Understanding AI Suggested Action Plans
We can now turn your engagement survey or performance review results into ready-to-use action plans. If your team doesn't have an in-house HR expert to interpret survey data and decide what to do next, or if your managers need a little nudge, AI Suggestions for Action Plans gives you a starting point. Action items are built on proven people-management frameworks and grounded in your organization's own context.
This article covers how the feature is set up, the frameworks it can draw on to build suggestions, and the context you can add to make those suggestions more relevant to your organization.
Turning on AI Suggested Action Plans
AI Suggested Action Plans is a premium feature. An admin needs to turn it on before anyone in your organization can use it.
To enable it:
- Go to Admin View → Action Plans → AI-Suggested Action Plans.
- Turn the feature on.
Once enabled, every user with permission to create action plans can generate AI-suggested action plans from the Engage and Performance modules.
To control who can use it, go to Roles and Permissions and assign the Manage Action Plans permission to the relevant roles.
Choosing a framework
A framework is the methodology the AI uses to structure its suggestions. For example, it may group action items into "Start, stop, continue" or set SMART action items with deadlines. ThriveSparrow comes with six built-in frameworks, each suited to a different kind of situation.
Framework | What it does | Best used for |
|---|---|---|
SMART | Sets goals that are specific, measurable, achievable, relevant, and time-bound. | The default framework — works for any kind of action plan. |
Effort x Impact | Prioritizes actions by comparing the effort they take against the impact they'll have. | Recommended for any kind of Engage surveys. |
Start-Stop-Continue | Sorts actions into what to start doing, stop doing, and keep doing. | Specially useful for Performance reviews. |
RCA (Root Cause Analysis) | Digs into the root cause behind an issue instead of just addressing symptoms. | Any kind of action plan. |
EX Lifecycle | Organizes actions around stages of the employee journey — onboarding, development, and exit. | Engagement, onboarding, and exit surveys. Not recommended for pulse surveys, since those focus on day-to-day activities rather than overall satisfaction. |
GROW | Structures actions around Goal, Reality, Options, and Way forward. | Recommended for Performance reviews. |
Your admin chooses which of these frameworks are available when your team generates suggestions.
Creating a custom framework
If none of the built-in frameworks fit how your organization likes to plan, an admin can create a custom one.
To create and test a custom framework:
- From Admin View → Action Plans → AI-Suggested Action Plans, choose to create a new framework.
- Give it a name and description.
- Write the prompt that tells the AI how to structure its response — including instructions, guidelines, and any resources you want it to reference.
- Test the framework with a sample scenario, such as "If an employee scores below 30%."
- Review the AI's response and let it know whether you're satisfied. If not, add comments explaining what you'd change — the AI uses this feedback to refine future suggestions.
You can come back and edit a custom framework at any time. Admins can also choose to let other action plan creators adjust the framework and context while they're generating a plan.
Adding context to improve suggestions
Context is extra information the AI uses alongside a framework to make its suggestions specific to your organization, rather than generic advice. By default, some context is automatically added to every prompt based on your account's usage. You can also upload your own files as additional context, along with a short description of how the AI should use them.
There are four types of built-in context:
- Survey Trends: Compares the current survey's scores against your last few surveys (you can choose to include 2, 3, or 4 past surveys), so suggestions account for whether things are improving or declining.
- Recognition History: Factors in how much recognition (kudos and awards) employees have received over a selected time period to see whether low recognition may be contributing to an issue.
- Analyse Performance: Brings in recent performance reports (anonymized) at the org, department, or manager level, depending on which focus area the suggestion is for, to see if high- or low-performing teams have different issues.
The more relevant context you add, the more grounded and specific the AI's suggestions will be.
Helping the AI learn from feedback
AI Suggested Action Plans improve over time based on feedback from the people using it.
When someone generates a suggestion, they can mark whether they're not satisfied with it. That feedback, along with the reason for any dissatisfaction and the focus area and prompts involved, is stored for our internal purposes of refining the suggestions.
Need help setting up AI-suggested Action Plans? Contact our support team to get started.