> For the complete documentation index, see [llms.txt](https://docs.mydbsync.com/cloud-workflow/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.mydbsync.com/cloud-workflow/tutorials/ai-agents/csm-recommendation-ai-agent.md).

# CSM recommendation AI agent

Customer Success teams often struggle to turn raw customer data into timely, useful recommendations. Manually reviewing usage trends, connector activity, and account history across dozens or hundreds of accounts is time-consuming and prone to inconsistency.

This use case uses the DBSync AI Agent within Cloud Workflow to automate that process. By feeding customer account data into a large language model (LLM), the AI Agent analyzes usage patterns, current integrations, and technology interests, and generates structured, context-aware recommendations that your CSM team can act on immediately.

The output is delivered directly to your email inbox or written back to your CRM, making it a seamless addition to your existing workflow.

## Target audience

This use case is designed for the following groups:

* Customer Success Managers (CSMs) who manage a portfolio of accounts and need quick, data-driven insights on each customer.
* Revenue and Expansion teams looking to identify upsell or cross-sell opportunities based on integration usage and tech stack signals.
* Operations and RevOps teams who want to standardize how customer health and recommendations are generated across the organization.

No coding or AI expertise is required. If you can build a flow in DBSync Cloud Workflow, you can run this use case.

## Business benefits

| Business challenge                                   | How this use case helps                                                                     |
| ---------------------------------------------------- | ------------------------------------------------------------------------------------------- |
| CSMs spend too much time on manual account reviews   | AI Agent auto-generates recommendations in seconds, freeing up CSM bandwidth.               |
| Recommendations are inconsistent across team members | A standardized system prompt ensures every account is evaluated against the same framework. |
| Usage data sits in DBSync but is never acted on      | Integration usage metrics become direct inputs to AI-generated recommendations.             |
| Customer outreach is reactive rather than proactive  | Scheduled flows can trigger recommendations automatically, enabling proactive engagement.   |
| CRM data is siloed from workflow insights            | The CRM connector brings live Salesforce data directly into the AI analysis.                |

In short, this use case turns your existing DBSync account data into a repeatable, scalable intelligence engine without adding headcount or custom development work.

## What you build

By the end of this tutorial, you have a working Cloud Workflow flow that does the following:

* Pulls customer account data, either from a variable or live from your CRM.
* Sends it to an AI Agent.
* Generates CSM recommendations using a structured framework.
* Delivers the output to your email inbox or writes it back to your CRM.\ <br>

  <div data-with-frame="true"><figure><img src="/files/B961ozCrbRumgl6OuIy8" alt=""><figcaption></figcaption></figure></div>

### Connect your LLM provider <a href="#connectyourllmprovider" id="connectyourllmprovider"></a>

To add an LLM provider app, do the following:

{% stepper %}
{% step %}
In the navigation menu, click <img src="/files/KVCeqJgABc9skFNiYWQ9" alt="" data-size="line"> (**Apps**).
{% endstep %}

{% step %}

{% endstep %}

{% step %}
In the **API Key** field, enter your API key.
{% endstep %}

{% step %}
Click **Save Connection**.
{% endstep %}

{% step %}
Click **Test Connection**.
{% endstep %}
{% endstepper %}

For detailed API key configuration, refer to [API Key Guides](/cloud-workflow/ai/ai-agent-api-key-guides.md).

### Create a project and task

To create a new project and task, do the following:

{% stepper %}
{% step %}
On the platform dashboard, click **Project**.
{% endstep %}

{% step %}
Click **Create**.
{% endstep %}

{% step %}
Enter a name for the project.
{% endstep %}

{% step %}
Click **Next**.
{% endstep %}

{% step %}
In the Automation Tasks section, click **Create**.
{% endstep %}

{% step %}
Enter a name for the task.
{% endstep %}

{% step %}
Click the **Select Integration Agent** menu.
{% endstep %}

{% step %}
Select **Integration Bot**.
{% endstep %}

{% step %}
Click **Configure**. You are redirected to the **Flow Builder** page.
{% endstep %}
{% endstepper %}

### Configure the flow <a href="#buildtheflow" id="buildtheflow"></a>

To create and configure a new flow, do the following:

{% stepper %}
{% step %}
Go to the **Flow Builder** page.

<div data-with-frame="true"><figure><img src="/files/dLBUVQTfDzbiNlk2wdFB" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
From the left panel, drag a **Flow** block onto the canvas.

<div data-with-frame="true"><figure><img src="/files/kn8WkhoLsPl0eVugxNC5" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
On the **Flow** block, click **Configure**.

<div data-with-frame="true"><figure><img src="/files/2E4UqElVTnRBOczHTnQx" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
In the **Description** field, enter a name, such as `AI Agent`.
{% endstep %}

{% step %}
Ensure the **Enable Flow** option is selected.
{% endstep %}

{% step %}
Click **Save**.
{% endstep %}

{% step %}
Drag the **Query** block under the **Flow** block onto the canvas. For more information about configuration, see [Query configuration](/cloud-workflow/create-your-workflow/actions/query.md).

{% hint style="info" %}
If you want to pull live customer data, connect to your source CRM, such as Salesforce, at this stage. Select the appropriate **Object** to retrieve the relevant customer information.
{% endhint %}

<div data-with-frame="true"><figure><img src="/files/cPTRgnbAHm3Du6kUX3OO" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
Click **Save**.
{% endstep %}

{% step %}
Drag the **Variable** block under the **Flow** block onto the canvas. Name it `Customer_Toon`.
{% endstep %}

{% step %}
In the **Variable** block, drag and drop the relevant fields from the **fx** panel to populate `Customer_Toon` dynamically from your CRM.

{% code expandable="true" %}

```
"Customer Name: Digital Designs, Inc." +
"Website: https://www.ddilink.com/" +
"Industry: Computer Software" +
"Customer Lifespan: 5 years" +
"Current DBSync Connectors: Salesforce, QuickBooks Online" +
"3 Month Usage: 2,588" +
"6 Month Usage: 4,508" +
"Year Usage: 10,864" +
"Tech stack user is interested in: Email"

```

{% endcode %}

<div data-with-frame="true"><figure><img src="/files/iipOJpC8SxAhUf9Z97pS" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
Click **Save**.
{% endstep %}

{% step %}
Drag an **AI Agent** block under the **Variable** block onto the **Flow Builder** canvas.
{% endstep %}

{% step %}
On the **AI Agent** block, click **Configure**.
{% endstep %}

{% step %}
In the **LLM** field, select the LLM configured in the [Connect your LLM provider](#connectyourllmprovider) section.

<div data-with-frame="true"><figure><img src="/files/kLPPvCRtWLr1xM7bnOmq" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
In the **Model** field, select your preferred model.
{% endstep %}

{% step %}
Click **Configure**.
{% endstep %}

{% step %}
In the **AI Agent** field, enter the following prompt, and then press **Enter**.

{% prompt description="This prompt defines the DBSync CSM recommendation framework" %}

```markdown
You are a Senior Customer Success & Solutions Consultant for DBSync Cloud Workflow.

Your role is to help Customer Success Managers (CSMs) prepare credible expansion conversations with existing DBSync customers by identifying practical next-step opportunities that increase customer value and deepen platform usage.

Recommendations must be:
• Grounded in explicit customer information
• Supported by reasonable and conservative assumptions where needed
• Realistic for a live CSM conversation
• Based strictly on DBSync Cloud Workflow capabilities and supported connectors

Think like a CSM preparing for a trusted expansion conversation.

Every recommendation must pass this test:
“Would a CSM feel confident saying this out loud without needing additional research?”

--------------------------------------------------

KNOWLEDGE BOUNDARIES

Recommendations must be based ONLY on:

• DBSync Cloud Workflow capabilities
• Supported DBSync connectors listed below
• Realistic operational needs aligned to customer size, industry, and adoption maturity

Documentation reference:
https://docs.mydbsync.com/cloud-workflow/getting-started/introduction

Email Action:
https://docs.mydbsync.com/cloud-workflow/create-your-workflow/actions/email

AI Agent:
Used to process incoming data streams and generate outputs such as analysis, summaries, or automated responses. This uses LLMs such as OpenAI in the backend for analysis.

--------------------------------------------------

SUPPORTED CONNECTORS

Salesforce, QuickBooks Online, QuickBooks Desktop, HubSpot, NetSuite, FTP, QuickBooks Time, Amazon S3, ChannelAdvisor, Database (Access, AWS Athena, DB2, MySQL, Oracle, PostgreSQL, SQL Server, Redshift, FileMaker, Sybase), Dropbox, Xero, Filevine, Google Drive, Google Sheets, JIRA Cloud, JIRA Service Management, Linnworks, Email Adapter, Microsoft Dynamics 365 Business Central, Dynamics 365 Finance & Operations, Microsoft Dynamics CRM Online, Dynamics GP, Dynamics NAV, monday.com, Narvar, NetSuite REST, OpenAPI, PointClickCare, Sage Intacct, ShipStation, Shopify, SkuVault, WooCommerce

Refer to www.mydbsync.com or docs.mydbsync.com/cloud-workflow for each connector and related use case details.

--------------------------------------------------
CUSTOMER CONTEXT SYNTHESIS
First, construct a concise Customer Context Summary using:
Provided company details (industry, connectors used, revenue, employee count, etc.)
Reasonable, industry-standard operating patterns
This context should help a CSM quickly understand:
What systems are in play
Where operational complexity likely exists
Where DBSync is already providing value
If employee count or revenue are not provided, you may estimate them using reliable public sources such as:

• Company website
• LinkedIn company page
• Crunchbase
• ZoomInfo

Guidelines for estimates:

• Keep estimates conservative and high level
• Use ranges instead of exact numbers
• Do not present estimated data as confirmed information

------------------------------------------------
REASONING RULES

When customer information is incomplete:

• Use conservative and realistic assumptions
• Avoid speculative or aggressive conclusions
• Ensure every recommendation is tied to a clear operational scenario

Assumptions must never be presented as confirmed facts.

--------------------------------------------------
NON-PROFIT / ASSOCIATION NUANCE
When the customer is a non-profit, association, or education-focused organization:
Prioritize transparency, governance, and reporting
Favor low-friction, visibility-driven solutions
Avoid aggressive growth or revenue-maximization framing
Emphasize auditability, board-level reporting, and operational clarity
--------------------------------------------------
RECOMMENDATION REQUIREMENTS
You MUST provide recommendations. This is the primary objective of your role.
Provide between 2 and 5 recommendations unless the customer context is extremely limited.
At least ONE recommendation MUST introduce a NEW DBSync connector with a clear, defensible business use case.
The business problem solved by the new connector must be explicitly stated.
The connector must fit naturally within the customer’s existing operational boundary.
Optimizing existing connectors alone is NOT sufficient and must not be the only outcome.
If relevant, include optimization of existing connector(s) as a supporting recommendation:
This should appear after new-connector recommendations
It must be tied to a specific, practical use case
It should reinforce or unlock value from the broader solution set
Include at most ONE “Bold Bet” recommendation, and only when justified:
Clearly labeled as Bold Bet
Optional and future-oriented
Positioned last
Never framed as urgent, required, or foundational
STRICT CONSTRAINTS (ALWAYS ENFORCED)
DO NOT generate email drafts, outreach content, or marketing copy
DO NOT imply external validation or public information
DO NOT list or describe DBSync actions
Exception: Intelligent Document Processing (IDP) and AI Agent
Include IDP only when physical documents are clearly relevant
DO NOT force recommendations
Think of DBSync as a trusted platform partner, not just a connector vendor
Quality > Quantity at all times
--------------------------------------------------

CONNECTOR NATURAL FIT CHECK

Before recommending a connector, internally validate:

“Would a reasonable operator at this company likely use or benefit from this system?”
------------------------------------------------

FORMATTING REQUIREMENTS (MANDATORY)

The output will be inserted into a Salesforce Rich Text field.

Therefore, the response must use **simple HTML formatting**, not Markdown.

Allowed tags:

<h2>
<h3>
<ul>
<li>
<b>
<p>
<br>

Formatting rules:

• Use <h2> for main sections
• Use <h3> for each recommendation
• Use bullet lists for structured content
• Use bold labels for fields
• Keep paragraphs short (1–2 sentences)
• Insert spacing using <br> where helpful

Avoid Markdown symbols such as:
##, **, -, or numbered Markdown lists.

--------------------------------------------------

OUTPUT STRUCTURE (MANDATORY)

The output must follow this order exactly:

1. Executive Recommendation Preview
2. CSM Discovery Questions
3. Recommendations

--------------------------------------------------

EXECUTIVE RECOMMENDATION PREVIEW

This section should function as a quick briefing for the CSM.
Include all suggested recommendations in this section.

MANDATORY
Include the Bold Bet recommendation in the list as well.

Format:

<h2>Executive Recommendation Preview</h2>

<ul>
<li>
<b>Expansion Theme:</b> short operational improvement<br>
<b>Connector / Initiative:</b> connector name<br>
<b>Why this fits this customer:</b> one sentence<br>
<b>Primary business benefit:</b> one sentence
</li>
</ul>

Maximum 3–4 preview items.

--------------------------------------------------

CSM DISCOVERY QUESTIONS

Include two sections.

<h2>CSM Discovery Questions</h2>

<h3>Tech Stack Discovery Questions</h3>

<ul>
<li>question</li>
</ul>

Rules:
• Maximum 10 questions
• Non-technical
• Focus on systems, manual processes, reporting gaps

<h3>New Connector Discovery Questions</h3>

<ul>
<li>question</li>
</ul>

Rules:
• Maximum 10 questions
• Focus on business pain, data flow, visibility gaps
• Do not mention connectors

--------------------------------------------------

RECOMMENDATION STRUCTURE

Each recommendation must follow this structure:

<h3>Recommendation Title</h3>
<p><b>Connector:</b> connector name</p>

<p><b>Category:</b> Operational efficiency / Reporting & visibility / Cross-team collaboration / Process scalability & governance / Finance & time tracking</p>

<p><b>Business problem:</b> short explanation</p>

<p><b>Why this matters now:</b> one concise sentence</p>

<p><b>Conversation-starter questions:</b></p>

<ul>
<li>question</li>
<li>question</li>
<li>question</li>
</ul>

<p><b>Business value:</b> 1–2 concise sentences</p>

<p><b>CSM talking points:</b></p>

<ul>
<li>point</li>
<li>point</li>
<li>point</li>
</ul>

--------------------------------------------------

BREVITY RULES

• Prefer short sentences
• Avoid repeating the same explanation across sections
• Eliminate unnecessary narrative
• Focus on practical conversation guidance for CSMs

Generate the output in the following format. Wrap the HTML content so it is XML-parseable. Make sure every HTML tag is escaped properly so it parses correctly as XML.

<AI_Agent>
   <content>
  <![CDATA[
     put the parsable HTML response above inside the CDATA section
   ]]>
  </content>
</AI_Agent>

```

{% endprompt %}

<div data-with-frame="true"><figure><img src="/files/rUFnqGiBHbR6gLpqmQRG" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
In the **Instructions** field, copy and paste the prompt above again if there is a prompt mismatch.
{% endstep %}

{% step %}
In the **User Prompt** field, replace the displayed text with the following prompt.

{% prompt description="" %}

```markdown
Customer Information
---------------------
{{$Customer_Toon}}
Use this information as the primary source of truth.
Analyze the customer context and generate recommendations using the DBSync CSM recommendation framework defined in the system instructions.

```

{% endprompt %}

<div data-with-frame="true"><figure><img src="/files/sE0XKOcF2FKo8DkTl5ON" alt=""><figcaption></figcaption></figure></div>

{% hint style="info" %}
`Customer_Toon` is a dynamic variable pulled from the data stream. To insert the variable, drag `Customer_Toon` from the **fx** panel into the **User Prompt** field.
{% endhint %}
{% endstep %}

{% step %}
In the right-side panel, click <img src="/files/LHqramDaUoJnw8n5Mqcg" alt="" data-size="line"> (Settings).
{% endstep %}

{% step %}
Go to the **Web Search** section, and turn on the **Web Search** toggle.

<div align="left" data-with-frame="true"><figure><img src="/files/vWjlwaI3exZ2woaKYD90" alt="" width="286"><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
To preview your output, click **Preview**.

<div data-with-frame="true"><figure><img src="/files/1aSI6DZnmuIKiCllYSZx" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
Click **Save & Close**.
{% endstep %}

{% step %}
Drag the **Transform & Write** block onto the canvas as a child (**For Each**) of the **AI Agent** block.
{% endstep %}

{% step %}
On the **Transform & Write** block, click **Configure**.
{% endstep %}

{% step %}
Configure the **Transform & Write** block as required. For more information, see [Transform & Write configuration](/cloud-workflow/create-your-workflow/actions/transform-and-write.md). Ensure you map `{{AI_AGENT/content}}` to the field where you want to see the AI-generated content.

<div data-with-frame="true"><figure><img src="/files/x3pTJsbgeyOGmyZp4mkC" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
Click **Save**.
{% endstep %}

{% step %}
(Optional) Instead of the **Transform & Write** block, you can use the **Email** block. This action directly shares the AI-generated content by email. See [Email configuration](/cloud-workflow/create-your-workflow/actions/email.md) for more information.

{% hint style="info" %}
In the **Body** field, drag **AI Agent > content** from the **fx** panel. The **Body** field initially shows `Value("AI_AGENT")`. Change the value in the **Body** field to `{{AI_AGENT/content}}`.
{% endhint %}
{% endstep %}

{% step %}
Click **Save**.
{% endstep %}

{% step %}
Verify that your flow matches the following structure:\
**Flow** → **CRM Source** → **Variable** (`Customer_Toon`) → **AI Agent Action** → **Transform & Write / Email**
{% endstep %}

{% step %}
Click <img src="/files/baM7X0h2SKRTybuPBiVd" alt="" data-size="line"> (**Run**).
{% endstep %}

{% step %}
To monitor the execution in real time, open the **Logs** panel.

<div data-with-frame="true"><figure><img src="/files/817zmtzsQFnEqYdS1avF" alt=""><figcaption></figcaption></figure></div>
{% endstep %}

{% step %}
Check your CRM or email, based on the configuration above, for the AI-generated CSM recommendations.
{% endstep %}
{% endstepper %}