Skip to content
Evergrowth Documentation

Create qualification and research criteria

A criterion is one question your agents answer about an account. Qualification Agents check the characteristics that define a good-fit account; Research Agents collect extra detail about accounts that pass qualification. This page walks through adding a criterion of either kind inside the Agent Training Center: naming it, writing its instructions, choosing how the answer is shaped, setting effort and speed, and giving each vertical its own expected answer. You would do this when you want your agents to score or research something they don’t yet capture.

Adding, configuring, and duplicating criteria is available to full-access members. If you are signed in with a read-only session, the Add and edit controls described here are hidden and you can only view existing criteria.

Before you start

Section titled “Before you start”
  • You need an ecosystem with at least one vertical defined, because each criterion is answered per vertical. See Set up ecosystems and verticals.
  • Decide whether the new criterion is a qualification check or a research field. They live in separate groups and behave differently downstream.

Steps

Section titled “Steps”

  1. Open Agent Training Center from the left navigation and select the ecosystem you want to work in. The ecosystem opens as a table: verticals run across the top, and the rows are grouped into Qualification Agents and Research Agents, each with a count.

  2. In the group you want, use the Add action. The tooltip reads Add Qualification Agent or Add Research Agent depending on the group. This opens the Agents templates picker.

  3. In Agents templates, you can pick an existing template to start from, or choose Add new to build a criterion from scratch. Either way the criterion editor opens, titled Add Qualification Agent or Add Research Agent depending on the group.

  4. Fill in the Agent Details section:

    • Name - what the criterion is called. This is the question shown in the ecosystem table.
    • Generate with AI - an expandable helper. Describe where and how you find this information today, then choose Generate to have Evergrowth draft the instructions for you. A name is required before you can generate; for Research Agents you also need at least one vertical answer filled in first.
    • Instructions for Research Agent - the instructions the agent follows when it works on this criterion. You can write these yourself or edit what the generator produced. (This field carries the same label in both the Qualification Agent and Research Agent editors.)
    • Will be used in - choose where this criterion’s answer feeds into. The options are Business model description, Account plans, Signals, and Competition landscape. You can select more than one.

  5. In the Effort & Speed section, set how hard and how fast the agent works on this criterion:

    • Effort - Low (0.5 credit) follows your steps exactly, Medium (1 credit) follows your steps and adds reasoning where needed, High (1.5 credits) handles complex, ambiguous tasks.
    • Speed - Standard (1x) or Fast (2x).
    • A Cost per run line shows the resulting credit cost (effort multiplied by speed). See Effort and speed for the full picture.

  6. In the Output Format section, pick the Response type that shapes the answer:

    • Text - free text.
    • Number - a numeric value.
    • Yes/No - a simple yes or no answer.
    • Select - one choice from a list you define.
    • Multi-Select - one or more choices from a list you define.
    • Range - a low-to-high range, entered with a minimum and a maximum.
    • Regions - a geographic selection.

  7. Depending on the response type, extra controls appear:

    • For Multi-Select, an AND / OR choice sets how the selected answers combine: AND means the account must match every selected value, OR means matching any one is enough.
    • For Number or Range, an optional Unit (Optional) group lets you attach a unit: %, $, €, or Other (with a free-text field for your own unit).
    • For Select, Multi-Select, and Range, a Possible answers list appears. Use Add to add each option and the trash icon to remove one. Options that are already in use by existing answers can’t be removed.

  8. In the Select Verticals section, set the expected answer for each vertical. This section appears once a response type is chosen. Each vertical gets its own cell so the same criterion can mean different things across verticals. Use Enable all at the top to switch every vertical on at once, and the Copy value to all action on a cell to push that cell’s value into every enabled vertical.

  9. Choose Add at the bottom of the editor to save. Use Close to discard.

Duplicate an existing criterion

Section titled “Duplicate an existing criterion”

To reuse a criterion as the basis for a new one, open the duplicate action from the criterion. The editor opens titled Duplicate Qualification Agent or Duplicate Research Agent, with the name pre-filled as the original plus (Copy) and all settings carried over. Adjust what you need and save with the Duplicate button.

Edit answers per vertical later

Section titled “Edit answers per vertical later”

You don’t have to set every vertical answer at creation. From the ecosystem table you can:

  • Configure all Qualification Agents or Configure all Research Agents for a group, which opens a side panel listing each criterion as a card. Expand a card to set or change every vertical’s answer, use Enable all and Copy value to all as above, then choose Save.
  • Configure agent on a single criterion to reopen its full editor (titled Configure Qualification Agent or Configure Research Agent), where Update saves and keeps it open and Update & Close saves and exits.

What happens next

Section titled “What happens next”

When you save, the criterion is created, attached to the ecosystem, and its per-vertical answers are stored. A confirmation appears (Agent successfully added), the editor closes, and the new criterion shows up as a row in its group in the ecosystem table with its answer filled in for each vertical you set. From there it is part of the agent’s work: a Qualification Agent run will score accounts against it, and a Research Agent run will collect it. See Run a Qualification agent and Run a Research agent to put it to work.

Eva-only: Where a criterion’s data comes from is fixed context on the template, not a setting the user picks in this editor. Don’t describe choosing a data origin as one of the steps here.

Section titled “Related”