Exploring the Upgraded AAVA: Conversational AI for Sports Science Data
- Jo Clubb
- Sep 24
- 4 min read
This post demonstrates the upgraded Action Apps Virtual Assistant (AAVA), an AI-powered chatbot that provides data viz from conversational prompts.
AI tools are developing rapidly, and sports science is no exception. I continue to focus much of my work on AI as I believe those who develop the skills to leverage such technology will set themselves apart.
In a previous post, I introduced AAVA, the AI-powered chatbot developed by Action Apps that can generate charts and insights from your athlete data with a single prompt. Now, AAVA has levelled up. The AAVA Chat update introduces an improved user interface and, crucially, a conversational chat mode. That means AAVA not only responds to individual prompts but also remembers the context of your earlier queries - building a dialogue that makes exploring data even more intuitive.
In this article, I’ll take you through the latest updates and demonstrate how conversational AI can support performance practitioners.
AAVA and Power AMS
Action Apps deliver their athlete management system (Power AMS) as a flexible platform built on Microsoft tools. It combines the structure of a ready-to-go back-end with the ability to customise front-end dashboards in Power BI.
AAVA sits on top of this system, driven by large language models, and allows practitioners to query their data using everyday language. Importantly, AAVA is designed to provide clear, specific answers rather than automated recommendations - a sensible step as AI becomes more integrated into applied sports settings.
From Prompts to Conversations
With AAVA’s new chat mode, practitioners can ask questions of GPS, VALD, or training load data in everyday language. More importantly, AAVA remembers the context. You don’t need to restate whether you’re looking at the last game of the season or excluding substitutes below a certain playing time, it carries that information forward.
In the demonstration below, I start by asking AAVA to display total distance by player for the final game of last season. Within seconds, it produces:
a chart,
an accompanying data table,
the underlying code that generated its response,
and a narrative summary suitable for sharing in reports or emails.
Refining the query was simple. By asking to exclude players with under 80 minutes of match time, the chart updated instantly — with no need to repeat the original context.
The real strength of chat mode comes when layering further insights:
Adding metres per minute as a secondary line chart on top of total distance.
Highlighting games where an athlete played fewer than 90 minutes.
Inserting an average line to benchmark individual performances.
Selecting particular players or positions using the legend.
Each time, AAVA builds on the existing chart without losing the context, which as practitioners know, is critical when analysing sports science data.
Going Deeper with Other Data Streams
The demonstration also explored VALD jump data, showing how AAVA can:
Average jump height by player,
Box plots to show distribution,
Removal of erroneous data from a single athlete,
Scatterplots linking jump height with relative peak power.
Everyday language was enough for AAVA to interpret “relative peak power” and map it correctly to the underlying metric definition.
Assessing a single data stream in isolation has limited insight for sports scientists. While so much emphasis is placed on GPS, the external training load alone doesn't tell us how the athlete is responding; it doesn't solve the Training Adaptability Prediction Problem.
We need a holistic approach to load monitoring. By integrating multiple data streams, such as GPS, heart rate, subjective monitoring, and force plate data, we gain a more complete picture of training load and adaptation. Power AMS provides the infrastructure to bring these datasets together, while AAVA Chat offers a way to interrogate them quickly and conversationally.
Why This Matters for Practitioners
The upgrade moves AAVA beyond being a one-prompt tool into a dynamic conversation partner. For practitioners, this offers:
Efficiency: quicker exploration of datasets without rebuilding reports.
Accessibility: no coding or expert data science background required.
Flexibility: the ability to ask ad hoc questions as they arise.
Communication: ready-made narrative summaries to share with coaches or athletes.
Final Thoughts
The upgraded AAVA Chat offers a smoother interface, conversational memory, and a more intuitive workflow for sports scientists and coaches. It shows how AI can support daily practice - not by replacing expertise, but by making it quicker and easier to access meaningful insights. I would like to extend my gratitude to Action Apps for sponsoring this exploration into AAVA and for permitting us access to the demo.
FAQs
What is AAVA?
AAVA (Action Apps Virtual Assistant) is an AI-powered chatbot that generates charts, tables, and narrative insights from sports performance data via written prompts.
How is the new version different?
The latest update introduces a conversational chat mode and an improved interface. AAVA Chat now remembers context, so you can layer queries without repeating information.
What data sources can AAVA use?
In this demo, we used fabricated GPS and VALD datasets. However, AAVA can be linked to your own performance data streams via their athlete management system, Power AMS, to generate insights.
Do I need coding skills to use AAVA?
No. You can interact with AAVA using natural language prompts. For those learning, AAVA also shows the code behind each query.
Is AAVA secure and compliant?
Yes, AAVA is backed by Microsoft security, ensuring GDPR and HIPAA compliance, making it secure for managing sensitive athlete data.
This article is support by Action Apps. For more information about their services, visit their website.
Subscribe to our blog to stay updated!
Comments