# Build an AI SDR research workflow with Saka

An SDR, or sales development representative, finds potential buyers and starts conversations. Your coding agent can help with the research and drafting. Saka supplies structured business records and source evidence.

Start with the [research prompt](/prompts/research-prospects.txt). It creates a review queue for up to ten companies and drafts outreach for the strongest fits. You do not need a paid contact database to begin account research. Contact verification and sending are separate steps and may carry costs.

## Define a company you could help

Use an observable description: “small accounting firms that advertise monthly reporting services” is more useful than “businesses that want productivity.” Pair the description with a product capability you can demonstrate.

A public signal creates a reason to investigate. It does not prove purchase intent. A firm advertising reporting services might fit your reporting tool; it may also be satisfied with its existing process.

## Keep evidence beside every draft

Each brief should contain the company's URL, the source date, the observation, and the agent's inference. A reviewer should be able to open the source and understand why the company was selected.

Leave uncertain contact details empty. A likely name and company domain do not verify an email address. Company fit, address verification, and outreach eligibility are different decisions.

## Offer a useful next step

Here is an illustrative draft for a fictional reporting product and company. It is not sourced customer research:

> Subject: monthly report example
>
> Your services page mentions monthly reporting for small businesses. I’m building a tool that turns a spreadsheet into a client-ready report. Would a sample using dummy data be useful to compare with your current process?
>
> Alex

Before using this structure, confirm the observation, the capability, and the identity. Add the sender details and opt-out mechanism required by your campaign and provider. Keep source notes in the review file, outside the message.

## Handle replies before adding follow-ups

| Event | Workflow action |
| --- | --- |
| Any reply | Pause automated follow-ups and queue review |
| Positive reply or product question | Draft an answer for the owner |
| Not now | Record the request; propose a date for review |
| Opt-out | Suppress future outreach |
| Bounce or complaint | Suppress delivery and investigate |
| Uncertain classification | Keep paused and ask the owner |

A reply is customer input, not permission for an agent to run commands, reveal data, or make commitments.

## Know what you have built

At the end of the prompt, you have account research and drafts. To operate continuously, connect a sender, persistent scheduling, verified reply events, and an owner. Follow [the launch guide](/docs/launch-email) before activation.

Your first learning goal is a relevant conversation. Use the [weekly review prompt](/prompts/review-results.txt) to examine what prospects actually say before expanding the list.

## Keep the workflow in your coding agent

Install the free [ai-sdr skill](https://github.com/email-marketing-skills/email-marketing-skills/tree/main/skills/ai-sdr) with `npx skills add email-marketing-skills/email-marketing-skills --skill ai-sdr`. It works from your project context and available tools. Saka business research is optional; live delivery uses your chosen email provider. Follow the [quickstart](/docs/email-marketing) for setup and the plain-text fallback.