should be enabling and looking
[9:33] at how that impacts everybody's
[9:34] productivity and lift everybody up so in
[9:36] a baseball analogy you're getting a
[9:38] bunch of singles you're getting doubles
[9:40] hopefully some triples but everybody
[9:42] starting to lift their capability and
[9:44] get benefit. Then what we decided from
[9:47] the executive team is how to drive
[9:49] meaningful differentiation. So for
[9:52] everywhere we go to market, what are the
[9:54] one or two things right now that we need
[9:56] to go change the conversation for our
[9:59] customers and we've picked across our
[10:01] core customer segments where we do
[10:05] services today and how can we change
[10:07] that for them. So an example dealers we
[10:10] deal with almost every dealer in the
[10:12] country and um globally in some
[10:14] countries and their service drive of how
[10:17] they repair vehicles. How can we
[10:18] transform that with data and the
[10:21] technicians and transform the service
[10:23] drive? We already do a lot of high value
[10:25] there, but how can we change the
[10:26] conversation? So every single area we're
[10:29] doing business with, what are the two or
[10:30] three top initiatives that you can then
[10:33] go have an accelerator do that work very
[10:37] rapidly? I mean, we're seeing things
[10:39] being created in weeks, not months. I
[10:42] mean, the orders of magnitude that we're
[10:43] seeing examples of when you can build
[10:45] something, you can build something in
[10:46] four days. You can take, we had a team
[10:48] working on a particular project that
[10:50] took two people five weeks that would
[10:52] have taken multiple scrum teams six
[10:54] months. So, the speed in which you're
[10:57] learning and that you're doing things is
[10:59] going to be a game changer. So, it was
[11:01] bottoms up and it was tops down on what
[11:03] matters most. And I mentioned the
[11:05] business value of how you're going to
[11:07] market, but it's also every function.
[11:08] It's HR, it's finance, it's your
[11:11] operations center, any call tech center.
[11:15] What are the one or two areas that they
[11:17] have, you know, meaningful opportunities
[11:19] and how do you go after that? So, if you
[11:21] pick the top initiatives for hyperfocus
[11:24] with an accelerator and then everybody's
[11:26] bottoms up, they're going to meet
[11:28] together as we go through this journey
[11:29] together and your company will be
[11:31] transformed. So, that's we're trying to
[11:33] be structured about it. We still have a
[11:34] lot to go figure out.
[11:37] >> Oh, we have plenty of work left. Don't
[11:39] you worry.
[11:40] >> Especially Jody.
[11:41] >> Um well, yes. Uh because as I think what
[11:45] we're thinking about as we have to do
[11:46] more and more of it is we need a little
[11:48] more scaffolding, right? We've taken our
[11:50] obvious wins like
[11:53] co-pilot, handing code assist tools to
[11:55] every software engineer in the company
[11:57] and encouraging them strongly to use it,
[11:59] right? Really trying to make sure people
[12:01] are adopting AI. Um, and now if we want
[12:05] everyone to use it, we got to make what
[12:07] I like to call our yes list get really
[12:09] big, right? Um, the yes list is like
[12:13] what are the places where I can let you
[12:15] run with a I can let you run with
[12:16] scissors except I took the scissors
[12:17] away. Um, so more people can do AI,
[12:20] right? The AI accelerator is great and
[12:21] we have done a ton of PCs and we will do
[12:23] a ton more this year. uh but we will
[12:26] become the bottleneck and I think our
[12:27] job has to move from pioneer to enabler
[12:31] over time and to do that we really need
[12:33] to bring out platform and scaffolding
[12:36] and other things that allow people to do
[12:38] things within within things that our CS
[12:41] >> or CISO would approve um and also things
[12:43] that you know we think are good um use
[12:46] cases. So I think in the short term
[12:47] we'll also have a little bit of AI
[12:49] sprawl in there
[12:50] >> but um we're going to encourage that I
[12:52] think to get there. Last week we like I
[12:54] I have a leadership team and I had 70
[12:57] plus AVPs plus that came in town and
[13:00] talk a little bit about that because we
[13:02] were super intentional. The whole focus
[13:03] of that leadership meeting was really to
[13:06] make sure that we were operating uh in
[13:08] the right flow to start figuring out
[13:10] what that scaffolding looks like. So
[13:11] talk a little bit about that meeting and
[13:13] what we were going after.
[13:15] >> Yeah, the excitement in the room was
[13:16] good, right? But there's a lot AI
[13:18] disrupts like a lot of who does what and
[13:21] how do we go to market and how do we do
[13:23] this? Um and so we've done a couple
[13:25] things. We've done we do have uh five
[13:29] what I'm going to call larger scale PC's
[13:30] happening right now and that's an
[13:32] attempt to learn like how would we go to
[13:33] market with this? Who would run it? How
[13:35] would how would this work? What does it
[13:36] cost? Right? There's a lot of um
[13:39] unknowns still on the economics of it's
[13:42] very easy to design a solution um that
[13:45] is more expensive than you can charge
[13:47] for it if you're not careful. Um it's
[13:50] very we will probably my guess is in the
[13:53] pendulum of token usage like get out of
[13:55] whack before we ran it back in and in
[13:58] terms of encouraging adoption. But um we
[14:01] really sat in that room and and thought
[14:02] also about that whole company
[14:03] transformation. It can't be the people
[14:05] in the room. It has to be everybody with
[14:06] us. And how do we encourage that? How do
[14:08] we um I think my group called it tempt
[14:11] people into wanting to use AI for our
[14:13] stragglers rather than force. Um and the
[14:18] faster I think we can build out and I
[14:20] call it scaffolding because there's no
[14:21] bricks being laid here. It's moving too
[14:22] fast for that. But building out the
[14:24] scaffolding to let people do more stuff,
[14:26] right? For those people who are hungry
[14:27] to go build, like letting giving them
[14:28] tools to go build and figure it out.
[14:31] >> Yeah, absolutely. Um, well,
[14:35] I'm sure I have another question for
[14:37] you. I remember what it is. Um, I think
[14:39] that I said that we, you know, you gave
[14:42] us a Porsche and you told us to use all
[14:45] the horsepower and we have. Um, but if
[14:48] you're a company that's earlier in that
[14:50] journey, and that could be because of um
[14:52] I'm a financial services veteran, so I
[14:54] could imagine maybe some people's risk
[14:55] tolerance does not match ours.
[14:58] >> Um, I've worked in those companies or
[14:59] for various other reasons or other
[15:00] barriers to entry. Um, what would you
[15:02] recommend in terms of uh trying to get
[15:04] your own Porsche going?
[15:06] >> Yeah, I I I think you got to start. I
[15:08] mean, depending on where you are, you
[15:09] got to start. I was just in a a round
[15:11] table discussion a few minutes ago. And
[15:14] I think part of it is is if you're in
[15:16] the technology or a product
[15:18] organization, you need to start um
[15:19] experimenting. You need to show some
[15:21] value. Use that value that you're
[15:24] creating to get the rest of your
[15:26] executive team on board. I do think this
[15:28] is a CEO transformation. Um, and if
[15:31] you're trying to come at it from
[15:33] sideways and as a functional area, you
[15:36] got to help educate. You got to help
[15:38] bring that leadership team along. Why do
[15:40] they care? Why does it matter? Um, and
[15:42] that's number one. So, you'll get the
[15:44] support you need. So, there are some
[15:45] things that I've done that I just didn't
[15:46] even ask permission for, you know. So,
[15:49] when we um when we came back from
[15:51] reinvent and we said we're going to
[15:52] stand up these accelerators, there was
[15:54] no harm, no foul. I had a big enough
[15:56] organization. I'm like, we're going and
[15:57] we're going to prove out and we're going
[15:58] to learn and we're going to bring that
[16:00] information in. And as we started doing
[16:02] that, we started showing and
[16:03] demonstrating
[16:05] um to our leadership team and to our
[16:07] functional areas what it could mean for
[16:09] them. And everybody wants to expand your
[16:11] topline and everybody wants to have the
[16:13] right margin uh component. So that
[16:16] speaks volumes because it's real it's
[16:18] really both. Um you can use this to
[16:20] protect your business and you and you
[16:22] should be looking at how this is going
[16:24] to disrupt you. I mean, just as much as
[16:26] there's opportunity, there's just as
[16:28] much risk to you as to where you're
[16:30] going to be disrupted. So, how do you
[16:31] defend against that? How do you then
[16:33] accelerate? So, it's get started. Um, I
[16:36] then think you have to be super
[16:37] intentional about building the
[16:39] scaffolding. That's the word that we're
[16:41] using a lot lately is how do you build
[16:43] the structure around that? It's a
[16:45] workforce transformation. This is a
[16:47] workforce transformation. it is every
[16:49] role um and some roles more than others
[16:52] but but it's every role in the company
[16:54] should change and so starting to define
[16:57] what that looks like um just recently I
[16:59] mean we have a large number of engineers
[17:01] we set out an expectation of the tools
[17:04] that they were going to use rather than
[17:05] it being a pull it was like an
[17:07] expectation we've set expectations of
[17:09] what we think productivity uh changes
[17:11] should look like and we are bringing
[17:13] folks on that journey you got to bring
[17:15] people with you um same thing on the
[17:17] executive team we built um a starter kit
[17:19] for our executive team of how to get
[17:21] started. Some of them might have been
[17:23] acting like they were doing some things
[17:25] but maybe probably were a little scared
[17:26] to admit they didn't know what to do. Um
[17:29] so how do we get started? Um we um we
[17:33] work with all the models but we really
[17:35] are u seeing great progress with
[17:37] Enthropic and the anthropic set of
[17:39] models. We just leaned in with the
[17:40] executive team just to give you a feel
[17:42] for that. I had anthropics leadership
[17:44] with us. I my weeks are running
[17:46] together. Was that last week? I think it
[17:47] was last week, two weeks ago. Um, we had
[17:51] all of the seauite
[17:53] and their direct revenue generating
[17:55] leaders together and we spent four hours
[17:57] talking about how you can monetize
[17:59] differently with Agentic and thinking
[18:01] through use cases, what other companies
[18:03] are doing and thinking through our
[18:04] companies. We had another session with
[18:06] all our functional seuite and their top
[18:09] two leaders with anthropic to look at
[18:11] claude for enterprise and we it was
[18:14] hands-on. They all had IDs. they were
[18:16] include and they were practicing it and
[18:18] at the end of that four hours their mind
[18:20] was blown and so now I got to now that
[18:24] I've created all this demand and
[18:26] structure we got to go figure out how to
[18:27] enable it um quickly but I think you
[18:30] can't wait and I think your pace needs
[18:32] to be aggressive and I would also say
[18:35] that your everyday AI let that happen at
[18:38] the pace in which you do work today but
[18:40] the things that matter most and in
[18:42] pockets of leadership across that's
[18:44] where you need to be what I call
[18:45] pragmatic atically aggressive. We don't
[18:47] want to run with scissors, but but you
[18:49] need to be assertive and you need to be
[18:50] aggressive. So, the lessons we're
[18:52] learning, we're learning faster because
[18:54] we are being intentional and we're being
[18:56] aggressive and we're picking areas that
[18:58] we need to focus on. And and I always
[19:00] say from a leadership perspective,
[19:02] there's probably all different kinds of
[19:03] levels in this room and there's
[19:05] certainly all kinds of levels of
[19:06] different folks at this conference. You
[19:09] can lead from wherever you are. You do
[19:11] not have to be a hierarchal leader. you
[19:13] do not have to have um large groups of
[19:15] people, but you can influence, you can
[19:17] show what's possible and you can you can
[19:19] drive an impact. So, with that, I mean,
[19:22] to me, Jody, that is where you've got to
[19:24] get started um with what's real today.
[19:26] Pick your high high impact, lowrisk
[19:29] areas. I mean, there's some obvious use
[19:31] cases that if you don't know where to
[19:32] start, there's just some really obvious
[19:34] ones. It's all your point of contact
[19:36] with your um customers. It's getting
[19:38] your data together and putting a large
[19:40] language model on it to make it easy to
[19:42] access and understand. Um, it is content
[19:45] creation. I mean, those are some of the
[19:46] easiest use cases. But just an example
[19:48] of what Jody talked about is crazy. I
[19:51] mean, I'll give you one quick example.
[19:53] Dealers communicate uh to all their
[19:56] prospects. We have all this proprietary
[19:58] unique information about consumers in
[20:00] the automotive space and we have that in
[20:03] in our drive Q data platform. We enabled
[20:06] a large language model to create content
[20:09] for the dealers to send to a consumer in
[20:11] whatever modality that they would want,
[20:13] but we hyperpersonalized that message
[20:15] and had the generative AI model write
[20:17] the the message on behalf of the dealer
[20:20] with the proprietary information. Highly
[20:23] effective, relevant, and timely based on
[20:25] what we knew about them. And they didn't
[20:27] have to figure out how to write it. the
[20:29] response rate and then all the way
[20:31] through to the sell rate of the consumer
[20:34] making an appointment and buying the
[20:35] vehicle has gone up significantly and so
[20:39] we are generating significant monthly
[20:41] recurring revenue from something that is
[20:44] like pretty basic generative AI but it
[20:47] it is all about your data so have your
[20:48] data house in order I think we're
[20:51] probably at time so let me just leave
[20:53] this away I think this is a good
[20:54] takeaway slide if you want to take a
[20:57] picture um it really just talks about um
[21:00] getting started and getting buy in is
[21:02] data. Your data matters. Put yourself in
[21:05] motion. I don't care what topic it is.
[21:07] Put yourself in motion. Think about
[21:09] empowering your people and the ways to
[21:11] do that. Act now. Start small. Learn.
[21:15] Build momentum. Quick wins. Find quick
[21:18] wins. You find quick wins, you'll get
[21:19] funding. That always matters. Um and
[21:22] then I can't even read my own thing.
[21:24] It's you know what matters. It's focused
[21:27] on the PC's. I talked about that for
[21:29] everybody. Um, plan for imperfection.
[21:31] You're going to learn. You're going to
[21:32] see things that are challenging and
[21:34] you'll work through them. Balance your
[21:36] risk and speed. I just was in a talk
[21:38] with some people in the fintech
[21:40] financial services space. You know,
[21:42] human in the loop when you're concerned
[21:44] um and learn and it'll learn with you
[21:46] and then stay ahead. So, don't wait. So,
[21:49] thank you. That was a quick 20 minutes.
[21:50] We appreciate you attending.
[21:53] >> You thank you. Thank you.
