AI assistant
Web, internal users
2025
Context
Waves AI Assistant is the first implementation of AI-driven workflows on the platform.
The project was designed to establish the foundation for future AI capabilities and to encourage greater adoption of AI technology across the company’s employees.
It is built on the OpenAI API, it uses the GPT-4o model for standard responses and GPT-5 for more complex reasoning.
The first version of the Waves AI Assistant doesn’t link to the platform database, but it marks the start of a new AI chapter.
The Problem
Employees were experimenting with external AI tools, but workflows were fragmented and lacked consistency. The business needed an AI solution that could be integrated into daily work while setting the stage for adding more AI tools and agents into workflows. There was also a growing risk of employees uploading sensitive client material to personal AI chatbots.
Goal
The AI assistant is designed to streamline internal employees workflows and simplify daily tasks through powerful AI technology, supporting the client service team and other company members across various areas of their work. Another goal is to provide employees with secure access to AI tools, allowing them to work with client materials without compromising confidentiality or security.
Interview insights
👨 No restrictions for file sizes
Users asked for unlimited file upload sizes and no text character limits to support the processing of large briefs and RFP documents.
👨🏼🦱 Dictation
Users would like the ability to think out loud first and then have the AI assistant process and refine their thoughts as needed.
👩🏻 Suggest follow up questions
Users find follow up questions and suggestions functionality in other assistants very useful for adding additional context.
🧑🏽 Image & chart generation
Users would like to have image and chart generation, for example, based on provided data to copy into decks.
👩🏻🦰 Indication of sources
Users would like to see the sources of the information so they can cite them in the files and documents they create.
👨🏾🦱 Context memory
Users would like the Assistant to have good context preservation because they tend to work on materials for the same client.
👨 Prompt library
It was a business request to add a company prompt library to the AI assistant to increase adoption of the product.
👨 Chat grouping
Users have expressed that they value the ability to organize chats into groups, such as those dedicated to a specific brand.
Actionable points
• Build a dedicated page for the AI Assistant, accessible from the main platform.
• Provide standard chat functionality, such as the ability to create, rename, delete, and share chats.
• Explore the addition of voice-to-text functionality.
• Implement chat search functionality.
• Design UI elements that indicate which tools are being used to generate responses.
• Add functionality for uploading files of different format and size.
• Outline tasks that can be completed with AI and feature suggested prompts for each.
• Create a company-wide prompt library with functionality for employees to contribute and browse prompts.
Results
👍 Settings page
The settings page allows users to select their title from a dropdown list and choose the personality they want the assistant to adopt. This page is only shown for the first ever session.
👍 Onboarding
This page provides users with best-practice guidance and reassures them that their data is secure, helping to build trust.
👍 Landing page
This landing page features a text input field and prompt badges sourced from our prompt library.
👍 Prompt suggestions
When a user clicks on a task badge they are show snippets of the three most popular prompts for this task from our prompt library.
👍 Prompt pre-filled
If the user opts to use a suggested prompt, it will appear in the text area automatically, where the user can revise or refine it before submitting.
👍 Chat page
The chat area will show the result and ask follow-up questions. If the text is over 150 characters, a banner will prompt users to save it to the company prompt library. Charts can be shared with other users.
👍 Saving prompt modal
This modal allows users to save new prompts. Users can select up to three tasks the prompt supports, choose to make it public or private, and decide whether to display their name or remain anonymous.
👍 Prompt library
Saved prompts appear on the saved prompt page. There, users can view a preview of each prompt with its rating and the creator's title.
Training sessions
As part of the project, I conducted two company-wide training sessions for end users: one for the EMEA team and one for the North American team. The purpose of these sessions was to explain both the capabilities and the limitations of the AI Assistant.
During the demo, I covered the following topics:
Differences between the available LLM modes
Key distinctions between the Waves AI Assistant and ChatGPT (as the most commonly used reference tool within the company)
The probabilistic nature of large language models
Best practices and strategies for effective prompting
Company data protection guidelines
Live demo of the AI Assistant
Example demo of a performance analysis task using the Waves AI Assistant
See the demo presentation here.
Results in detail
✅ No restrictions for file sizes
Users asked for unlimited file upload sizes and no text character limits to support the processing of large briefs and RFP documents.
✅ Dictation
Users would like the ability to think out loud first and then have the AI assistant process and refine their thoughts as needed.
✅ Suggest follow up questions
Users find follow up questions and suggestions functionality in other assistants very useful for adding additional context.
✅ Image & chart generation
Users would like to have image and chart generation, for example, based on provided data to copy into decks.
✅ Indication of sources
Users would like to see the sources of the information so they can cite them in the files and documents they create.
✅ Memory
Users would like the AI Assistant to have good memory and context preservation because they tend to work on materials for the same client.
✅ Prompt library
It was a business request to add a company prompt library to the AI assistant to increase adoption of the product.
✅ Chat grouping
Users have expressed that they value the ability to organize chats into groups, such as those dedicated to a specific brand.
Impact
The Waves AI Assistant was launched in September 2025. Unlike general chat tools, it connects to specialised tools that provide direct access to social media data.
The company successfully achieved its goal of generating 6,000 messages in the first quarter following the launch of the AI assistant.
The AI assistant also represents a major milestone, establishing a foundation for AI adoption and future workflow integration.
The next steps include integrating the Assistant with the Waves database, expanding its toolset to deliver deeper campaign insights, and refining usability based on employee feedback.
