SmartStop Deploys AI Agents to Automate 85% of Storage Pricing

SmartStop部署AI智能体自动化85%仓储定价 自主定价工具将执行85%的日常收益工作流程，并使绩效较低的分析师能够超越未采用该技术的顶尖人工人才。 SmartStop Self Storage概述了其从手动电子表格演变至智能体AI架构的历程，该架构旨在使其230处物业中约85%的日常定价工作流实现自动化。演讲者详细介绍了其Property Insight Agent，展示了对话式自然语言工具如何诊断入住率滞后问题、评估物理单元重构，并在疲软市场以及学生住房旺季市场中推荐针对性的定价干预措施。讨论重点指出了智能体AI作为组织绩效平衡器的作用，使收益分析师的表现大幅超越传统的人工作业流程。最后，会议还回顾了SmartStop如何处理高频竞对价格抓取，并围绕LLM定价自主权设置严格的算法护栏。 SmartStop 在美加两国运营 230 家自助仓储物业，依托收益管理系统长期保持超 93% 的行业领先出租率，其收益管理模式已从 2017 年的手动 Excel 逐步演变为基于代理型 AI（Agentic AI）的主动决策流程。 (0:04) 公司部署的 Property Insight Agent 整合数十亿级运营与竞对数据，能提前 15 至 60 天预测出租率下滑趋势，并通过自然语言对话辅助分析师快速诊断异常原因。 (6:38) 该 AI 系统可精准区分价格敏感与物理库存错配问题，不仅能评估调整仓位尺寸的资本支出（CapEx）投资回报率，还将分析师晨间诊断耗时缩减 50% 至 66%，引导一线团队通过跟进线索与局部调价避免陷入破坏利润的价格战。 (9:07) SmartStop 将 95% 视为最佳出租率上限，超过该水平将触发提价机制；公司当前正推动实现标准收益管理工作流 85% 的自主自动化执行，以支撑资产规模扩张而无需线性增加团队编制。 (12:36) 系统在决策架构上将大语言模型（LLM）的灵活性置于传统算法基准与硬性规则护栏之内，并正加速采集本地竞对日内动态价格，使 AI 成为抹平初中级分析师能力差距的高效赋能工具。 (15:38)

SmartStop Self Storage is deploying agentic AI to automate up to 85% of its routine revenue management workflows across more than 230 properties. Executives Joe Robinson and Rene Hernandez argue that combining domain-tuned language models with algorithmic guardrails enables proactive demand forecasting, targeted price adjustments, and inventory reconfiguration without linear headcount growth. Complex Dynamics of Self-Storage Pricing (1:53) Self-storage pricing faces structural challenges distinct from traditional commercial real estate. Demand is strictly needs-based and reactive to personal life events, meaning operators cannot induce demand through discounting if a consumer lacks an active storage requirement. Additionally, customer tenures are open-ended, ranging from one month to several decades without fixed contract horizons. Managing thousands of distinct unit types across omni-channel funnels—where digital channels generate approximately 75% of customer volume—creates an environment where competitors frequently discount base rates and concessions. SmartStop, which operates over 15 million square feet across 23 U.S. states and Canadian provinces, maintains that maximizing total revenue requires prioritizing occupancy over pushing rates on a smaller tenant base. To maintain its average occupancy above 93%, the company moved beyond manual spreadsheets used in 2017 and first-generation machine learning deployed in 2020, shifting toward agentic workflows that predict vacates and inbound rental volume. Agentic Architecture and Algorithmic Guardrails (6:38) SmartStop’s Property Insight Agent ingests operational records and localized competitor rate data, which is collected daily or multiple times per day depending on population density. Rather than generating static reports, the agent forecasts demand disruptions and occupancy erosion across 15-, 30-, and 60-day horizons. Through natural language queries, revenue analysts prompt the system to identify underperforming assets each morning, which Hernandez reports reduces diagnostic preparation time by 50% to 66%. The deployment pairs conversational AI flexibility with algorithmic guardrails. Robinson and Hernandez specify that large language models do not possess unrestricted pricing authority; instead, traditional historical models set the analytical baselines, and the agents execute rate modifications strictly within pre-configured rules. The high cadence of competitor data collection also generates substantial data overhead, prompting the company to implement archival strategies to manage storage growth. Root-Cause Diagnostics and Physical Reconfiguration (9:07) When properties face concurrent declines in occupancy and asking rates—a compounding dynamic that Hernandez noted caused $95,000 in lost revenue and $24,000 in lost profit across three struggling locations—the agent isolates root causes by unit size category. Instead of defaulting to broad price cuts, the system recommends operational lead follow-ups and surgical asking-rate cuts of 4% to 7% on specific 5x10 and 10x10 inventory to avoid widespread revenue cannibalization. If stagnant demand reflects a structural mismatch rather than a pricing issue, the platform models capital expenditure alternatives. In submarkets where prospective renters need larger storage formats but the facility has excess small inventory, the AI evaluates whether physically converting 5x5 units into 10x10 or 10x15 units will generate an investment return within six months, replacing continuous discounting of unmarketable space. Yield Optimization and Portfolio Scalability (12:36) At high-performing locations, such as a student-focused facility in Texas where seasonal moves pushed occupancy to 95%, the system implements yield optimization. SmartStop treats 95% occupancy as an operational ceiling, as higher levels restrict inventory turn and indicate underpriced units. In response, the agent recommends raising asking rates by at least 5% and directs site staff to transition tenants to autopay to reduce payment delinquencies. SmartStop targets 85% autonomous execution for routine pricing workflows, shifting analysts from manual entry to managing edge cases. Robinson observes that the self-storage industry remains fragmented and historically slow to adopt technology, noting SmartStop was among only a few sector participants at the AI event. He contends that agentic AI functions as an organizational equalizer, enabling lower-performing analysts equipped with AI tools to outperform skilled analysts lacking them. SmartStop is scaling its agentic revenue platform across the United States and Canada, though rapid data accumulation from high-frequency competitor scraping poses ongoing data management and archival demands. Long-term efficacy hinges on maintaining algorithmic guardrails to prevent autonomous agents from breaching operational pricing thresholds.

SmartStop Self Storage 高管详细介绍了如何将代理型人工智能（Agentic AI）引入自主仓储的收益管理中，以解决需求被动、租期不固定以及非标单元定价难的问题。通过将日常分析与调价工作转变为受算法护栏约束的自动化代理流程，公司旨在实现常规任务的高比例自主处理，同时提升下行与高景气资产的资本回报率。 自主仓储定价难题与收益系统演进 (1:53) 自主仓储行业具有独特且受事件驱动的经济特征。SmartStop 首席运营官 Joe Robinson 指出，仓储需求完全由突发生活事件触发且租期开放（客户留存可从一个月延伸至三十年，例如长期存放 1980 年代论文的案例），降价无法凭空刺激不存在的存储需求。在全渠道运营中，数字化移动端占据了约 75% 的客户量。行业普遍遵循的定价逻辑是在单位利润与整体出租率博弈时，优先保住高出租率底线（公司年均出租率保持在 93% 以上），而非单纯依靠少量租户博取高单价。 为精确预测退租与新进客流，SmartStop 的技术体系经历了三代更迭：2017 年主要依靠人工维护包含嵌套 IF/THEN 逻辑的 Excel 表格与基础物业管理系统；2020 年上线基于机器学习的初代“SmartRate”收益系统，引入动态收益曲线模型；目前则进入代理型 AI 阶段，通过对话式智能体实时调取运营数据，实现事前预警与交互式诊断。 智能代理架构与分析师工作流重构 (6:38) SmartStop 收益管理总监 Rene Hernandez 介绍了其部署的 Property Insight Agent。该架构整合了市场趋势、进驻速度和退租规律等数十亿级数据点，能够提前 15、30 及 60 天识别潜在的出租率滑坡风险（例如提前预警南加州资产群可能出现的 3% 出租率回落），直接向分析师提供针对性诊断，改变了以往依赖静态报表的被动模式。 该平台目前已成为收益分析师每日工作的分诊台。分析师每天早晨只需通过自然语言询问核心关注点，系统即可自动排查出出租率与挂牌价同步下跌的高危物业，替代了以往手动编写 SQL 脚本和拼凑临时表格的数据准备工作。管理层表示，这一工作流使分析师每日早期的诊断耗时减少了 50% 至 66%，使团队能够迅速转向策略制定与实操干预。 空间户型重构与极值收益调优实践 (10:24) 在处理出租率与挂牌价同时受挫的资产时，模型会细分小、中、大各类仓型以进行归因诊断。以面临此双重压力的三处试点物业为例，该趋势累计造成约 95,000 美元的营收损失和 24,000 美元的利润折损。AI 建议现场团队侧重线索跟进、带看与推行自动扣款，同时在价格端对特定的 5x10 和 10x10 单元实施 4% 至 7% 的精准降价。针对部分因物理户型与当地需求脱节而滞销的资产（如 6116 网点的小型单元），模型可辅助进行资本支出可行性评估，指导资产端将过剩的 5x5 单元拆改重组为紧缺的 10x10 或 10x15 单元，力争在 6 个月内收回改造投资。 而在高景气市场，AI 则执行截然相反的收益最大化策略。以得克萨斯州学生公寓仓储为例，5 月高校离校潮推动其出租率冲至 95%。Hernandez 说明，95% 是公司认定的最理想上限，超过该水平会阻碍合理周转并暗示定价偏低。对此，模型发出提价信号，要求将挂牌费率至少上调 5%，并协同推进自动还款绑定，以锁定现金流并降低违约逾期概率。 算法护栏、高频数据与组织效能 (13:59) SmartStop 为收益管理制定了 85% 常规任务由 AI 自主执行的长期目标，以支撑其在美加两国跨 23 个州和省份、拥有 230 处资产的规模扩张。在技术架构上，系统并未赋予大语言模型无约束的定价权，而是将其锚定在传统历史模型与算法基线所设定的参数护栏之内，仅允许智能体在限定区间内微调。此外，系统根据周边人口密度在特定半径内每日或一日多次抓取竞对费率；面对同行日益频繁的日内调价趋势，高频抓取带来的海量数据增长已促使团队规划数据归档方案。 Robinson 认为，自主仓储行业在技术应用上整体滞后于其他商业地产板块，智能工具的应用正从根本上重塑团队的人才结构。他提出，先进的 AI 工具具有均衡业务水平的效果，配备这套辅助工具后，过往绩效偏弱的收益分析师能够超越未获工具赋能的顶尖分析师，从而帮助企业在不呈线性增加人力成本的前提下管理更大的资产组合。 SmartStop 的案例展示了代理型 AI 如何结合行业特定护栏实现自主仓储的动态收益管控，但同行日内调价所引发的高频数据膨胀压力，以及如何在算法既定边界与大语言模型的自主判断之间保持稳健平衡，仍是这一模式在大规模落地中需要持续应对的现实问题。

## Background 该视频由频道 Ai4 于 2025 年 8 月 27 日发布在 YouTube 平台上。视频记录了 SmartStop Self Storage 的首席运营官 Joe Robinson 与收益管理总监 Rene Hernandez 的联合演讲，内容主要围绕他们在自助仓储定价决策中开发与部署 AI 智能体（Agentic AI）的实战路径与应用成效展开。 ## TL;DR 演讲系统回顾了 SmartStop 作为自助仓储 REIT 在应对海量 SKU、不确定租期与激烈竞价等复杂场景下，定价系统从早期 Excel 演进至机器学习乃至 Agentic AI 的全过程。团队重点展示了 Property Insight Agent 如何通过数十亿级数据前瞻性预警出租率下滑、提供单元重构与靶向调价方案，并将分析师排查耗时缩减一半以上以迈向 85% 的常规业务自动化。同时，主讲人还分享了高频竞对数据抓取治理经验，以及在严格业务护栏约束下结合传统算法与 LLM 智能体进行动态调价的混合决策架构。 ## Atmosphere & style 整场演讲呈现出专业务实且幽默风趣的现场氛围，讲者开场通过调侃租户存放大学毕业论文等真实趣事迅速拉近与听众的距离。叙事节奏明快且极具实战说服力，在剖析行业科技滞后痛点与展示 AI 赋能产出的过程中充满自信与洞察，尾声问答环节讲者与台下技术团队“点头确认”的默契互动更平添了生动的技术探讨张力。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=4s) SmartStop 管理着 230 处物业资产与超 1,500 万平方英尺面积，全年平均出租率超 93%，近 75% 客户通过移动端进行线上预订与租赁。 - [04:02](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=242s) 自助仓储属于强需求驱动型消费且租期极不确定，在收益管理策略中维持高出租率通常优于追求单一高利润。 - [04:02](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=242s) SmartStop 收益管理经历了从 2017 年 Excel 规则、2020 年 Smart Rate 机器学习，到当前 Agentic AI 智能体体系的三代系统演进。 - [06:36](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=396s) 基于数十亿级数据点打造的 Property Insight Agent 能够前瞻性预警未来 15 至 60 天的出租率走势，促使团队从被动应对转向主动出击。 - [06:36](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=396s) 统一的 AI 工作台替代了繁琐的 SQL 临时报表制作，分析师可通过自然语言直接质询归因与获取日常重点物业清单。 - [09:07](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=547s) 面对出租率与租金双降的困境，AI 能够下钻诊断并评估“单元重构（Unit Reconfiguration）”的投资回报率，避免盲目打折。 - [10:24](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=624s) 针对双降网点，AI 制定了线索跟进、核心房型 4%-7% 靶向降价及主动客户留存策略，为分析师减少了 50% 到 66% 的日常排查时间。 - [12:36](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=756s) 在高出租率（如 95% 峰值）的稳定网点，系统识别出 5% 以上的提价空间并协同推广自动扣款（Auto-pay）以降低逾期欠费率。 - [13:59](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=839s) SmartStop 致力于将 85% 的常规收益管理任务交由 AI 自主处理，让分析师集中精力攻克异常疑难网点。 - [15:38](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=938s) AI 工具带来了强大的“拉平效应”，使普通分析师在先进工具赋能下的产出能够超越缺乏工具的最顶尖人才。 - [18:03](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=1083s) 系统按辐射半径高频抓取竞对租金以应对日内动态调价，同时通过历史数据归档来化解海量竞对数据带来的存储压力。 - [19:05](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=1145s) 定价系统并非任由大模型自由发挥，而是将 LLM 智能体的决策权严格限制在业务规则与传统历史算法模型的边界约束之内。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=4s) 开场介绍与自助仓储行业的定价挑战 SmartStop Self Storage（SmartStop 自助仓储）的首席运营官 Joe Robinson 携手收益管理总监 Rene Hernandez 共同登台，分享他们如何将 Agentic AI（智能体 AI）引入收益管理与动态定价决策中。Joe 在开场时与台下观众进行了轻松幽默的互动，询问现场有多少人使用过自助仓储、又有谁不喜欢这项服务，并打趣道台下没举手的人“非常聪明”。 作为一家专注于自助仓储的房地产投资信托基金（REIT），SmartStop 目前拥有并运营着 230 处物业资产，总管理面积超过 1,500 万平方英尺，分布于美国和加拿大的 23 个州与省份。今年全年的平均出租率更是保持在 93% 以上，在当前行业整体环境中表现十分亮眼。公司于今年 4 月成功完成 IPO，成为公开交易市场上仅有的四家自助仓储上市公司之一。 自助仓储看似简单，但面临着极其复杂的定价难题： - **多渠道与海量 SKU**：管理着数以千计的“单元类型”（Unit Types，即产品），且覆盖到店走访、电话呼叫中心以及移动端线上预订。其中将近 75% 的客户都是直接通过手机移动端进行预订与租赁。 - **刚需属性（Needs-based Business）**：自助仓储属于强需求驱动型消费，无法仅凭推销凭空创造需求，只有当客户真正遇到搬家、缩容或物品存放需要时才会寻找服务。 - **激烈的价格竞争**：市场竞争对手会毫不犹豫地通过降价或激进的促销手段争夺客源，且价格调整极其频繁。 - **多元客群与不确定的租期**：租户的使用场景千差万别（如小型分销商、存放季节性装饰品、搬家过渡等），且租期完全不可预测——有的只租 1 个月，有的租 6 个月，甚至有人一租就是 30 年。Joe 特别分享了科罗拉多州一位租户的真实趣事：该租户租了一个 5x10 的仓位却只放了几个箱子，现场经理曾建议他缩减仓位以节省开支，租户却坚持不换，因为里面装的是他早在 1980 年代初完成的毕业论文，对他而言具有无与伦比的纪念意义。 ```mermaid flowchart TD accTitle: 自助仓储定价业务复杂度维度 subgraph S1["多渠道交互渠道"] A1["线下门店走访"] A2["呼叫中心咨询"] A3["移动端在线预订 (占比近 75%)"] end subgraph S2["动态外部环境"] B1["强需求驱动 (非冲动消费)"] B2["竞品高频调价与促销打压"] B3["租期不确定 (1个月至30年不等)"] end subgraph S3["核心决策目标"] C1["数千种单元类型 (SKU)"] C2["平衡出租率与利润率"] end S1 --> C1 S2 --> C2 ``` --- ### [04:02](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=242s) 收益管理核心逻辑与 SmartStop 的系统演进 在谈及收益管理的底层逻辑时，Joe 强调“数据是开展收益管理的基石”。若没有海量且高质量的数据，收益管理将寸步难行。在自助仓储业务中，收益管理的核心在于精准的远期预测（Forward Forecast），涵盖退租（Vacates）与新租（Rentals）：退租往往是不可避免的自然流失，相对容易预测；真正的难点在于精确模拟不同挂牌价格（Asking Rates）下的新租需求曲线。 虽然行业内对“保利润率还是保出租率”存在策略分歧，但 Joe 凭借多年的行业经验明确指出，在自助仓储领域，**维持高出租率（High Occupancy）普遍优于追求单一高利润**。 SmartStop 自助仓储的定价系统演进经历了三个关键阶段： 1. **2017 年（初创期）**：成立了小规模的收益管理团队，依赖最基础的工具（如带有嵌套 IF-THEN-ELSE 逻辑的 Excel 表格，以及物业管理系统自带的简易功能），严重依赖分析师的纯手工计算与经验拼凑。 2. **2020 年（自动化与机器学习期）**：全面部署首代“Smart Rate”收益管理系统，将机器学习预测与分析算法结合，自动化拟合最优收益曲线。 3. **当前阶段（Agentic AI 智能化升级）**：引入自主智能体架构（Agentic AI），让系统从被动运算跃升为主动洞察与人机协同。 ```mermaid flowchart LR accTitle: SmartStop 收益管理演进路径 subgraph Stage1["2017 初创探索"] E1["Excel IF-THEN-ELSE"] E2["初级 PMS 内置功能"] E3["人工拼凑分析"] end subgraph Stage2["2020 算法升级"] M1["Smart Rate 系统"] M2["机器学习预测"] M3["最优收益曲线拟合"] end subgraph Stage3["当前 Agentic AI"] A1["Property Insight Agent"] A2["自然语言交互与学习"] A3["主动式经营决策"] end Stage1 --> Stage2 --> Stage3 ``` --- ### [06:36](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=396s) 引入 Property Insight Agent 实现前瞻性洞察 收益管理总监 Rene Hernandez 详细介绍了他们所研发的 **Property Insight Agent（物业洞察智能体）**。该智能体能够全天候扫描公司多年沉淀的数十亿级数据点，实时监控市场动态趋势、退租预测及新租表现。 与以往事后统计不同，该智能体能够进行高维度的前瞻性预警。例如，它能主动识别并提示：“由于未来 15 天、30 天或 60 天内可能出现出租率下滑，某区域物业预计将面临收益承压。”例如智能体可能会发出预警：“南加州的一组特定物业预计将在未来 60 天内面临 3% 的出租率下滑。” 这一架构使收益管理团队从传统的**被动应对（Reactive）转向主动出击（Proactive）**： - **自然语言人机挑战**：收益经理可以直接利用自然语言模型（Plain Language Model）质询 AI 智能体背后的原因（如：“这是季节性需求波动，还是发生了未预料到的市场竞争变化？”）。 - **护栏机制下的双向学习**：智能体在预设的业务护栏（Guardrails）内给出合理解释，并在收益经理的反复提问与交互中不断优化学习。 - **统一工作台替代人工报表**：SmartStop 为收益经理打造了统一的“落地页（Landing Page）作为每日工作大本营。过去团队每天早晨需要手动编写 SQL 查询、在庞大的数据库中拉取数据并制作临时报表（Ad-hoc Reports）；现在，收益经理每天上班只需询问智能体“未来两天出租率走势如何”或“今天我应该重点关注哪些物业”，系统便会直接指出出租率和要价同时下降的异常站点，从而节省出大量人力以专注于策略解决。 ### [09:07](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=547s) 仓储单元重构与双降困境的智能诊断 当仓储网点同时面临出租率下降与挂牌租金（asking rates）走低时，系统会迅速拆解数据并生成诊断摘要。模型按小、中、大各类房型逐层细分，甚至能精准定位到特定单元类型。例如系统会提示：“网点 6116 的小型单元已有 X 天未成功出租”。此时收益管理团队需要评估：究竟是定价失误、营销推广未达预期，还是面临更深层的结构性问题？ 系统给出的第二步关键建议是“单元重构（unit reconfiguration）”。这意味着现有库存结构可能与当地市场需求脱节——例如市场主力客群需要 10x10 或 10x15 的大单元来存放较多物品，而网点配置了过多无人问津的 5x5 小型单元。AI 赋能让团队无需盲目打折或追加营销投放，而是转向投资回报率（ROI）分析：评估投入改造资金后，是否能在 6 个月内通过快速去化实现更优的收益回报。 ```mermaid flowchart TD accTitle: 收益下滑诊断与应对决策流程 subgraph S1["数据监测与归因"] A["监测到指标双降<br/>(出租率/租金)"] --> B["颗粒度下钻分析<br/>(小/中/大单元类型)"] B --> C{"归因判断"} end subgraph S2["策略建议路径"] C -->|"市场需求错配"| D["空间重构 (Reconfiguration)<br/>评估改造 ROI"] C -->|"潜客转化不足"| E["线索与跟进优化<br/>(Lead Management)"] C -->|"定价偏离"| F["靶向降价 (4%-7%)<br/>拉动核心户型入住"] end ``` ### [10:24](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=624s) 应对双降网点的行动建议：线索管理、靶向调价与客户留存 针对旗下同时出现出租率下滑和租金走低的 3 个典型网点，如果不加以干预，粗暴的激进降价只会让营收状况雪上加霜。从业务影响层面上看，这 3 个网点预计将面临 95,000 美元的营收损失和 24,000 美元的利润缩水。AI 代理系统能够直接给出针对性行动方案，将收益分析师的日常排查时间缩短一半甚至三分之二： 1. **线索管理与预订跟进（Lead Management & Reservation Follow-up）**：聚焦转化率提升。在获取线索后，引导客户到店体验优质服务与产品附加值。 2. **重点房型靶向调价**：针对核心的 5x10 和 10x10 单元精准降价 4% 至 7%，快速打开出租率突破口，免去分析师早晨繁重的数据翻查工作。 3. **定向客户留存（Targeted Customer Retention）**：主动联系现有按月租赁的在租客户，关怀使用体验与设施状况，依托优质服务建立品牌忠诚度。此外，分析师还可以通过自然语言对话框向系统进行交互式追问。 - **潜在财务风险**：3 个双降网点如不及时干预，预计产生 9.5 万美元营收损失及 2.4 万美元利润下滑。 - **分析效率跃升**：AI 预先完成繁琐的数据挖掘，分析师工作耗时减少 50% 到 66%（"cutting that time in half, or even a third"）。 ### [12:36](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=756s) 案例剖析：德州高校网点的 95% 出租率优化与自动扣款推广 在另一个场景中，系统展示了对德克萨斯州某高校周边网点的调控逻辑。学生群体具有明显的季节性特征：5 月份学生离校过暑假时仓储需求激增，而 8 月份开学前后退租流动频繁。该网点的出租率已达到 95% 的理想峰值（SmartStop 将 93% 视为良好，95% 视为黄金上限，不建议盲目追求更高比例）。 在入住量与退租量保持健康周转率、市场自然需求强劲的背景下，系统识别到了提价空间，建议将挂牌租金至少上调 5% 以扩充营收。与此同时，针对该网点庞大的客群基数，AI 建议大力推行“自动扣款（Auto-pay）”专项行动。这不仅为租户提供了免于月度账单繁琐流程的增值便利，确保物品安全存放，同时也有效压降了逾期欠费率（delinquency），保障了整体现金流的稳定性。 ### [13:59](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=839s) 收益管理智能化转型与 85% 自动化目标 SmartStop 在推进 AI 转型过程中确立了三大核心抓手，致力于将 85% 的常规收益任务交由 AI 自主处理。其核心逻辑在于将收益分析师从琐碎的例行公事中解放出来，全力聚焦于“异常网点（exception opportunities）”——即那些业务遇阻、增长不达预期的疑难网点；而运行稳健的优质网点则由系统自动化承接运转。 Agentic AI（代理型 AI）全面接管了重复性任务与即时分析需求，为分析师赋予了极高的战略决策与执行优势。随着 SmartStop 在美国和加拿大市场的快速扩张，智能代理体系在营销与定价端的支撑使收益管理部门不再是规模化扩张的瓶颈，而是转化为强劲的增长引擎。正如团队的总结所言："自主化不再遥不可及，它正在迅速化为现实（Autonomy is no longer distant. It's rapidly becoming reality）。" ### [15:38](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=938s) 科技填补行业鸿沟：AI 如何颠覆传统收益管理竞争格局 SmartStop 首席运营官 Joe Robinson 指出，自助仓储（Self-Storage）行业整体在科技应用上相对滞后且高度分散。参加本次大会的仓储运营商凤毛麟角，行业龙头甚至全数缺席。然而，SmartStop 长期坚持在技术平台建设上深耕，并与长期战略伙伴 Kaizen Analytics（展位号 200）携手推进收益管理与人工智能的深度融合。 谈及 AI 对商业规则的重塑，Joe Robinson 强调了其强大的“拉平效应（leveling the playing field）”。十年前组建团队必须争夺行业最顶尖的分析天才，但优质 AI 工具的出现彻底打破了这种壁垒。通过赋能强大的 AI 工具集，即便是团队中表现垫底的收益分析师，其实际产出也能超越那些没有工具加持的最顶尖分析师。AI 正从根本上重塑 SmartStop 乃至整个仓储行业的竞争生态。 ### [18:03](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=1083s) 实时竞对数据采集与数据量挑战 针对现场关于“是否在数据集中实时引入竞品数据”的提问，演讲者给出了非常明确的肯定回答。SmartStop 会根据各物业所在区域的人口密度设定特定的辐射半径，每天定时抓取周边竞争对手设施的租金价格；在部分核心区域或特定情况下，抓取频率甚至会达到一天多次。随着市场上越来越多的竞争对手开始实施日内动态调价（Intraday Pricing Changes），团队未来可能还需要进一步提升数据采集的频率，以保证价格敏锐度。 ```mermaid flowchart TD accTitle: 竞品数据流与定价决策 subgraph Ingestion ["竞对数据采集"] A["周边竞品物业<br/>(按人口密度设定半径)"] --> B["每日/日内多次抓取租金"] B --> C["高容量数据存储与归档"] end subgraph Decision ["定价引擎处理"] C --> D["趋势预测与影响分析"] D --> E["LLM 智能体 + 历史算法"] F["业务规则与边界约束"] --> E E --> G["最终定价决策/调整建议"] end ``` 这种高频采集机制也带来了意料之外的数据治理挑战。由于不间断地持续拉取全量竞对信息，数据规模膨胀极其迅速。演讲者分享了一个生动的幕后插曲：前几天团队查看后台存储时忍不住惊呼“我的天哪，这玩意儿已经变得太庞大了（oh my goodness this thing has gotten really big）”，甚至已经到了不得不着手进行历史数据归档的阶段。尽管数据存储负担较重，但监测竞品动态依然是不可或缺的关键环节，它对于把握市场整体走势、评估自身物业的价格敏感度具有决定性作用。 - **采集半径与频率**：依据周边人口密度划定地理范围，常规为每日拉取，部分场景一天多次，未来将进一步提高频次以应对竞对日内调价。 - **数据压力**：连续高频采集导致数据量激增，团队正规划数据归档方案，以平衡存储成本与实时计算需求。 --- ### [19:05](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=1145s) LLM 智能体的决策权限与算法边界约束 在随后的问答中，观众询问了定价机制的核心逻辑：“系统究竟是完全交由大语言模型（LLM）来生成调价建议，还是结合了其他传统算法？”演讲者对此进行了解释，并幽默地望向台下的技术团队以确认表述是否严谨。他指出，系统并非毫无约束地任由大模型自由发挥，而是赋予 LLM 智能体在明确设定的业务约束（Constraints）范围内拥有灵活的调价决策权。 演讲者进一步说明，智能体底层的构建在很大程度上融合了传统的时间序列或历史定价算法模型，由这些成熟的基准算法提供基本驱动力；与此同时，系统给予 AI 智能体足够的自主空间，让其能在既定边界内根据实时输入做出微调与优化。在确认台下技术同事给出了一个“认可的半点头（a semi nod）”后，演讲者笑着确认自己的技术解释基本到位，并随后对全场听众的聆听表达了诚挚的感谢，圆满结束了本次关于智能体定价落地的分享。 - **人机协同与约束边界**：LLM 并非脱缰运作，而是在预设的业务底线、涨跌幅限制等约束机制下行使调价裁量权。 - **混合算法架构**：将传统历史定价算法的稳定性与 LLM 智能体的上下文推理和情境灵活性相结合。 ## Notable quotes > [04:02](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=242s) 数据是开展收益管理的基石 — Joe Robinson > [04:02](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=242s) 维持高出租率（High Occupancy）普遍优于追求单一高利润 — Joe Robinson > [06:36](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=396s) 南加州的一组特定物业预计将在未来 60 天内面临 3% 的出租率下滑。 — Property Insight Agent > [09:07](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=547s) 网点 6116 的小型单元已有 X 天未成功出租 — Property Insight Agent > [10:24](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=624s) cutting that time in half, or even a third — Rene Hernandez > [13:59](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=839s) Autonomy is no longer distant. It's rapidly becoming reality — Rene Hernandez > [15:38](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=938s) leveling the playing field — Joe Robinson > [15:38](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=938s) 通过赋能强大的 AI 工具集，即便是团队中表现垫底的收益分析师，其实际产出也能超越那些没有工具加持的最顶尖分析师。 — Joe Robinson > [18:03](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=1083s) oh my goodness this thing has gotten really big — Rene Hernandez > [19:05](https://www.youtube.com/watch?v=jEEcCvkGyTY&t=1145s) a semi nod — Joe Robinson ## Mentioned SmartStop Self Storage, Joe Robinson, Rene Hernandez, Ai4, Property Insight Agent, Smart Rate, Kaizen Analytics, Agentic AI, REIT ## Video description Rene Hernandez, Director of Revenue Management, SmartStop Self Storage Joe Robinson, Chief Operations Officer, SmartStop Self Storage Abstracts We will present the approach we took to develop and deploy AI agents in our Pricing decision making. [Source](https://www.youtube.com/watch?v=jEEcCvkGyTY)

[4:02] Joe Robinson: 数据是开展收益管理的基石 [4:02] Joe Robinson: 维持高出租率（High Occupancy）普遍优于追求单一高利润 [6:36] Property Insight Agent: 南加州的一组特定物业预计将在未来 60 天内面临 3% 的出租率下滑。 [9:07] Property Insight Agent: 网点 6116 的小型单元已有 X 天未成功出租 [10:24] Rene Hernandez: cutting that time in half, or even a third [13:59] Rene Hernandez: Autonomy is no longer distant. It's rapidly becoming reality [15:38] Joe Robinson: leveling the playing field [15:38] Joe Robinson: 通过赋能强大的 AI 工具集，即便是团队中表现垫底的收益分析师，其实际产出也能超越那些没有工具加持的最顶尖分析师。 [18:03] Rene Hernandez: oh my goodness this thing has gotten really big [19:05] Joe Robinson: a semi nod

[0:04] Joe Robinson: We'll touch base afterwards... it's kind of a relatively small community. [8:22] Where should we put our focus today? [13:59] Rene Hernandez: Autonomy is no longer distant. It's rapidly becoming reality. [15:38] Joe Robinson: If you have the right artificial intelligence tool set that can supplement what revenue analysts do, you can take your worst performing revenue analysts, hand them this tool set, and they can outperform your best performing revenue analyst without the same tool set. It's just that much of a game changer. [18:03] Speaker: Oh my goodness, this thing has gotten really big, we may have to start to archive some of this stuff. [19:04] Speaker: I'm going to look at my tech people. They're going to nod at me or not nod at me.

[0:04] My name is Joe Robinson. I'm the chief [0:05] operating officer at Smart Stop Self [0:07] Storage. I'm joined by Renee Hernandez, [0:09] who's director of revenue management at [0:12] uh at uh at Smart Stop. Um today we're [0:14] going to be talking to you about uh [0:16] using uh Gent AI in revenue management [0:19] and in self storage. I mean, it could be [0:21] any other industry uh but uh we work in [0:23] self storage, so that's what we're going [0:25] to we're going to talk about. So, kind [0:27] of jumping into this, let me just ask a [0:28] question of everyone. Um, who here you [0:31] currently uses self- storage? [0:34] Nobody. A bunch of smart people. Wait, [0:36] we got a couple of people. Who here [0:37] doesn't like using self storage? All [0:40] right. No. No less hands. Okay, you [0:41] won't offend me. I get it. I get it. [0:43] Does anyone here actually work in self- [0:45] storage? Nobody. Nobody. You're here. [0:48] Right here. Okay. We'll have to We'll [0:50] touch base afterwards. Um, it's kind of [0:52] a relatively small community. I know a [0:54] lot of people in the in the industry. [0:56] Um, but let me start off by just giving [0:58] a quick uh primer on kind of Smart Stop [1:00] and I'm going to introduce kind of the [1:01] pricing problem uh that uh that we face [1:04] and kind of what we're solving solving [1:06] to using uh Aentic AI. Smartstop is a [1:09] self-s storage rate. Uh despite whatever [1:11] numbers are up on the screen, we [1:12] actually uh own and operate uh 230 [1:16] properties uh over 15 million uh square [1:19] feet. Our occupancy on average actually [1:21] for this year has been just 93 north of [1:24] 93% for the entire year. And just from a [1:27] compar comparable perspective in the [1:29] industry, that's actually pretty good [1:30] right now. Um we uh we actually did an [1:34] IPO earlier this year back in April. So [1:36] we're now part of we're one of four uh [1:40] publicly traded uh self- storage on on [1:43] the uh on the uh stock exchange. We [1:46] operate in both the US and Canada and I [1:48] believe that we are in 23 states and uh [1:51] and provinces. [1:53] So let's start talking about the uh the [1:55] pricing problem. I'm going to introduce [1:56] this just in terms of what we're faced [1:58] with. We have thousands of what we call [2:00] unit types, but those are actually our [2:02] products uh that uh that we that we [2:05] price and we do it across multiple [2:07] channels. Whether it's people that are [2:08] walking into our offices, people that [2:10] are calling our call center, or the [2:11] majority of folks that are reaching us [2:13] online and it's not, you know, down in [2:15] their basement on their computer, uh [2:17] almost 75% of them, they all do it from [2:19] their their mobile phone. It's a [2:20] needs-based business. When someone needs [2:22] self storage, that's when they're going [2:23] to go purchase. And uh you can't just [2:25] talk them into going and and and using [2:28] uh self- storage. It's pretty aggressive [2:30] in terms of competition, especially [2:32] right now. um you know people they they [2:35] will they will not hesitate to undercut [2:37] you in terms of in terms of pricing uh [2:40] and in terms of the promotions that [2:42] they're that they're going to that [2:43] they're going to offer and they will [2:45] change their prices all the time uh [2:47] especially the more sophisticated uh [2:50] ones will do it. The other challenge [2:51] that we face is we don't have like, you [2:54] know, a one- type product or one type [2:56] renter, right? So, people use self [2:58] storage for a lot of different reasons. [3:00] Um whether they're they've got a a [3:03] little debut distributorship, whether [3:04] they're seasonal decorations they're [3:05] storing, whether they're downsizing, [3:07] whether they're moving, a lot of [3:08] different reasons. There's no silver [3:10] bullet. And the other thing that's [3:11] interesting about our customers are that [3:14] they don't there's no preset length of [3:16] time that they're going to stay with us. [3:18] Um and so they will they will stay um [3:21] you know some will stay a month, some [3:23] will stay uh six months, some will stay [3:27] 30 years. Uh the story of one gentleman [3:30] that uh that was that was running with [3:32] us in Colorado. Uh he had a 5x10 and I [3:35] know site managers going out there's [3:36] like you just have a couple of boxes in [3:38] there. Why don't you downsize this or go [3:39] do something else with this? You don't [3:41] need this space. Um and the gentleman [3:44] was like nope. we're just going to leave [3:46] those boxes right there. Turns out that [3:49] was his dissertation from back in the uh [3:51] back in the early 80s. Um and he was it [3:53] was so special to him. He just wanted to [3:55] leave it in that one spot and uh never [3:57] change it. So, how do we attack this [3:59] pricing problem? Well, if any of you [4:02] actually can I get a quick raise of [4:03] hands? Who here works in revenue [4:04] management or pricing there? That's why [4:07] you guys are here. Okay, great. That's [4:09] the majority of the people. If you guys [4:11] all know, anyone who works in revenue [4:13] management or pricing, you need data and [4:14] you need lots of data. If you have data, [4:16] lots of data, you you're off to a great [4:18] start. If you don't have data, it it's [4:20] it's going to be a little trickier. So, [4:22] you got to have lots of data. In our [4:23] business, you've got to have a good [4:25] forward forecast in terms of what's [4:27] going to happen with your rentals and [4:28] what's going to happen with your [4:29] vacates. Vacates are vacates are going [4:32] to happen no matter what your p your [4:33] asking rates are. They're they're going [4:34] to they're going to go up or down. You [4:36] can get pretty good on those forecasts. [4:38] where you've really got to dial it in is [4:40] where what what level of rentals are you [4:43] going to have at um at various price [4:46] points. So, you've got to be really good [4:47] at the uh forecast. And then, you know, [4:50] in terms of like complexity and [4:52] sophistication that you're adding to [4:53] revenue management, you're going to want [4:55] to have, you know, uh machine learning [4:58] um you know, AI based kind of analytics. [5:00] And that's some of what we're going to [5:02] talk about uh to today. But really what [5:05] you're trying to do is maximize your [5:07] overall revenue by managing your margin [5:09] uh versus your occupancy. I can tell you [5:11] as a general rule in self- storage um [5:13] these gentlemen may disagree with me um [5:16] but I've been doing this a really long [5:17] time um but as a general rule in self [5:19] storage you want to have higher [5:21] occupancy as opposed to lower occupancy. [5:24] And I can talk to you for hours about [5:27] why that actually is laid out like that [5:29] and give you actually current examples [5:31] of why that's the uh that's the case. So [5:34] let me talk real briefly briefly about [5:36] the smart stop evolution in terms of [5:39] pricing. Uh back in 2017 we actually [5:42] formed a small revenue management team [5:44] and you know you bring in a couple of [5:45] smart people and we gave them really [5:48] rudimentary tools. You're talking about [5:49] like Excel spreadsheets with the if then [5:51] else statements and then some some [5:53] functions that were built in kind of [5:55] property management systems for self [5:57] storage. Um but they weren't really [5:59] sophisticated. So you're relying on your [6:01] revenue analyst to do a lot of kind of [6:03] that that work and and make things uh [6:06] fit fit together. Uh back in 2020, uh we [6:10] actually deployed kind of our first kind [6:12] of whole scale uh smart rate revenue [6:15] management system which really brought a [6:18] lot more of kind of that machine [6:19] learning sophistication forecasting that [6:22] kind of stuff and married it together to [6:24] to to kind of drive towards kind of [6:26] optimal revenue revenue curves. Um, and [6:29] what we're going to talk about today is [6:31] actually how we're transforming that uh [6:34] using um Aentic AI. And with that, I'll [6:36] turn it over to Renee Hernandez. So, as [6:38] we jump into this, uh, the one thing is [6:40] the AI has really powered our revenue [6:42] management system here at Smart Stop to [6:44] the next level. As Joe mentioned, we [6:46] were pretty rude of entry back in 2017 [6:49] and the biggest opportunity for us was [6:51] to really get an understanding of our [6:52] properties, right? So, we led with our [6:54] property insight agent. And really what [6:56] this does is it scour our billions of [6:58] data points that we've accumulated over [6:59] the years. And what it's doing, it's [7:01] it's it's reviewing the market uh [7:03] trends. It's reviewing our vacates, our [7:05] rentals, and essentially forecasting [7:07] what what we would anticipate, but at a [7:10] higher level. So, it's giving us the [7:12] insight of saying, "Hey, you're [7:13] potentially going to see some hindrances [7:14] in your revenue due to occupancy loss [7:16] within maybe the 15, 30, 60 days." And [7:20] that's giving us the opportunity to give [7:21] our revenue managers a highle [7:23] understanding of that data a little more [7:26] proactive than reactive. So our revenue [7:28] managers essentially they get this data [7:30] pool from the AI agent and what the AI [7:32] agents telling you is hey Southern [7:34] California your subset of properties or [7:36] we're anticipating 3% loss of uh [7:39] occupancy within the next 60 days. So [7:41] what our revenue managers do, they take [7:44] that data and then what they do is they [7:45] question the uh the plain language model [7:48] uh that we've developed. So what they're [7:50] doing is they're challenging the AI [7:52] agent and at the same time that agents [7:54] learning based off the questions that [7:55] our revenue managers are asking the [7:57] system. So it's telling you, hey, you're [7:59] going to lose this occupancy. Great. The [8:00] question is why are we going to lose [8:02] this occupancy? Is it seasonal demand? [8:04] Is there something changing in the [8:05] market that we just didn't anticipate? [8:07] So, as they ask these questions, the [8:09] model's responding to it within the [8:11] guardrails that we've set, and it's it's [8:13] helping us learn more about our business [8:15] in anticipation of of of, you know, [8:17] future hindrances that may affect our [8:19] revenues. [8:22] So, we've developed uh, you know, a a [8:24] landing page for our revenue managers [8:26] and really this is the home base for [8:28] them. So, they essentially go in here in [8:29] the morning and what they're saying is [8:31] what do we anticipate in terms of [8:33] occupancy for the next two days or where [8:35] should I put my focus on today? Whereas [8:37] in the past our revenue managers would [8:39] have to develop ad hoc reports uh review [8:41] our data you know scour our database [8:44] pull uh queries and really that was a [8:46] lot of time right a lot of manpower that [8:48] we could be using to solve some of these [8:49] issues that the uh AI agents developing [8:52] for us. So this here's an example of a [8:55] of a site or of a question that we asked [8:57] the system uh essentially where should [8:59] we put our focuses today and really what [9:02] it's doing it's telling us there's a [9:03] couple sites that are decreasing in [9:05] occupancy and asking rates. So great, [9:07] we're decreasing occupancy and asking [9:09] rates. What's the next step there? So [9:11] really, it's going to break it down for [9:13] us. It gives us a short synopsis of [9:15] what's uh what's anticipated. So your [9:17] unit occupancy, it carves it up by [9:19] category, small, medium, large, and it [9:21] goes even deeper into that uh based on [9:23] unit types. And it's telling you, hey, [9:25] your small units at site 6116 has not [9:28] been running for x amount of days. [9:30] Great. We haven't been renting those [9:31] units. Is it our pricing? Is it our [9:33] marketing efforts? are are they not [9:34] achieving what we're trying to achieve [9:36] in terms of occupancy? The second step [9:38] here, the the model is telling us, well, [9:41] you may want to look at unit [9:42] reconfiguration. And what that's telling [9:44] us is that this the uh inventory that we [9:46] have for that market may not be playing [9:48] correctly, right? We may have too much [9:50] or too small of an average square foot [9:52] for that market where a customer may be [9:54] looking for 10 by 10 x10s, 10 x 15s, and [9:57] 5x5s just don't work for them. They have [9:59] more things to store. So this op this [10:01] gives us the advantage of going and and [10:03] doing additional analysis and saying hey [10:05] is it worth doing the reconfiguration at [10:07] our property? Should we spend a little [10:08] more money and get that return on [10:11] investment you know 6 months down the [10:12] line because we would fill those units [10:14] up quicker than we would by pricing [10:16] those aggressively in that market or or [10:18] putting more marketing effort and [10:19] dollars to those units. [10:24] So one of the examples we wanted to go [10:25] over was uh decreasing occupancy and uh [10:28] decreasing rates. So, what this is [10:30] telling us is there's three sites here [10:31] that are both experiencing decreasing [10:34] occupancy and decreasing rates. So, [10:35] we're losing occupancy and we're pricing [10:37] very aggressive. So, if anybody knows [10:39] here in revenue management, when you do [10:40] that, that's really not going to grow [10:42] the revenue for you. Um, so the business [10:44] impact high level would be essentially [10:46] with those three sites, it would be a [10:48] revenue loss of $95,000 and a profit of [10:51] 24 $24,000. Essentially, [10:55] the system's going to recommend the [10:57] actions, right? We have very smart [10:59] revenue managers that go in there and [11:00] determine what's the best action they [11:02] should take to solve these issues. Now [11:04] they're cutting that time in half, [11:06] right? Or even a third of what they're [11:08] doing there. And it's telling us we [11:09] should focus on our lead management. [11:11] Let's focus on reservation followup and [11:13] initially improve our conversion. So, [11:15] we're getting the leads, but we need to [11:17] put a little more focus on talking to [11:19] our customers, getting them in into the [11:21] store, showing our products, showing [11:22] them our customer service, really [11:24] showing the value ad to them with what [11:27] we're offering. And additionally, it's [11:29] telling us we should focus on our 5x10 [11:31] and 10 x10 units and decreasing those [11:33] rates by four and 7%. So, those are our [11:36] areas of opportunity for occupancy. We [11:38] can see that as we dig through our data [11:40] and and determine, okay, this is [11:41] probably what's holding a site back, but [11:43] the model's telling us now beforehand, [11:45] right? We don't have to go do that [11:47] analysis. And our managers have so much [11:49] domain knowledges uh domain knowledge [11:51] that they can see that that that bullet [11:53] point there and be like, okay, that's [11:55] probably something I need to go focus [11:56] on, cutting out all the leg work that [11:58] they would have to do in the morning. [11:59] And then lastly, implement target uh [12:02] customer retention. So really what that [12:03] is is touch base with your customers, [12:05] right? you have a lot of customers that [12:07] are rent renting from you monthly, touch [12:09] base with them. Ask them how their [12:10] experience has been. Uh how's the [12:12] property doing? How's their unit [12:13] treating them? Uh we we really pride [12:15] ourselves on customer service. Uh we we [12:17] like to think we hire the best managers [12:18] out there because we really want to [12:20] create that experience for the customer [12:21] and when they decide to rent again, they [12:24] decide to use Smart Stop. [12:27] And then there's an example too where we [12:29] can just use our our natural language [12:30] prompt where we can ask additional [12:32] questions to get feedback from our uh [12:34] from our system. [12:36] Example two. Uh this property here in [12:38] Texas, it's a student property. Uh [12:41] occupancies are pretty high around [12:42] student uh around August and and May. Uh [12:45] August is when we typically lose the [12:47] occupancy due to student vacates, but [12:49] May is when people use the units as [12:51] they're going home for the summer or [12:52] whatnot. Um this property in particular [12:54] has hit 95%. This is really like where [12:57] we like to optimize our occupancy. 93 is [13:00] great. 95 is even better, but we don't [13:02] like to go higher than that. And what [13:04] that means is we're having stable [13:05] moveins and stable vacated activity. Our [13:07] turn rate rates healthy, right? We're [13:08] bringing in more people than we are [13:09] vacating people. And there's natural [13:11] demand there. And what that means is we [13:13] have opportunity for revenue. So the [13:15] model is going to uh uh recommend an [13:17] increase of 5% at at minimum to drive [13:21] additional uh revenue here because of [13:23] the the stable volume that we're seeing [13:25] from Reynolds. [13:27] And then some of the recommended actions [13:29] implement higher asking rates as we [13:30] mentioned above and then develop a [13:32] targeted initiative to drive autopay. So [13:34] what that means is you have a large set [13:36] of customers now at this property. Let's [13:38] focus on the autopay. let's provide more [13:40] of a service value ad to the customers [13:42] where you know hey let's set you up with [13:44] autopay you won't have to worry about [13:45] your monthly billing go do your thing [13:47] your your stuff is stored safely in our [13:49] unit and really that also helps us [13:51] mitigate some of our our higher higher [13:53] delinquencies and um overall driving [13:56] revenue [13:59] as we transform with the AI at smarttop [14:02] uh really what the the three main items [14:04] that we were focused on was really [14:06] streamlining what our uh what our [14:08] revenue task costs are we we we're [14:10] targeting 85% of AI uh usage really and [14:14] what that means is we really want our [14:16] revenue analysts to focus on exception [14:18] opportunities and what that means is [14:20] problem sites or sites that just don't [14:22] seem to be getting off the ground the [14:24] volume that we're anticipating or [14:25] revenues so they can spend more time on [14:27] solving those issues and let the system [14:29] really run the properties that are doing [14:31] great right um we have the data points [14:33] we have the system running that's what [14:34] we like to achieve uh Agentic AR takes [14:37] over repetitive tasks in on demand [14:39] analysis. I mean, I think that's why [14:40] we're all here, right? We're trying to [14:42] utilize AI to really cut down on those m [14:44] mundane tasks that we have and really [14:46] giving us giving the analyst a strategic [14:48] edge and opportunity execution. So, [14:50] letting them make decisions faster, [14:52] quicker quicker pricing decisions, [14:54] quicker uh execution of do we need to go [14:57] focus on this aspect of the business or [14:58] that aspect and then driving exponential [15:01] growth. Uh we are we are growing as an [15:04] organization and really we want to set [15:06] oursel up to have a lot of agents in [15:08] place that can handle a lot of the [15:10] mundane tasks pricing marketing and [15:12] allow our analysts to really scale up [15:14] with the growth of our property. So [15:16] we're not using revenue management as a [15:18] hindrance. It's more of a value ad for [15:19] us where we can keep growing but we can [15:21] keep up with the demand of making [15:23] pricing changes and really driving [15:25] revenue in markets in Canada and the US. [15:28] And really, I'll lastly, I'll leave with [15:30] uh my little quote. Their autonomy is no [15:32] longer distant. It's rapidly becoming [15:34] reality. [15:36] And I'll turn it back over to Joe. [15:38] >> Thanks, Renee. Grab this. You know, um [15:42] it's smart stop. Well, first of all, [15:44] self storage is [15:47] it's kind of an interesting business. [15:48] It's like behind in technology. You have [15:50] a lot of fragmentation in terms of [15:51] operators and different things. In fact, [15:53] I can tell you that we're probably one [15:56] of maybe two or three self- storage [15:58] people actually at this conference and I [16:00] will tell you the major players are not [16:02] here period. Um, so it's it's just it's [16:06] something that we kind of lag behind and [16:08] catching up, but like doing this stuff [16:10] is super interesting to us. And I think [16:12] from a Smart Stop perspective, making an [16:14] investment in technology um to really [16:17] advance our platform is something that's [16:19] been critical to us for a long time. uh [16:22] we've made a long-term investment uh in [16:25] in in in kind of this the uh revenue [16:27] management and artificial intelligence [16:30] especially with our our long-term [16:31] partners Kaizen Analytics. Um and they [16:34] actually have a uh they have a booth [16:36] here in the exhibit hall number 200. You [16:38] you see them right when you right when [16:40] you uh right when you go in. But I want [16:42] to leave you with just a couple of [16:43] things and I'm really only going to [16:44] speak about one of this one of these [16:46] things. And this is how we think that [16:48] artificial intelligence will change the [16:51] playing field uh for us. If you go back [16:54] maybe a decade and you're like, I'm [16:56] putting together a revenue management [16:57] team, you're like, I need the I need the [16:59] best. I need the the the the the top the [17:01] top people to come and and work with me, [17:04] the smartest people that are out there. [17:06] But I think one of the things that [17:08] artificial intelligence does is it [17:10] really starts to level that playing [17:12] field. And so, you know, I'm going to [17:14] leave you with a couple of thoughts. And [17:15] one of them is, you know, if you have [17:17] the right artificial intelligence uh [17:19] tool set um that can supplement what [17:21] revenue analysts do, you can take your [17:23] worst performing revenue analysts, hand [17:26] them this tool set um and they can [17:28] outperform your best performing uh [17:31] revenue analyst without the same tool [17:33] set. It's just that much of a gamecher. [17:36] Um and and we got some other thoughts [17:38] here on the screen. I'm not going to go [17:40] uh through all of them, but um anyway, [17:43] this is something that's that's that's [17:44] that's really important to us, something [17:46] that we're going to continue to uh [17:47] pursue, and I'm super excited about how [17:50] this is going to change not just our [17:52] organization, but actually our our [17:55] entire industry. Uh so with with that, [17:58] um we can open up for any questions. [18:01] Uh I I think your question was, are you [18:03] using competitive data real time as part [18:04] of your data set? The answer is yes. Uh [18:08] we actually pull competitive data on a [18:09] daily basis for uh facilities that are [18:12] um around, you know, within a certain [18:15] radius of each of our properties [18:16] depending on kind of the the population [18:18] density uh around those properties. But [18:20] we pull those we pull those rates uh [18:22] daily and in some in some cases may [18:24] maybe multiple uh times a day. we may [18:27] need to up that frequency um you know in [18:29] the coming future just because you're [18:31] starting to see more intraday uh pricing [18:33] changes uh with um with with some of our [18:37] with with some of our uh competitors and [18:39] the data by the way is actually gets [18:41] quite large really fast if you're just [18:43] pulling this all the time and we were [18:44] just I was I we were just looking the [18:46] other day and I'm like oh my goodness [18:47] this thing is this thing has gotten [18:48] really big we may have to start to [18:49] archive uh some of this stuff but it but [18:52] it's critical like to look at that [18:54] because it helps to uh influence inuence [18:56] uh kind of the trending that you're [18:58] expecting uh for for you know how how [19:01] price is going to influence your [19:02] property. Are we letting the LLM make [19:05] the pricing recommendation or are there [19:08] are there other say the last part [19:11] >> other algorithms? Um good. Uh I uh the [19:17] answer to the question is the LLM is [19:21] going to have flexibility to make [19:23] pricing changes within the constraints. [19:26] Um and I'm going to look at my tech [19:27] people. They're going to nod at me or [19:28] not nod at me. U within the constraints [19:31] that we set up for those agents. So in [19:32] other words, we we establish those [19:35] agents using I would say almost [19:37] historical models or algorithms to some [19:39] degree and they would drive drive that a [19:42] bit. But we do we we have to offer [19:44] flexibility for the the uh agents to [19:46] make some changes. [19:48] Get a nod on that. Yeah, a semi nod. I [19:51] said it kind of right. Other questions? [19:55] All right. Thank you everyone. Much [19:56] appreciated.