Before you begin: This article explains how to view and change the prompts that AI agents use in TapClicks. You can change the prompt for the Executive Summary (Widget) agent and for Build-Your-Own Agents (BYOA Agents). AI Insights Agents use fixed prompts that you can view but not change. Managing AI agent prompts requires a Super Admin role. For more information on AI in TapClicks, see FAQs: Using AI in TapClicks.
Which Agent Prompts You Can Change
The AI Agents list includes three types of agents. The icon next to each agent shows whether you can change its prompt.
| Agent | Prompt | Icon in the AI Agents List |
|---|---|---|
| Executive Summary (Widget) | Editable. Used when you send widget data to AI. | Edit (pencil) |
| BYOA Agents | Editable. Uses the prompt written when the agent was created. | Edit (pencil) |
| AI Insights Agents | View only. The prompt cannot be changed. | View (eye) |
To create a BYOA Agent, see How to Build-Your-Own AI Insights Agent. For more information on AI Insights Agents, see AI Insights Agents Overview.
Default Prompt for Sending Widget Data to AI
When you send widget data to AI to generate an AI-assisted summary, the Executive Summary (Widget) agent uses this default prompt:
"Generate optimization recommendations in executive summary format for campaigns based on the attached marketing campaign data, written like a campaign manager describing to an advertiser who understands marketing. Be concise, highlight recent campaign performance."
This prompt generates focused, strategic insights in a clear and concise format, so the information is relevant and actionable for decision-makers familiar with marketing trends.
| Note: If you change the Executive Summary (Widget) agent prompt and want to restore the default, copy the prompt above and paste it into the Prompt field by following the steps below. |
How to Change an Agent Prompt
1] Log in to TapClicks, and on the Data menu, click AI Agents.
2] Locate the Executive Summary (Widget) agent or the BYOA Agent whose prompt you want to change by scrolling or using the available filters, and click the Edit (pencil) icon. The agent editor opens.
3] In the Prompt field, edit the prompt as needed.
4] Click Save at the bottom of the page.
Suggestions for Quality AI Prompts
To write effective prompts, be clear, specific, relevant, and well-contextualized. The right details guide the model and lead to more tailored and useful responses.
- Clarity: Make sure the prompt is clear and specific, with a well-defined topic or question. This helps the AI understand the requested information and generate more relevant and useful responses.
- Context: Providing context about the problem or question you are addressing can improve the accuracy and usefulness of the AI's responses. Include background information, relevant data or statistics, or other details that help frame the issue or question.
- Relevance: Ask for analysis the AI can perform with the data you send, such as trends, comparisons, and recommendations based on the metrics in your widget.
- Grammar and Spelling: Using proper grammar and spelling helps the AI correctly understand your prompt, leading to more accurate responses. Avoid slang or informal language that might confuse the model.
Considerations for Using AI Effectively
AI is a powerful tool capable of generating accurate and informative responses across a wide range of topics, but there are situations where it may provide incorrect information. Here are a few examples:
- Misunderstanding the context: The AI may generate inaccurate responses if it does not fully understand the context of the question or prompt. Providing additional information or rephrasing the prompt can help improve response accuracy.
- Limited knowledge or training data: The AI's responses are based on the knowledge and training data it has been exposed to. If the question or topic is outside its training scope, or if the training data is limited or biased, it may generate inaccurate or incomplete responses.
- Errors in the training data: If the data used to train the AI contains errors or biases, this can affect the accuracy of its responses.
- Ambiguity or complexity of language: Natural language can be ambiguous or complex, and AI may struggle to generate accurate responses to questions that are open-ended or require a nuanced understanding of language.
- Programming or technical errors: Like any software, AI can sometimes encounter technical errors or issues that affect its performance.
AI is not infallible. Keep these limitations in mind and use AI as a supplement to human expertise and knowledge, not as a replacement.
Example Prompts for Widget Data
The following example prompts work with the Executive Summary (Widget) agent. You can also use them as a starting point for BYOA Agents that analyze similar data.
- "Summarize campaign performance for the selected date range. Highlight the top three and bottom three campaigns by conversions, and recommend where to shift budget."
- "Compare spend, clicks, and conversions to the previous period. Explain the biggest changes and suggest one action for each."
- "Identify which channels have the lowest cost per acquisition (CPA), and recommend how to reallocate budget to improve return on investment (ROI)."
- "Flag any campaigns where click-through rate (CTR) or conversion rate dropped by more than 20% compared to the previous period, and suggest possible causes to investigate."
- "Summarize website traffic and conversion trends by week, and note any weeks that stand out from the rest."
- "Using the key performance indicators (KPIs) in this data, list which metrics are on track and which need attention, with one recommendation for each."
- "Write a short executive summary for a client who is not a marketing expert. Avoid jargon and explain what each key metric means."
- "Write the summary as three bullet points followed by one recommended next step."