AI Operation Management

AI operation management involves optimizing and automating business processes using artificial intelligence to enhance efficiency, decision-making, and resource allocation, here is the real case for your reference.

What is AI Automation?

AI automation refers to the use of artificial intelligence (AI) technologies to automate various tasks and processes without the need for direct human involvement. Some key aspects of AI automation include:

1. Robotic Process Automation (RPA): Using AI-powered software robots to automate repetitive, rule-based tasks such as data entry, form processing, and administrative workflows

2. Machine Learning (ML) Automation: Applying machine learning algorithms to automatically identify patterns, make predictions, and make decisions without explicit programming.

3. Natural Language Processing (NLP) Automation: Using AI techniques to automate language-based tasks like document processing, customer service chatbots, and language translation.

The key benefits of AI automation include increased efficiency, reduced errors, 24/7 availability, and the ability to scale operations.

By automating repetitive or tedious tasks, businesses can free up human workers to focus on more strategic and creative work. AI automation is rapidly being adopted across various industries, from manufacturing and healthcare to finance and customer service.

Use cases for Finance team

1. Finance Ticket Triage

This demo shows how Krista can help answering questions related to invoicing. In this demonstration the user asks an outstanding invoice payment question to Krista.

Krista is able to understand the context of the question (Intent mining) and classify the type of question using its propriety machine learning (ML) capabilities. Krista then extracts relevant answers based on the corresponding process flow, connecting to downstream systems.

document repositories, and other workflows incl. humans in the loop to orchestrate the desired outcome. The user can ask questions pertaining to invoice amount, category of invoice, number of invoices from a specific vendor, etc.

In case Krista is unable to answer a question based on the available data she will inform the user and can be configured to handle this via an exception process e.g. raising a ticket with the finance team. Finally, Krista will manage the ticket workflow (assign to an agent, close the ticket, etc.) and archive the original mail.

2. Finance Spend Data Classification

This demo video shows how to perform advanced data classification techniques using different models, including machine learning (ML) and large language models (LLM).She can parse the question and understands the context (intent).

In the first example (ML), Krista leverages a large body of data based on both historic classifications as well as industry data like the UNSPSC database for the classification.

Krista is able to use its built-in classifiers to accurately detect the UNSPSC Family / Segment / Class levels.

In the second example (LLM), where not enough training data exist to perform accurate classification, Krista will use LLM’s to aid in the classification. She will ask the LLM to categorize the line-items. She will then perform a second pass to validate the results against a different LLM.

AI automation then provides the employee with the appropriate answer including the obtained classification.

3. Finance Email Read and Respond

This demo shows how Krista can help accelerate how emails are handled by customer service organizations. An email is initially send to Krista asking to provide a duplicate bill.

A common usecase for e.g. utility companies. Similar ask from customers can be found all across. Just think about simply asking for when an order going to be shipped or delivered, when the next appointment is scheduled, a copy of the last invoice, etc, etc.

Krista is able to understand these inbound emails and classify the intent of the emailer.

In this case the duplicate bill will be handled as such meaning that the relevant information has to be extracted from the email and the erp will be queried to obtain the appropriate bill to be send back to the customer. In the second instance an email is sent to Krista that she is unable to classify with sufficient confidence.

To address this, she asks a person for input which provides her with an understanding on how to handle the particular email and teach her how to handle similar emails in the future. While handling both emails Krista kreeps track of what she is doing in SFDC by creating cases and updating them even including the confidence scores and intent classifications she is producing.

Use cases for Customer Services team

1. Multi Lingual Chat via Mobile

This demo shows Krista is truly omni channel and capable to deal with a variety of customer interactions in whatever language the come. Krista has been embedded in an example mobile application.

This app is not the equivalent of the Krista Client on mobile but rather a showcase that Krista can be easily embedded in other platforms and channels. The mobile application depicted here has a Krista Chatbot embedded within it.

Customers could embed the exact same Krista Chatbot if the configurations for the chatbot are suitable or integrate directly on the Krista Chatbot APIs.

The first messages send by the user states ‘Good day’ in German and Krista easily understands this is a way of engaging a new conversation. She replies cordially with a joke. The tone and temperature of the interaction by Krista, it can be fully configured and if desired these can strictly formal. Next the users asks, in Dutch what the weather is like of there. Again, Krista understands the intent of the user and replies with a cordial joke. Finally, the user asks about the supported android version.

Krista understands this question and recognizes she is unable to provide an answer to this query and responds that she will as a person to join the chat. Note that Krista is able to deal with multiple languages even in a single session when engaging with people.

When customer or users engage with Krista via any omni channel interface the following (generally) occurs.

  1. Krista receives the message from the given channel
  2. Krista establishes the context for this interaction including intent, customer specific context information, language, sentiment, etc.
  3. This can include prior interactions in the same conversation.

3. Krista acquires interaction specific data e.g. manuals, product information or data in other connected systems

4. Krista uses available context and data to generate a response.

5. Krista assesses the response and determines if she was successful in producing a response

6. Based on assessment Krista will either sends the response back to the user/customer

7. Notify the user/customer she was unable to understand or reply properly and ask an agent to join the chat.

Use case for Sales, Business Operation team

This demo illustrates how Krista’s AI features and process automation capabilities can be used to automate the completion of lengthy and complex GRC assessments.

Krista can dramatically reduce the amount of time necessary to complete what is typically a difficult and tedious manual process.

In this demo Krista is asked to complete a SIG Lite assessment, an assessment often used in a business’s Third-Party Risk Management (TPRM) program.

Krista will use a provided collection of evidence to complete the assessment, For this example, Krista is using a set of evidence that includes SOC2 and other reports, internal security policies, and previously completed assessments. Note that Krista can use a very wide variety of documents and data sources as the evidence she uses to complete assessments like this.

Krista accepts the incoming evidence, prepares a new copy of the SIG Lite assessment, and begins to complete the assessment, responding specifically to each question. Upon completion, Krista can optionally compare the newly completed assessment to a control or benchmark version (a version that represents the expected results) and identify deviations from expected results.

These findings can then be processed in any number of ways – loaded into a case management or GRC solution for follow-up, sent to the vendor who provided the evidence, and/or informing the analyst responsible for the assessment. Note that a wide variety of findings criteria can be applied, too – low confidence scores, missing evidence, etc.

Contact us