How to Build an AI Agent for Salesforce, HubSpot, and Slack

The workflow problem

Revenue operations teams struggle with a fundamental data fragmentation problem. Marketing runs campaigns in HubSpot, generating leads and tracking engagement. Sales works opportunities in Salesforce, managing pipeline and closing deals. And the real-time coordination between these two functions happens in Slack, where context evaporates within hours.

Consider the lifecycle of a single lead. Marketing captures a form submission in HubSpot, scores the lead based on website behavior and email engagement, and eventually marks it as an MQL. At that point, someone — usually a marketing ops person — has to manually create or update a lead in Salesforce and notify the right SDR in Slack. This handoff alone can take hours or days, during which the prospect’s interest cools.

The problem deepens after the handoff. When a sales rep updates a deal stage in Salesforce, marketing has no immediate visibility into whether their campaigns are actually producing revenue. They check dashboards once a week, or worse, ask in Slack. When a deal stalls, nobody connects it back to the original campaign source to understand what went wrong.

Meanwhile, critical signals get trapped in the wrong system. A prospect who received a demo (tracked in Salesforce) starts engaging with competitor comparison content on the website (tracked in HubSpot). Neither system alerts the account executive, and the competitive threat goes unaddressed until the deal is lost.

Slack channels dedicated to deals, accounts, or pipeline reviews become graveyards of outdated information because updating them requires someone to manually pull data from two different platforms and synthesize it.

Why an AI agent, not just automation

Native integrations between Salesforce and HubSpot exist, and both platforms have Slack apps. But these connections are mechanical — they sync fields and post notifications. They cannot reason about what the data means for your revenue process.

An AI agent understands the relationship between marketing engagement and sales readiness. When a lead’s HubSpot engagement score suddenly spikes — three page views, a whitepaper download, and a pricing page visit in one afternoon — the agent does not just pass along a raw notification. It checks Salesforce to see if this lead already has an open opportunity, identifies the assigned rep, reviews the deal stage and last activity date, and crafts a Slack message that tells the rep exactly why this moment matters and what to do next.

The agent also resolves conflicts between systems intelligently. When HubSpot and Salesforce have different email addresses for the same contact, or when lifecycle stages are out of sync, the agent can flag the discrepancy and suggest a resolution based on which system has the most recent data. Simple automation would either overwrite one with the other or silently ignore the conflict.

Revenue intelligence requires synthesis across data sources, and that synthesis is inherently a reasoning task. The agent acts as a RevOps analyst who never sleeps, constantly monitoring the intersection of marketing activity and sales motion and surfacing what matters to the right person at the right time.

How it works with ACP

The Agentic Control Plane gives your agent authenticated, scoped access to all three platforms. Here is how the agent uses each tool:

Salesforce tools: Read and update leads, contacts, opportunities, and accounts. Query custom objects and fields. Read activity history and task records. Create tasks and log activities. The agent works with your existing Salesforce data model, including custom fields your team has added.

HubSpot tools: Read contacts, companies, and deals. Access marketing engagement data including email opens, page views, form submissions, and content downloads. Read workflow enrollment history. Update contact properties and lifecycle stages. The agent can see the full marketing journey for any prospect.

Slack tools: Post messages to channels and threads, read channel history, search conversations, and send direct messages when configured. The agent delivers insights where your team already works rather than requiring them to check dashboards.

Here is an example of the agent coordinating across all three platforms:

9:14 AM — Agent posts in #sales-alerts: “Priority alert for @mike: Your contact Lisa Chen at Acme Corp (Opportunity: $85K, stage: Proposal Sent, last activity: 12 days ago) just downloaded our ROI Calculator and visited the pricing page three times this morning in HubSpot. Her engagement score jumped from 42 to 78. This looks like renewed buying interest after a quiet period. Suggested action: send a follow-up email referencing the ROI calculator she downloaded.”

Mike in #sales-alerts: “Thanks, that’s perfect timing. I was about to deprioritize this one.”

Agent: “I have created a follow-up task in Salesforce on the Acme Corp opportunity due today, and updated the HubSpot contact timeline with a note that sales was alerted about the engagement spike.”

This coordination would have required Mike to independently notice the HubSpot activity, cross-reference it with his Salesforce pipeline, and remember the deal context — all while managing thirty other opportunities.

Governance and security

Revenue data is among the most sensitive information in any organization. Access to deal values, customer contacts, and marketing engagement data must be tightly controlled.

Salesforce OAuth scopes: The agent authenticates via OAuth 2.0 with scopes limited to api and refresh_token. Within Salesforce, object-level and field-level security still apply — the agent respects your existing Salesforce permission sets. If your org restricts access to revenue fields for certain profiles, the agent cannot see them either. The ACP admin selects which Salesforce objects the agent can read and write.

HubSpot OAuth scopes: The agent requests crm.objects.contacts.read, crm.objects.deals.read, content, and analytics.read. It can read marketing engagement data but cannot modify marketing emails, workflows, or campaign settings. Write access is limited to updating contact properties.

Slack OAuth scopes: The agent uses chat:write, channels:history, and channels:read. It posts only to channels it has been invited to, and it cannot access private channels or direct messages unless explicitly granted by a workspace admin.

Data isolation: The agent processes data in transit but does not store copies of your CRM records. Salesforce and HubSpot data are queried in real time when the agent needs them and are not cached between interactions. This means your existing data residency and retention policies remain in effect.

Audit and compliance: Every cross-platform action is logged. If your compliance team needs to verify that customer data was not improperly shared between marketing and sales, the ACP audit trail shows exactly which records were accessed and what information was included in Slack messages. This is particularly relevant for organizations subject to GDPR or CCPA, where demonstrating controlled access to personal data is a regulatory requirement.

Example use cases

• Marketing-to-sales handoff: When a HubSpot lead hits MQL status, the agent creates or updates the Salesforce lead, assigns it based on territory rules, and notifies the SDR in Slack with the lead’s full engagement history and recommended talk track.

• Deal stall intervention: When a Salesforce opportunity has been in the same stage for more than a configurable number of days, the agent checks HubSpot for any recent prospect engagement and posts a Slack summary to the rep with suggested re-engagement strategies.

• Campaign-to-revenue attribution: At the end of each week, the agent traces closed-won Salesforce deals back to their original HubSpot campaign source and posts a revenue attribution summary in the #marketing-results channel.

• Competitive intelligence alerts: When HubSpot detects that a prospect associated with an open Salesforce opportunity is viewing competitor comparison pages, the agent immediately alerts the account executive with battlecard links.

• Pipeline review preparation: Before weekly pipeline meetings, the agent compiles a Slack-formatted briefing for each deal, combining Salesforce stage and value data with HubSpot engagement trends and recent Slack discussions about the account.

Getting started

1. Sign up at cloud.agenticcontrolplane.com — free tier includes connections to all three platforms.

2. Connect your tools by authenticating Salesforce, HubSpot, and Slack via OAuth. Choose which Salesforce objects, HubSpot properties, and Slack channels the agent can access.

3. Describe your workflow in plain English. Specify your MQL criteria, deal stage thresholds, and which Slack channels correspond to which sales teams. The agent adapts to your revenue process, not the other way around.