Amplify Chat Best Practices

Get the most out of Amplify Chat by following these simple best practices.

These tips will help you write better prompts, interpret results effectively, and use Amplify Chat efficiently in your day-to-day workflow.

Writing Effective Prompts

Be specific.

Provide clear context for what you want Amplify to do. For example, “Summarize Acme Staffing’s open jobs for this quarter” is more effective than “Summarize Acme.”
Include time ranges or filters.

Adding details such as “this month,” “last 30 days,” or “active jobs only” helps Amplify deliver more accurate, targeted results.
Use natural language.

You don’t need to know query syntax. Simply phrase your requests as if you were talking to a colleague. For example: “Show me my top-performing candidates from the last quarter.”
Select records instead of typing names.
Type @ to choose a candidate or contact, or # to choose a job. Amplify then knows exactly which record you mean, instead of interpreting a name in your message. This is especially helpful when records have similar names. For example, “Summarize my last three interactions with @[name]” is more reliable than typing the name in full.
Choose a tool when you know which one you need.
Type / and select a tool such as Bullhorn ATS, Bullhorn Analytics, or Amplify Prospect. This tells Amplify which tool to use instead of leaving it to guess. For example, “Find me some fintech companies in London /Prospect.”
Refine instead of restarting.

Use follow-up questions to clarify or adjust results instead of starting a new chat. Amplify keeps the context from previous messages to improve accuracy.

Check the tools.

If something looks off, check which tool Amplify used for your request. If it isn’t the one you wanted, ask again and select the right one. Tools include Bullhorn ATS, Analytics, Job Feed, and Web Search.
Use the summary explanation.

When Amplify outlines how it’s approaching your request, read this carefully — it shows you how it’s interpreting your prompt.
Answer Guided Questions.
When Amplify needs more information, it may show a short form with checkboxes, dropdowns, or text fields instead of asking you to type. Answer each question as precisely as you can, since your selections shape the result. You can also ask Amplify to gather your requirements this way: “Ask me a few questions in a guided form to narrow down the candidates we should shortlist.”
Review previews before you approve.
When Amplify proposes to add or change records, notes, tasks, or tearsheets, it shows a preview first, and nothing is saved until you confirm. Take a moment to check the details, especially for batch actions, before you approve.

Start a new chat for new topics.

Use New Chat when you want to switch focus. For example, when switching from candidate activity to client engagement. This ensures results are clean and not influenced by earlier context.
Use Chat History for reference.

Your recent conversations are saved for 30 days on the History tab. Use the search bar to find a chat or message by keyword, then open it to revisit data and reuse prompts without retyping.
Combine tools for richer insights.
You can select more than one tool in a single request. For example, use Bullhorn ATS data for internal metrics, then add Web Search to see recent news or market activity for the same company.
Work in batches.
If you have a list of records to add or update, ask for them in one request instead of one at a time. Amplify can add up to 25 candidates, contacts, or companies at once, and prepare notes and tasks for several records together. See Taking Action on Records with Amplify Chat.
Try suggested next steps.
After some responses, Amplify suggests what to do next. These can save you from writing a follow-up request yourself.

Your feedback helps refine Amplify’s accuracy and usefulness over time.

Click Thumbs Up () if a response is helpful or accurate.
Click Thumbs Down () if it’s incomplete or incorrect, and add a brief note explaining why.

This information goes directly to the Amplify development team to guide improvements.

If your prompt doesn’t return the expected results:

Try simplifying or rewording your request.
Select the exact records with @ or #, and the tool you want with /.
If results still seem off, test the prompt in another large language model (such as ChatGPT) to see how it interprets the phrasing. If similar models struggle, refine your wording or narrow the scope.
See Using Generative AI to Create Amplify Prompts and Amplify Prompting Best Practices for guidance on writing and testing effective prompts.