Entreprise Information Retrieval
The Challenge: Lack of Expertise Visibility
A large company in the construction industry with 100,000 employees and several hundred distinct entities across multiple countries had accumulated over 25 million documents across various ECM systems, CMS systems, intranets, file systems, Cloud-based documents, internet pages, and more. Their systems were sustaining close to one million queries per month, or up to 60 requests per second.
The company needed a focused application that would enable HR and project managers to select employees according to different requirements located in various systems, including structured applications and unstructured documents.
The employees created, and edited documents related to projects, methodology documents, and meeting minutes. The system needed to extract skills and experiences from this unstructured content and attach the documents to their author, while also considering the varying weight of the extracted skill, based on the density of the concept in each document.
The Solution: Getting to Expertise Faster and More Productively
Using our Enterprise Search technology, our team built a focused application for this company by combining a knowledge management approach with an expertise management system.
In addition to integrating all of the company’s information together, the application also required an enrichment process for the documents using the additional metadata collected on each employee to improve the accuracy of results. This allowed a very granular, filtered search. For example, the application could find a senior field engineer located in Texas with a strong expertise in a specific specialty who had worked on more than five projects related to that specialty and who was available in March or April.
Currently, the application is used in a variety of contexts:
- an employee directory that unifies all employees and sub-contractors across regions and organizations into a single portal with a very high level of pertinence due to semantics, indexing, and application federation;
- a company and sub-contractor directory with enriched metadata originating from company house and internal databases;
- expertise management software combining self-declarative data, resume descriptions, HR management, and document production; and
- knowledge management that leverages author profiles.
The Results Speak for Themselves
The directory of employees and their related expertise generates more traffic and requests than any other information management application. Progressively, this application, with its dynamic categorization capacities, is replacing various ECM systems.
The employee portal now enables a user to search over a wide range of criteria, including an employee’s past projects, current projects, most recent publications, company resume, languages spoken, colleagues, communities, and more. HR managers use the platform to enable users to search for job opportunities, submit skills requests, communicate internally, and more.
This advanced application enabled users to define alerts based on their own requirements, like a project request for a specific geographic region, a document on a specific issue written by an expert, or a job proposal that meets specific criteria.
The research center for an international group specializing in environment management across 8 business units. This entity was created through the merger of separate division research centers. There were a number of problems to solve:
- Pooling highly heterogeneous applications (originating from several different entities) while maintaining the terminological and lexical specificities of each division. With an attempt at unification in the scope of a single DMS having partially failed and each entity wanting to keep its own tools, the search engine made it possible to pool these various tools without forced migration.
- Exchanging information among the various groups while maintaining a high level of security and confidentiality. We therefore developed a search module capable of notifying the user that a document exists without granting him or her access to it, but rather enabling him or her to submit an access request to the owner of the document resource.
- Defining a classification system capable of pooling the resources common to different divisions while setting apart documents only of interest to one of the fields.
- Creating a directory of human expertise on key and sometimes cross-disciplinary subjects by analyzing the document production of the authors.
- Pooling external document resources thus making it possible to avoid multiple purchases.
The application has now been in use several years and new sources are being integrated into the project.
The French homeland security agency manages several million records on judiciary proceedings, including items, persons and operating modes. These records mix a highly detailed information structure – with over one hundred attributes for each type of item – and full-text for each content. The system must be able to guarantee service of over 200 extremely complex requests per second.
The homeland security organization needed a focused application to search for this content and navigate through it. And to harness the full power of their structured information, they needed a powerful search engine.
Using our Enterprise Search technology, a large systems integration firm built a custom administration console, combining the full-text search capabilities and nested document support provided by our platform to allow the user to formulate highly complex queries.
In addition to the standard indexation process, the application also included an enrichment process for documents using semantic analysis in order to extract structured information from free text and identify similarities between operating modes.
✔ Accelerate access to information
✔ Combine structured and unstructured search
✔ Distributed environment permitting very high usage
Managing expertise is a very common requirement that is not addressed effectively in many corporations due to the constant evolution of expertise domains, their lack of normalization and the absence of metadata attached to employees in most systems.
To effectively address this requirement, the system needs to learn automatically from incoming information in real time and on a continuous basis. The information concerning expertise requirements (customer case, request for project, tenders, crisis management, etc) and expertise resources (FAQ, documents written, emails, social media profiles, resumes, projects to which an employee participated, minutes of meetings, blogs, etc) is fragmented and siloed.
bee4sense Learning Engine analyses all information originating from an employee or from external requirements where expertise can be detected. The system leverages its semantic capability to cluster concepts extracted from these applications and create relevant clusters of expertise adapted to the business jargon of the company, all the while taking care to avoid dupes, as an expertise can be described through several different terms.
Expertise identification is updated continuously based on the appearance of new information. The classification system adapts automatically to emerging domains of expertise so that the company can identify new domains of expertise to be addressed and identify others where the company lacks resources.
The continuous self-learning capability of the system guarantees that the expertise taxonomy never becomes obsolescent, even in an ever-changing environment.
Lastly, the system can also associate a user’s expertise with his or her immediate availability, by displaying the user’s Lync or Skype status, as well as his or her work schedule obtained from project management applications or calendars such as Exchange, Google calendars or Notes.
✔ split-second expertise identification to ease solving a customer or employee issue
✔ benchmarking existing expertise resources against incoming requirements to ensure prospective HR strategy and business sustainability
✔ knowledge and best practice dissemination to speed employee ramp-up and business impact
✔ ensuring that expertise available in the company is leveraged to maximize revenue impact
Large organizations usually manage a large portfolio of products or services, across multiple business units and product ranges. No single user can acquire the expertise needed to manage product enquiries effectively and potentially sell. In addition, depending on the product, the user has to deal with a variety of characteristics, sometimes very technical, to be able to recommend the right product or service.
Enabling customer-facing employees to benefit from any customer interaction to maximize revenue is a critical success factor. The non-ability of most companies to address upsell or cross sell opportunities is real.
The reason why current systems fail to deliver such expertise to users is related to the heterogeneity of internal product management systems (technical documents, FAQs, sales brochures, white papers, multiple product catalogs, pricing databases, ERPs, order management systems, quote management systems, etc) and also external information (social media, competition, etc). Such heterogeneity increases complexity.
To effectively address this requirement the system needs to be continuously up-to date and the GUI needs to be self-adaptive so that options and selection criteria evolve as the user refines requirements. In addition, as companies launch new products, new ranges as norms and constraints may shift, it becomes imperative to automate taxonomy upgrades. Lastly, it is essential to leverage internal business jargon to customer requirements that may originate from reviewing a product from a competitor (“do you have a product or service like…”).
The bee4sense Learning Engine analyses all information originating from product related environments, applications, social media, competitive intelligence to maximize customer-facing employee efficiency. The system handles a holistic vision of the product characteristics, permitting the swift and accurate matching of customer requirements to internal products by leveraging its semantic and taxonomy management layers. Lastly, the contextual GUI adapts dynamically to the category of products handled.
✔ enable customer facing employees to effectively respond to customer product enquiries to maximize expertise and sales potential
✔ allows product suggestions to be adapted to customer profiles
✔ leverages all product information to maximize relevancy
By leveraging data enrichment and search, bee4sense provides a unique approach to helping users focus on critical information and proactively choose the right strategies to maximize performance. In the following example, opportunities originating from the CRM application are benchmarked using the predictive engine and pushed to the user using ActionAlerts™, helping the user avoid missing critical windows.
✔ use a KPI to prioritize opportunities
✔ challenge forecast from experience
✔ disseminate best practices to maximize success
The French DCPJ (Direction Centrale de la Police Judiciaire) handles several million pieces of heterogeneous content, in different formats and originating from different data sources. Because of the highly time-constrained nature of their work, it was imperative for them to be able to process these documents as efficiently as possible. The time required to search across this corpus and extract valuable information was simply too high, particularly considering the increased responsiveness required in the judicial context.
After having trained customer resources, the organization built a platform combining relevant and performant search capabilities with Artificial Intelligence based enrichment processes. The application leverages embedded natural language processing capabilities to extract structured information from unstructured documents. This information consists of different types of named entities linked together in order to reveal relations networks across entities.
On one single platform, the customer is able to index the documents, extract all the value from the available information and access this value faster.
✔ Accelerate access to the right information
✔ Reduce the time required to recover available information from unstructured texts
✔ Improve information discovery by combining search capabilities and graph navigation
The digitalization of processes has transformed complaints management. In response to increased e-fraud, the French homeland security agency implemented a self-service declaration system. It then faced a huge amount of complaints, which had to be processed in a timeframe compliant with the investigation requirement.
This agency now implements a solution based on the bee4sense platform, enabling them to automatically process this vast quantity of complaints and identify similarities between them. These similarities allow declarations to be grouped together in batches, which helps investigators focus on relevant operating modes. The application also provides network graphs to better understand the links between complaints.
✔ Automatic processing of huge volume of data
✔ Similarities between complaints detection
✔ Tailored display to help the investigation
One of the largest electrical goods companies in the world with 2,500 customer care agents in 100 countries, has recognized that to stay competitive, you need to not only over-deliver on customer expectations, but also find new opportunities to add value to customers. The organization is dedicated to delivering outstanding service and is taking an innovative approach to customer care excellence with a customer lifetime value-driven approach.
The challenge: delayed customer identification time
The leaders of the company realize that at the heart of the most important, and often the only, interaction point with customers, customer care agents are a true asset in establishing rapport with customers and in increasing the company’s bottom line.
The company is now in the process of re-inventing their support organization. In the past, improvement initiatives were handled one at a time: one application at a time, one process improvement at a time. This led to silos in technologies, processes and the ability to quickly find important customer information. All this also hindered the utilization of this information to make the most relevant decisions and recommendations, in the field, during customer calls.
The average time customers spent on the phone with the live agent was over 7 long minutes, of which 5 minutes was spent waiting for the agent to find customer information, and to enter basic information into their support application. This activity wasted the majority of the valuable time the customer is willing to spend on the call, instead of utilizing it in the most satisfactory and productive way possible.
The goal: increase customer lifetime value through reduction of customer identification time and product upsell
The new goal is to reduce customer identification time in each interaction by being able to quickly map the incoming number with the full customer profile, enrich this information intelligently, and suggest the most appropriate actions. This frees up time for a customer care agent to become ‘solution consultants’ for the customer, thus not only increasing overall customer satisfaction but also generating additional revenue from relevant upsells and promotions.
The challenge required the company to further their innovation in this field, and they have partnered with bee4sense to bring its Information Insight Platform into their ‘Big Data Technology Innovation Lab’, in conjunction with Accenture. A well-planned phased approach ensures that each use case is incrementally deployed, brings quick value to the business, and that new use cases build on top of the previous one.
The solution: technology-enabled approach to personalize each customer care interaction
bee4sense is working with the company’s customer care and technology innovation teams to enable their agents to be more proactive in the following context:
The “Welcome” Phase
- reduce the time required to identify a pre-existing customer or contact;
- reduce the data-entry workload for creation of new contacts and customers; and
- ensure accuracy and quality of information of current and new contacts and customers.
The “Resolution and Recommandation” Phase
- reduce the time needed to understand the requirement of the contact or customer;
- ease access to the right information by shortening the research cycle;
- deliver high-quality recommendations for problem resolution; and
- identify potential need and recommend product upsell.
The “Closing” Phase
Automate recording of the customer case to avoid time-consuming manual entry.
The new, bee4sense-powered customer care solution enabled increased productivity of customer care agents by decreasing the overall time-to-resolution, as well as increasing time spent on selling of value-add services.
✔ lowered call duration by 60%
✔ increased resolution rate by over 100%
✔ increase customer satisfaction by 60% (as determined through surveys collected after the case closing)
✔ increased upsell by over 10%
✔ reduced churn by over 5%