Stratefuse

Four ways we extend your CRM.

HubSpot Breeze, Salesforce Agentforce, Claude MCP connections and custom API software connections. Each one is scoped, built and priced on its own.

1. HubSpot Breeze

HubSpot's built-in AI, configured and governed.

2. Salesforce Agentforce

Salesforce's built-in agents, built and governed inside your org.

3. Claude MCP connections

Anthropic's Claude, put to work in your business by Claude architects.

4. Custom API software connections

Your CRM connected to any system. Separate from the AI work.

1. HubSpot Breeze

HubSpot's native AI, built into the platform you already run. No external model and no external connection. We configure it, ground it in your CRM data and set the permissions it works within.

What it includes

  • Breeze Assistant set up for each team
  • Breeze agents deployed and tuned: customer, prospecting, content and data
  • Custom agents in Agent Builder, and Breeze Intelligence for enrichment and buyer intent
  • Permissions, approval steps and reporting on what the agents did

Best for: HubSpot customers who want AI inside the platform, with nothing leaving HubSpot.

2. Salesforce Agentforce

Salesforce's native AI agents, built inside your org. We design each agent's job, build it, test it against real records and govern what it may do.

What it includes

  • Agents built in Agent Builder with topics, instructions and actions
  • Prompt templates grounded in your Salesforce and Data Cloud records
  • Flow and Apex actions for anything the standard actions cannot do
  • Sandbox testing, escalation rules and usage reporting

Best for: Salesforce customers who want agents for sales, service and operations without adding another vendor.

3. Claude MCP connections

We put Anthropic's Claude to work in your business, as Claude architects. First we design the AI brain: what Claude knows, what it may do and how it is judged. Then we connect it to your CRM and your other systems through the Model Context Protocol (MCP), the open standard for giving a model controlled access to tools and data.

Our architects are completing Anthropic's Claude Certified Architect program, and our Claude work is built to its standards.

What it includes

  • The AI brain design, approved by you before anything is connected
  • Claude connected to HubSpot through HubSpot's MCP server and to Salesforce through Agentforce's MCP support, with scopes set per job
  • Claude connected to anything else with an API: ERP, support, billing, data warehouse, your own software. Where no MCP server exists, we build one
  • Automated jobs and multi-step agents, every step logged and every write reviewable
  • Delivered where people already work: HubSpot workflows, Salesforce Flow, Slack, email or a small internal app

Best for: one model across HubSpot, Salesforce and everything else you run, or jobs that need reasoning across systems.

4. Custom API software connections

Separate from the AI work. We build and maintain custom API connections between your CRM and any system: ERP, finance, support, e-commerce, industry software. Every connection is specified in writing, built on the system's API and monitored after launch. Where the far side needs specialist hands, our trusted partners work under our project management, so you have one contract and one point of accountability.

What it includes

  • A written specification: systems, direction of each data flow, trigger, field map, authentication and error handling
  • Connections to ERP and accounting, billing, support, e-commerce, telephony, data warehouses and industry software
  • Trusted partners for the far side: ERP configuration, vendor platforms, legacy systems
  • Monitoring with alerts, a runbook, the source code and a named owner at handover

Best for: data that has to move between systems correctly and on time, every time. No AI involved unless you ask for it.

Designing the AI brain

Every Claude engagement starts here, before anything is connected.

  1. Inventory the knowledge

    CRM data, documents, pricing, playbooks and policies. We decide what Claude may read and what stays out.

  2. Define the jobs

    Each use case written as a job: input, output, measure of success and the person who checks it.

  3. Set the rules

    What it may write, where a human approves, how it escalates and what it logs.

  4. Map the connections

    Which systems, through which MCP server or tool, with which scopes. Nothing connected that a job does not require.

  5. Connect, test, measure

    Sandbox first, then a pilot group, then expansion as accuracy, adoption and cost are measured.

The output is a written AI design document with a monthly cost estimate. You approve it before anything is built.

Common questions

Is a custom API connection an AI project?

No. A custom connection moves data between systems on fixed rules and needs no AI. It is scoped, built and priced on its own. If a step later needs judgment, Claude can be added as a separate line.

We already have Breeze or Agentforce. Why add Claude?

Breeze and Agentforce each work inside their own platform. Claude connects to both and to everything else you run, so one model can reason across systems, long documents and custom logic.

Will the AI change our CRM data without a person checking?

Only where you decide it should. Our default is that AI reads freely, writes to designated fields, and anything customer-facing waits for human approval until its accuracy has been measured on real records.

Who owns what gets built?

You do. Code, prompts, configuration and documentation are yours at handover, in your accounts and your repositories.

Book a call