As of March 2022, the average employee spent nearly four hours each day searching for information to answer a question about anything from PTO to a bug in their computer. Employees who submit IT quests often have to wait for days before a support team fulfills their request. 58% of employees attribute this excessive time spent searching for information to the overwhelming number of SaaS applications they navigate. As of 2023, businesses used an average of 371 SaaS applications. Even when the needed information is available, the volume of irrelevant resources leads to information overload, making it difficult for employees to find what they need.

AI-powered chatbots in the workplace have grown to become a common solution to this problem as AI can identify common themes and trends from employees to best address their questions. A 2023 study found that AI can help employees of an organization save up to 20 hours per week in internal communications. With the integration of natural language processing (NLP) and machine learning, as of October 2024, chatbots recognize and interpret a wider range of human language and intent. They can provide answers by connecting with backend systems of the business, recognizing who they’re talking to, and providing personalized support. AI chatbots could help modernize internal communication and improve operational efficiency by enhancing access to information and reducing administrative burdens.

Moveworks uses AI to automate workplace support tasks, helping companies resolve employee issues like IT or PTO questions without having to rely on employees. Its AI Copilot integrates with dozens of apps that employees may already use like Slack and Microsoft Teams so they can get their questions answered instantly, without interrupting their workflows. As of October 2024, Moveworks serves over 350 enterprise customers including Salesforce, Palo Alto Networks, Pinterest, and DocuSign.

Source: The Software Report

Moveworks was founded in June 2016 by two-time ex-founder Bhavin Shah (CEO), Varun Singh (President), Jiang Chen (AI CTO), and ex-founder Vaibhav Nivargi (CTO).

Growing up in Silicon Valley during the 1990s, Shah was influenced by the region's abundance of technology giants and an “influential” breakfast he had with Steve Wozniak at Apple in the sixth grade, which instilled in him “the belief that technology would be central to [his] career and that the concept of starting a company seemed almost natural.”

Shah began his career, after receiving his master's degree in education, technology, and business from Stanford, when he joined Leapfrog as the director of new business development in July 2000, helping scale its revenue from $30 million to $450 million over five years. Leapfrog was an educational entertainment developer for children.

In 2002, Shah was recruited by the US government to help educate Afghans on health basics, which inspired him to launch Gazillion Entertainment, an educational game company, in 2005. His experience in Afghanistan sparked the idea of a digital briefing book that could prepare users with key insights, leading him to found Refresh.io in 2011, a tool for salespeople acquired by LinkedIn in 2015.

Singh, meanwhile, started his career in 2010 at Sefaira, an energy-efficient design software. While there, he built and launched three software products including Sefaira Concept: the world’s first web-based building design analysis software. In 2014, Singh started a new role as a lead product manager for recruiting products at Facebook. He led the development of an NLU-driven bot in Facebook Messenger, that was used by several thousand Facebook employees to streamline the hiring process.

Chen started his tech career in 2008 at Yahoo! In December 2009, he joined Google as a staff software engineer where he was a tech lead on the Open Domain question/answering project. The project went live in 2014 and continues to account for a “large portion of all searches, including Google Search and Google Home. In 2015, he continued his search-focused machine learning career at Airbnb, having improved user booking recommendation algorithms on the platform.

Nivarghi began his career in 2007 as a software engineer at Aster Data, which was acquired for $263 million by Teradata in 2011. He then founded ClearStory Data, a cloud-based, interactive data analytics and visualization product that was acquired by Alteryx for $19.6 million in 2019.

The team found that there was a pressing, unsolved problem of inefficient IT help desk processes. When employees needed help with simple troubleshooting or had questions, they’d have to wait hours or even days to get their issue resolved, hindering their ability to perform their jobs.

Motivated by the belief that “no employee should have to wait for help” through a user-friendly self-service experience, the team co-founded Moveworks in June 2016. Shah mentioned that a key driver in their decision to start Moveworks was the rise of Slack. As “people were spending hundreds of hours a month” in internal collaboration tools, the team believed that these tools would be a “perfect venue in which to give people instant support.”

Moveworks is an AI-powered platform designed to automate and streamline employee support in large organizations. It uses natural language understanding (NLU) to resolve issues by handling requests related to IT, HR, and other business functions directly through chat or messaging platforms like Slack and Microsoft Teams. By integrating with an organization’s existing systems, Moveworks can process and resolve common employee questions, automate repetitive tasks, and help reduce response times, improving both efficiency and employee experience.

Copilot

Source: Moveworks

Moveworks’ core product is a generative AI chatbot copilot designed for enterprise use. The Copilot enables enterprise employees and end customers to engage with a single, conversational interface that automatically pulls in data from an organization’s business apps such as Slack and Microsoft Teams. Users can retrieve the precise information they are seeking, without leaving their current workflow.

Copilot pulls in all the resources that an enterprise has across its systems into its Enterprise Cache. These documents are broken into snippets, and the product’s Semantic Match functionality enables it to intelligently search the Enterprise Cache for the most relevant information to return to the user, in seconds.

Source: ZDNet

Expansion Beyond IT Use Case: In March 2021, Moveworks expanded beyond the IT support use cases to provide modules for HR, finance, facilities, and employee communications. Nivarghi noted that IT was where Moveworks validated its product, but that it “extends very naturally to other corporate functions.”

As of October 2024, Moveworks also documented use cases for marketing, sales, and engineering teams. For example, marketers can use Copilot to retrieve metrics on their ongoing social media campaign or generate ad copy for a project. HR teams can deploy Copilot to immediately answer employees’ insurance coverage questions or onboard new employees. Alternatively, finance teams can seamlessly adjust their accounts payable after an invoice has been paid or instantly retrieve information from financial statements.

Enterprise Search

Launched in October 2024, Moveworks' enterprise search feature aims to provide organizations with a tool to locate and retrieve information across multiple internal systems and knowledge bases. Unlike traditional search functions that have a specific database or type of information, this search capability is designed to integrate across various platforms and applications, offering a unified view of information stored in cloud services, intranets, document repositories, and internal databases. This approach allows employees to access necessary information directly from their primary communication tools, such as Slack and Microsoft Teams, where Moveworks is embedded, enabling access to organizational knowledge.

The enterprise search feature is structured to use natural language processing (NLP) to interpret employees' questions and locate relevant information, whether it’s in the form of documents, tickets, FAQs, or other content types. Moveworks' use of NLP means the search can identify and deliver specific answers or documents, even if the language used in the query differs from the way the information is stored. This functionality helps reduce redundant support tickets and minimizes time spent searching for resources, aiming to improve overall productivity by centralizing access to the knowledge employees need across the organization’s ecosystem.

Service Management

Moveworks' service management product is built to assist organizations in automating and streamlining their internal service request workflows. The tool is designed to work with established service management systems, such as ServiceNow and Jira, by integrating directly into workplace messaging platforms like Slack and Microsoft Teams. This integration enables employees to make requests and receive help within the chat platforms they already use daily, minimizing the need to switch between applications. By automating certain repetitive or predictable tasks, the service management product reduces response times for issues commonly handled by IT, HR, or other support teams, helping to maintain service levels with less manual intervention.

MoveLM: MoveLM, Moveworks’ proprietary LLM that is a part of its service management product and released in 2023, was designed explicitly to understand and solve the unique problems that enterprise employees face. It was trained on over six years of proprietary Moveworks data spanning a variety of enterprise departments such as IT, HR, and employee service. This training set included 500 million support tickets, 14 million bot conversations, and 400K knowledge base articles.

In addition to its proprietary data, MoveLM was also trained on over 500K instruction examples from public datasets including Dolly, Natural Instructions, and OpenAssistant. This was done to “increase [the model’s] exposure to terminology and task formats that may be outside the scope” of Moveworks’ owned data.

Source: Moveworks

Once MoveLM was trained and fine-tuned, its performance was benchmarked against OpenAI’s model, GPT-3.5 Turbo. It was evaluated on its ability to correctly choose the correct enterprise API to use within a response and its ability to provide helpful responses to users (summarize API outputs in a format that is clear and useful to the user).

The results indicated that MoveLM (77% of all cases) outperformed GPT-3.5 Turbo (55% of all cases) in its ability to accurately choose which enterprise API to call for a given user inquiry. MoveLM also returned helpful responses to users 75% of the time while only 66% of GPT-3.5 Turbo’s responses were classified as helpful.

Source: Moveworks

Provision Management

Moveworks’ Provision Management product is designed to streamline the access provisioning process across an organization by using AI to automate requests for access to various tools and applications. Moveworks’ approach aims to eliminate delays by automating the approval and configuration process. With this system, employees can obtain the resources they need without manual intervention, helping maintain operational efficiency and reducing the load on IT teams.

Provision Management leverages generative AI to handle access requests from submission through approval, reducing the wait times typically associated with setting up software or tool access. It integrates with an organization’s existing IT infrastructure to ensure that permissions are managed according to security policies while minimizing time spent on administrative tasks. The goal is to allow employees to securely access essential tools like requesting PTO or resetting passwords, all through an autonomous system.

Employee Experience Insights

CIOs face the challenge of enhancing both operational efficiency and employee experience without a clear understanding of the obstacles their employees encounter. This lack of insight often hampers their ability to proactively address issues and improve productivity. Recognizing this need, Moveworks introduced Employee Experience Insights (EXI) in October 2022. According to Shah, “every help desk is sitting on a treasure trove of information that could prevent future issues before they happen.”

EXI empowers enterprise administrators by providing a complete view of the entire employee experience on a single screen through its Employee Issues Map (shown below). It condenses issues into a single, easy-to-navigate bird's eye view. Singh stated that “EXI analyzes unstructured data in tickets and gives CIOs clear insight into the exact drivers of poor employee performance, without the need for manual data tagging.”

Source: Moveworks

Users can delve deeper into specific issues by selecting a specific square on the Employee Issues Map. The pop-up window displays a breakdown of each type of issue commonly present with employee requests that relate to the issue. When discussing his responsibility to empower the digital workforce, Databricks CIO Naveen Zutshi stated that “with Moveworks’ sophisticated NLU and Employee Experience Insights, they have meaningful and actionable data to turn thousands of requests into an efficient plan of action — and a best-in-class employee experience.”

Source: Moveworks

Knowledge Studio

Source: Moveworks

Knowledge Studio enables users to generate articles based on employee demand, ensuring that the content is relevant and useful. Shah explained that “companies have content all over the place and when there’s a gap in content already created, you have to give them a helping hand.” The system can retrieve information from disparate sources and create a document draft that is 85% to 90% complete.

The system recommends article topics by mining submitted tickets from across the enterprise to identify the most frequent issues reported by employees — making it easier for users to identify and address knowledge gaps. These issues are displayed in the Knowledge Studio interface which users can click on to begin editing the draft generated using employee conversations with the Copilot.

When a user selects a document to edit, they can make manual edits to the text or submit natural language prompts within an input box to further refine the article. Complete citations from employee Copilot conversations that were used to generate the article draft are also displayed.

Customer

In a July 2024 interview, Shah stated that Moveworks aims “to make work easier for employees, and act as the bridge to information locked in systems across their organization.” Moveworks was purpose-built for large enterprises, evidenced by its list of case studies done with enterprise customers across many industries. Shah emphasized that “what they’ve built is applicable to almost any business in the world”, but it becomes truly valuable when an organization reaches a size at which information is spread across many systems, with different access privileges provided to each employee.

According to a Moveworks executive, as of October 2024, Moveworks targets large enterprises with at least 2K employees. Specifically, organizations that seek to immediately resolve commonly repeated questions that would otherwise require human attention.

Additionally, a former Moveworks leader emphasized that Moveworks “was most effective with companies who had a vision or commitment to supporting its employees and [with] companies that cared about the employee experience.” The individual also emphasized that the COVID-19 pandemic drove many enterprises to invest in productivity technology with the intent of making the transition to remote work as simple as possible, saying that “Moveworks was able to attach itself to those bigger employee experience-related initiatives rather than trying to sell based on cost savings alone”.

While enterprises are the target customer for Moveworks, there are two distinct end users of its product. In addition to enterprise employees, a second end user is the enterprise’s end customer. One Moveworks customer at a large data analytics and risk assessment firm implemented the Copilot to enable the firm’s own clients to get their questions resolved quickly. While the firm only had 12 humans on the IT support desk, the team was fielding over 100 support requests per day, which meant the amount of time required to process a client’s request was probably longer than desired. Upon implementing Moveworks’ Copilot, the support desk only needed to manually fulfill 10 to 20 requests daily which saved the firm from needing to hire an additional 50 human support desk agents.

Additional Use Cases

Palo Alto Networks is a cybersecurity protection software company, with about 14K employees as of 2023. The company doubled its revenue from $3.4 billion to ~$6.8 billion between 2020 and 2023 and needed to support this growing demand by hiring employees globally. Consequently, its help desk needed to provide support at a much broader scale. In April 2020, the company deployed the Moveworks Copilot to its entire workforce.

This implementation was a part of FLEXWORK, Palo Alto Network’s vision for the future of work designed to empower employees to determine how they collaborate while receiving personalized support at work. Former Senior Director of Digital Experience Management at Palo Alto Networks, Nicole Tate-Pappas, emphasized that prior to the pandemic, Palo Alto Networks had limited support options, but knew that “technology would play a critical role in meeting people where they are.”

The Moveworks chatbot, private-labeled as ‘Sheldon’ for Palo Alto Networks employees, played a role in enabling FLEXWORK. Tate-Pappas continued, “[Sheldon] doesn’t just file tickets — he takes action and solves issues automatically, just by asking Slack. It’s an experience our people rely on.” This resulted in 145K interactions with Sheldon, over 90% of employees having used Moveworks, and 351K hours of productivity saved.

Separately, Mercari is a Japanese online marketplace that connects millions of people across the US to shop and sell items to others. Its simple user interface that appeals to the everyday shopper and casual seller resulted in the company’s net sales more than doubling from $519.9 million in 2020 to $1.2 billion in 2023. Due to this growth, “Mercari’s US operations faced the challenge of having enough support staff to keep up.” In July 2021, Mercari US deployed its Moveworks Copilot, which helped employees solve a variety of problems such as resetting passwords, editing email groups, troubleshooting devices, and more.

Before Mercari used Moveworks, “employees would directly Slack the IT team for support, which posed productivity challenges. [With Moveworks], the IT team can automate workflows, and employees can receive help without leaving Slack.” The Copilot has reduced ticket volume by 74%, as 94% of employees use the Copilot first when they have questions. Subsequently, 8K hours of routine IT work are saved per month.

Market Size

The global conversational AI market size was valued at $7.6 billion in 2022 and is anticipated to reach an estimated $41.4 billion by 2030, reflecting a 23.6% CAGR between 2023 and 2030. The chatbot segment comprised 67% of the total global conversational AI market in 2022.

The global AI chatbot market size had an estimated value of $6.4 billion in 2023. 58.2% of the market is comprised of solution offerings, which represents Moveworks’ addressable market of $3.7 billion. It’s anticipated to grow at a 26.4% CAGR between 2024 through 2033, to reach an estimated value of $66.6 billion in 2033.

Source: Market.us

There are three primary drivers of this market. First, the heightened customer expectations for immediate responses to their questions. Two-thirds of respondents from a 2023 customer patience study say speed is as important as price. As a result, companies have sought to automate customer support with chatbot technology to instantly fulfill customer inquiries.

Second, not only do companies need to provide instant access to desired information, but the response provided must be tailored to the customer’s specific issue. As of July 2023, “71% of customers expect businesses to provide personalized experiences, and a staggering 76% feel frustrated when this expectation goes unmet.”

A side effect of providing instant access to desired information is reduced employee churn. A 2021 study of 1K respondents found that 49% of US workers said that they are likely to leave their jobs if they’re unhappy with the technology they use at work. This underscores the growth in demand for chatbots in an enterprise setting, as they “reduce the time required to find the information you need to get your job done.”

The third primary component driving the AI chatbot market is the desire to reduce support desk spend. As of 2022, it was forecasted that conversational AI would reduce contact center agent labor costs by $80 billion by 2026. Labor costs represent up to 95% of total contact center costs, and conversational AI chatbots could automate all or part of a contact center customer interaction.

Incumbents

ServiceNow: Founded in November 2003, ServiceNow went public in 2012 at a market cap of $2.1 billion and offers a cloud-based platform that automates and streamlines work processes across organizations. Like Moveworks, ServiceNow sits on top of customers’ key automation incumbent platforms. While ServiceNow’s offerings span beyond AI chatbot, Now Assist for Virtual Agent was released in June 2023 to create conversational experiences for intelligent self-service. While this offering is competitive to Moveworks, a former executive indicated that ServiceNow’s Virtual Assistant is offered at a lower price point than Moveworks but provides a lower ticket resolution rate and requires more manual fine-tuning than the Moveworks Copilot does out-of-the-box.

Box: Box was founded in 2005 and went public in January 2015 with a $1.6 billion market cap. Its core product is the Intelligent Content Cloud, designed to enhance how organizations manage and utilize their content (sales presentations, product spec sheets, employee handbooks). It features Box AI, released in May 2023, which parses documents and produces new content. The conversational AI chatbot within Box AI is similar to Moveworks’ core product as it summarizes text from documents to find specific details about things like system update information or contract expirations. The chatbot can also generate new text for resources like social media posts and engineering spec sheets, which the Moveworks Knowledge Studio is also capable of.

Startups

Aisera: Aisera was founded in 2017 and offers AI Search, a product enabling employees to search for contextually relevant information across all business apps and knowledge bases within the enterprise. Respective users of Aisera and Moveworks reported that both companies are user-friendly, but Moveworks’ initial setup experience is simpler. As of October 2024, Aisera had raised $180 million in total funding. It was valued at $638.3 million pre-money in March 2022 after raising a $90 million Series D co-led by Thoma Bravo and Goldman Sachs.

Espressive: Espressive was founded in 2016. Its core product is an autonomous virtual agent that enables employees to get their questions answered quickly. Espressive raised a total of $53 million in funding as of October 2024. It raised a $30 million Series B in 2020 led by Insight Partners. Other investors included General Catalyst and Wing Venture Capital. In May 2023, it was noted by a former executive at Espressive that Espressive tended to focus on SMBs and had robust out-of-box capabilities, despite lacking analytical reporting features. Moveworks contrasts with Espressive due to its focus on the enterprise customer. The same executive also mentioned that Moveworks’ dashboard is more robust than Espressive’s.

Kore AI: Kore AI was founded in 2013. In December 2023, it was valued at $650 million after it raised a $150 million Series D round led by FTV Capital with participation from NVIDIA. As of October 2024, Kore AI has raised a total of $223.5 million. One of its core products, Search AI, is similar to the Moveworks Copilot. Kore AI has more customization optionality to hyper-specific use cases but requires more fine-tuning out-of-box.

Moveworks uses a per-user pricing model on a custom quote basis. A former executive at Moveworks indicated in October 2023 that Moveworks charged per employee who had access to the product. The price per employee typically costs around $5 monthly. This can fluctuate based on how many modules (modules for IT requests, HR, finance) the customer purchases and decreases as the total number of employees to be onboarded increases. Additionally, the “baseline minimum” deal size for Moveworks is about $75K.

In November 2023, Moveworks was recognized on the 2023 Deloitte Technology Fast 500 as one of the fastest-growing technology companies in North America, having achieved an 860% growth rate. Moveworks had 447 employees as of August 2023 and increased to about 600 employees as of October 2024. As of October 2024, Moveworks serves over 300 companies across various industries including more than 5 million employees using its Copilot. As of October 2024, notable Moveworks customers include Salesforce, Slack, Pinterest, Broadcom, DocuSign, Stitch Fix, Autodesk, and GitHub.

In June 2021, Moveworks was valued at $2.1 billion when it raised $200 million in a Series C round led by Tiger Global Management and Alkeon Capital. This reflected a ~5.4x up-round from its $390 million Series B post-money valuation in November 2019. As of October 2024, Moveworks has raised a total of $315 million from investors including Bain Capital Ventures, Lightspeed, and Kleiner Perkins.

Despite its June 2021 financing occurring during a period in which valuations were historically over-inflated, Moveworks’ 860% growth in fiscal year 2023 along with direct competitor Kore AI achieving a 6.5X valuation step-up in its December 2023 Series D raise indicate that, should Moveworks elect to raise a new round of capital, it may be able to do so at favorable terms.

Copilot for Web

As of July 2024, Moveworks is expanding its core offerings beyond chat to include a browser-based experience. This web-based platform will allow users to toggle between two modes: assistant mode (the existing Copilot chatbot) and search mode. The search mode is designed to help employees discover internal content across enterprise applications through a single search bar with filters to quickly narrow down results.

An opportunity for scaling Moveworks’ web-based offering lies in its potential to serve new employee segments that typically lack access to enterprise messaging apps like Slack and Microsoft Teams. These segments include retail employees in brick-and-mortar stores, manufacturing employees on the shop or factory floor, field workers going door-to-door, and independent contractors.

Moveworks’ new web-based offering could enable employees to find answers to common shopper questions, check stock availability, submit purchase orders, access local field offices, and understand contractor-specific benefits. This expansion could increase Moveworks’ user base and impact. For context, as of 2023, there were approximately 4 million retail sales workers and over 12.7 million manufacturing employees in the US.

Scaling Data Centers Internationally

Moveworks has expanded its global footprint by opening new data centers in key regions addressing the growing demand for data privacy and residency. Nivargi commented that this move highlighted its “dedication to meeting the diverse and growing needs of its global customers.” In May 2023, Moveworks launched its first regional data center in the European Union (EU), which adheres to data residency requirements across the region. This move was subsequently followed by the July 2023 opening of a Canada data center, and most recently in January 2024, with the announcement of its Australia data center.

A 2021 report found that the number of countries that have enacted data localization requirements has nearly doubled from 35 in 2017 to 62 in 2021. The total number of data localization policies has also more than doubled from 67 in 2017 to 144 in 2021. As a result, many firms have been forced to physically locate data in the country where it originates to be granted permission to do business within these regions.

While entities including the EU don’t explicitly require data localization, it does limit data transfers outside the region. Moveworks’ establishing data centers within these regions enables the company to fulfill local data requirements within them while also providing customers with piece of mind that their data isn’t compromised as it travels across international borders.

User Base Expansion

In April 2024, Moveworks announced a partnership with Microsoft. This collaboration leverages Microsoft Azure which as Shah mentioned, “provides the stability, security, and privacy standards” necessary to deliver a top-tier Copilot employee experience. Moveworks also gains access to Microsoft’s flexible offerings, including the latest NVIDIA GPUs and Azure OpenAI Service’s provisioned PTUs, enhancing its ability to deliver cutting-edge Generative AI use cases.

Additionally, Moveworks Copilot is available through the Microsoft Marketplace as of October 2024 which allows customers to apply their existing Azure Consumption Commitments (MACC) toward its purchase. This integration presents a potential sales pipeline opportunity for Moveworks due to gaining access to Microsoft’s enterprise customer base.

Security Concerns

Despite the desire among enterprises to implement AI solutions, a recent survey revealed that 85% of over 1.4K executives plan to invest in generative AI. However, barriers to AI adoption persist, which include concerns around data lineage or provenance (61%), data security (57%), and regulation and compliance (53%).

To combat this apprehension, Moveworks prioritizes compliance with global privacy and security laws, masks all sensitive data that was used to train its proprietary MoveLM LLM, and anonymizes personal data required to serve customers. Despite this, concerns among CIOs and CTOs could continue to present an obstacle for the company.

Price Point

Moveworks charges a higher price point compared to competitive chatbot providers like ServiceNow, Espressive, and Aisera. As more natural language chatbots came out, a former sales executive at Moveworks indicated that “Moveworks tends to be on the more expensive side compared to the competitors”, which “developed a lot more challenges in the marketplace.”

These competitors often compete with Moveworks on price to win deals. This premium pricing strategy may pose a risk as it may drive potential customers toward more cost-effective alternatives.

As of 2023, 60% of CIOs were investing in employee experience to improve the productivity and performance of distributed workforces. This trend, combined with the 26.4% CAGR of the global AI chatbot market, the need to reduce employee churn by providing essential tools, and the desire to lower support desk overhead through automation, indicates that AI chatbots could stick around to help employees.

Moveworks is an AI-powered platform that helps companies solve employee support issues. It works inside messaging apps like Slack or Microsoft Teams, so employees can request help with things like IT issues, HR questions, or access to software directly through chat. Moveworks uses AI to understand these requests, handle routine tasks automatically, and connect people to the right resources, which speeds up response times and reduces the need for manual support from IT or HR teams. In short, it’s like having a virtual assistant that simplifies and automates support requests across an organization.

Moveworks’ Copilot product provides a more effective user experience due to its proprietary MoveLM model, which is built upon over six years of enterprise data from customers. This model is specifically trained to understand the challenges that enterprise employees face and has a versatile understanding of user background and prompt context. Moveworks needs to differentiate itself, even after its 860% fiscal year 2023 growth, as the enterprise chatbot market becomes more fragmented.