Run a Research agent on accounts or contacts

A Research agent gathers and writes up specific information about an account or a contact - the kind of detail you would otherwise dig out by hand from a website, news, or a profile. Run it when you want richer context on records before you reach out, or to refresh what you already have. For what these agents are and how they differ from Qualification, see Account signals and Contact signals.

Running agents is for editing users. If you are signed in with a read-only session, the agent action buttons do not appear.

Before you start

The list of agents you can run is built from the records you select:

For accounts, the workspace looks at the verticals linked to your selected accounts and offers the Research agents tied to those verticals. If your accounts have no linked verticals, you will see “No linked verticals found for selected accounts” and nothing to run. Qualify the accounts first so they carry a vertical.
For contacts, it looks at the personas linked to your selected contacts and offers the Research agents tied to those personas. If your contacts have no linked personas, you will see “No linked personas found for selected contacts”. Qualify the contacts first so they carry a persona.

If the records do have verticals or personas but no Research agents are linked to them, the workspace tells you so and closes the window. Linking agents to verticals and personas happens in the Agent Training Center.

Steps

You can run a Research agent on a single record from its detail page, or on many records at once from the list.

On one record

Open the account or contact detail page (see Account detail page or Contact detail page).
In the row of agent buttons near the top, select Research agent.
The window opens. Continue from step 3 in the bulk steps below.

On several records at once

Open the Accounts or Contacts list and select the records you want by ticking their checkboxes (see Filter accounts or Filter contacts to narrow the list first).
In the action bar that appears, select Research agent.
The window opens with the title Research agent for accounts or Research agent for contacts. The footer shows how many records are selected, for example 12 accounts selected.
The window lists every Research agent available for the selected records’ verticals (accounts) or personas (contacts). Each agent is a card showing its name, the field type it produces, and which ecosystems or personas it is used in. By default every available agent is selected.
Choose which agents to run:
- Use the search box at the top to filter the list by agent name.
- Toggle an individual agent on or off with its switch.
- Use Select all to turn on every agent currently shown, or Unselect all to clear them.
- You must keep at least one agent selected. Submitting with none selected shows “Select at least one research agent”.
Optionally turn on Boost, the switch in the window’s header, to run this launch at Fast speed - quicker results at double the credits for the qualification and research work in it. The credit estimate re-prices and reads (2x boosted). See Boost a single run.
On more than one record, optionally switch on Stop early at a target - Stop at [number] researched accounts (or researched contacts). A record counts once any one of the agents you selected above has a result for it, or, for an agent that reports whether it found something, once it has found it; so the target follows your selection. You can also count records matching your own conditions instead. See Stop a run early at a target.
Select Start to run, or Cancel to close without running.

Straight from an empty research cell

If you have a research agent’s own column showing on the Accounts or Contacts list, a record with nothing in that cell offers the Research agent right there - and opens the window with that column’s agent already selected, so you are running exactly the one whose gap you spotted rather than the full set.

The button only appears where the agent could genuinely run: on accounts, when the ICP is Yes and the agent is assigned to the account’s vertical; on contacts, when the IBP is Yes and the agent is assigned to the contact’s persona card. An empty cell with no button means one of those is not true yet, which is a more useful answer than a run that would be skipped. See Accounts list: columns and filters and Contacts list: columns and filters.

Straight from an empty Score

An account with no Score yet offers Research in that cell too, on any qualified account. This one opens with only your signal research selected, not the full set - the Score measures signal coverage alone, so running your data-point agents alongside would spend credits without moving the number you were trying to fill. The same button sits on the Account Score column of the Contacts list, where it runs against the person’s company rather than the person. What the Score does and does not mean is covered in Account Research and signals.

On the accounts behind a set of contacts

From the Contacts list you can research the companies rather than the people. Select the contacts, open Launch agent, and switch to the Accounts tab. The workspace resolves your selection to the accounts those people belong to - several contacts at the same company count once - then opens the account version of the window against those accounts. If none of the selected contacts has an account linked, a message says so and nothing runs. While the run is going, the contacts on those accounts show the working indicator in the list, so you can see the research is under way from the screen you started on.

What happens next

When you select Start, the workspace creates a research job for each selected record and confirms with the message “Job created successfully”. You can watch progress under Agent tasks on the Activity tab of the record’s detail page, where each run shows a status, the action, and the source. While a record has work in progress, its Activity tab shows an in-progress indicator.

Each agent works through the records independently, so results arrive as they finish rather than all at once. When a run completes, its output appears on the Research tab of the record’s detail page, in the Research agents section for an account and the Agent research section for a contact. Each result shows:

the agent’s name and its written answer,
a timestamp of when it was produced,
a Source and reasoning chip you can hover to see the sources used and the reasoning behind the answer.

If an agent has been edited since the result was produced, the timestamp turns orange and a note prompts you to re-run it. You can re-run a single agent on that record with the restart button next to its result.

For accounts, positive findings also feed the Score shown next to the account name - a percentage calculated from the Account Research agents that identified positive signals. Running more Research agents, or re-running them, keeps that score current.

If a run errors or never finishes, see Agent job error and Task stuck.

Eva-only: the accounts and contacts Research actions are both labeled “Research agent” in the UI and present the same surface to the customer; accounts run the account research path and contacts run the contact research path. The account action was previously called “Enrich”; the customer-facing name is now “Research” everywhere, so if a customer still says “enrich an account” they mean running the Research agent.