> ## Documentation Index
> Fetch the complete documentation index at: https://deepline.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Research accounts with source evidence

> Build one evidence-linked account brief with explicit unknowns and a proposed next question.

[Human guide: work with your agent →](https://deepline.com/docs/human/guides/account-research-automation.md)

Build a small account brief that separates sourced facts from hypotheses. Keep enough evidence to review each proposed outreach angle.

## Before you start

Use an authorized workspace and a small file with stable source IDs.
Have source access and your team's acceptance policy available.
This guide describes customer-authored logic, not an automatically enabled product policy.

## 1. Understand the decision

An observation needs its source and scope. An inference must remain distinct from an observed fact.

Test this fictional sample before calling providers:

| Case | Expected decision under this example policy |
| - | - |
| Dated source reports an office opening | Record the observation |
| Hiring page lists open roles | Do not infer an approved budget |
| Source retrieval fails | Report unavailable evidence |

## 2. Inspect supported operations

```bash theme={"languages":{"custom":["/languages/deepline-prompt.json"]}}
deepline auth status --json
deepline search "company website account news research" --type tools,prebuilts --json
```

Inspect a returned ID with `deepline describe <returned-id> --json`.
Compare required identifiers, outputs, access mode, and price.
A similar title does not establish that a contract fits.

## 3. Implement the bounded workflow

Ask your agent to author `account-research-automation.play.ts` using [Write Plays](https://deepline.com/docs/sdk-v2/plays.md).
Have it declare the CSV input and preserve source IDs.
Reuse existing contracts when they fit the transformation.

Implement these rules explicitly:

1. Define the account identity and three bounded research questions.
2. Retain source URLs, retrieval time, and available observation dates.
3. Connect findings to evidence. Mark unsupported conclusions as hypotheses.
4. Return missing evidence and the next question without creating outreach or CRM changes.

Keep external changes disabled for the first pilot.
That control belongs in the authored logic, not a universal Play dry-run flag.

## 4. Check and pilot

Prepare `pilot.csv` using [CSV-safe selection](https://deepline.com/docs/sdk-v2/batch-csv.md).
Then check the source:

```bash theme={"languages":{"custom":["/languages/deepline-prompt.json"]}}
deepline plays check account-research-automation.play.ts
```

Resolve authoring errors first. After reviewing source and price, run only the pilot:

```bash theme={"languages":{"custom":["/languages/deepline-prompt.json"]}}
deepline plays run --file account-research-automation.play.ts --csv pilot.csv --watch
deepline runs get <run-id> --full --json
deepline runs export <run-id> --out account-research-automation-review.csv
deepline billing usage --run-id <run-id> --json
```

Use the ID returned by the pilot. Select a returned dataset path when several
exist. Provider-backed execution can incur usage.

## 5. Verify the outcome

Open the evidence behind each conclusion. Confirm that unavailable access did not become a claim that no activity exists.

> [!TIP]
>
> Every source ID has a result or explicit exception. The fictional cases follow
> the stated policy, and actual results retain their evidence.

Ask your agent why a record was accepted, held, or rejected.
Correct the rule before expanding input or enabling external writes.

## Continue

[Research workflow](https://deepline.com/docs/recipes/research-workflow.md) covers sources. [Provenance](https://deepline.com/docs/guides/data-provenance-in-crm.md) explains evidence history.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.