Agent Spark works best when your prompt includes a few key elements. The more context you provide, the more relevant your insights will be.
A strong prompt includes:
Instruction
Topic
Audience
Location (optional - defaults to all markets covered in the data set used unless specified otherwise)
Timeframe (optional - defaults to the last year of available data unless specified otherwise)
Don't worry about the order. The more context you give, the sharper your insights will be.
Example: "I am preparing a social media strategy targeting Gen Z women with an interest in fashion in the UK. Could you identify their key reasons for using social media? Please focus on 2025."
Instruction: Could you identify their key reasons for using social media?
Topic: Reasons for using social media
Audience: Gen Z women with an interest in fashion
Location: UK
Timeframe: 2025
Additional context: I am preparing a social media strategy
Describe at least one audience and topic
Describing an audience and topic helps Agent Spark surface the most relevant insights for your question, the more specific you are, the better.
Example: Tell me about Gen Z NBA fans in Europe.
If you are a user on the Pro plan you can add a pre-built audience directly to your prompt by clicking the plus icon in the left corner of the prompt field, selecting "Add audience" and typing the name of the audience you'd like to add.
Free or Plus users can add pre-built GWI Audiences to their prompt.
Teams users can add pre-built GWI Audiences and audiences shared by someone in their org.
Note: Country local regions can't be used as part of your audience or location.
Define timeframe and location
By default, Agent Spark generates insights based on the most recent available data for all markets covered in the data set used. Mention a location or timeframe in your prompt to narrow this down.
Timeframes are best understood when you specify a year or research wave.
Example: What was the most popular social media platform in Q1 and Q2 2022 in the UK?
Ask follow up prompts
Agent Spark retains the context, audience, location and timeframe from your initial prompt, so you don't need to repeat these in follow-up questions unless you want to change them.
For example, if you asked "What was the most popular social media platform in Q1 and Q2 2022 in the UK?", you could follow up with "And what about France?" to get the same insights for a different location.
Using Agent Spark in the GWI platform vs. via an LLM
Agent Spark surfaces insights through natural language in both environments, but each has its own strengths.
In the GWI platform:
Prompts work best when focused on a single topic at a time.
You can move directly from results into Charts to explore the underlying data.
Audiences can be applied directly within Agent Spark for tailored analysis.
Via an LLM:
The LLM can break down complex prompts automatically, calling Agent Spark multiple times behind the scenes to deliver a comprehensive answer from a single prompt.
You can combine text prompts with images, charts, or documents for richer context.
You gain access to the LLM's asset creation capabilities and can generate visualisations, written artifacts, and connections to third-party tools depending on the access you have.