The Background
We're a cybersecurity firm based in Oakville, Ontario. We help professional services firms secure their Microsoft 365 and Google Workspace environments through transparent, month-to-month security services.
At the time of this project, we were early-stage. One client. Twelve managed seats. Roughly six months of runway. We had built our initial business through referrals and a handful of warm introductions, but growth had plateaued and the pipeline was thinning.
The Challenge
Our inbound approach had produced one anchor client, and nothing was replacing the momentum. Three proposals had gone cold in a row. We knew outbound prospecting was the answer, but the typical sales tooling stack presented its own problem.
A conventional outbound setup, combining a tool like ZoomInfo for contact enrichment, Outreach for email sequencing, and a CRM with paid integration layers, would run $300 or more per month. For a company managing twelve seats, that kind of overhead doesn't pencil out. Even assembling a lighter stack would mean weeks of evaluating vendors, configuring integrations, mapping data fields between systems, and debugging broken webhooks. All of that before a single prospect gets contacted.
We needed to build a functional outbound pipeline on a near-zero budget, without spending weeks on setup.
The Solution
We built an end-to-end prospecting workflow using three tools: Claude Pro ($100/month), Attio CRM (free tier), and Apollo.io (free tier with API access). LinkedIn Sales Navigator, which we already had, provided the initial prospect identification.
The critical piece was Attio's MCP (Model Context Protocol) connector. MCP is an open standard that lets AI assistants interact directly with external tools. In practice, it meant Claude could create CRM records, update contact fields, and manage pipeline stages inside Attio through natural conversation. No middleware. No manual data entry. No CSV imports.
The entire build happened in a single afternoon across four phases.
Identifying the target market. We defined four verticals in the Halton region: law firms, accounting practices, financial advisors and insurance brokerages, and management consulting firms. These are businesses where a single compromised email or ransomware incident can shut down operations, and where most lack dedicated IT staff. Using LinkedIn Sales Navigator, we pulled 33 contacts across 32 companies and organized them into a structured prospect list. This phase took about 30 minutes.
Building the CRM through conversation. Instead of manually clicking through Attio's interface to create records one by one, we used Claude and the MCP connector to handle all CRM setup. Claude created 33 people records and 32 company records. It built a "LinkedIn Outreach" list with a custom pipeline tracking each prospect through six stages: Not Connected, Connected, Messaged, Replied, Meeting Booked, and Proposal Sent. It tagged every record with its source for attribution and set all companies to Prospect status. Nobody opened the Attio UI during this phase. Twenty minutes, start to finish.
Enriching contacts through the Apollo API. We needed verified email addresses, LinkedIn profile URLs, and company domains for all 33 contacts. That kind of data normally requires either a paid enrichment integration or hours of manual lookup. Instead, Claude called Apollo's People Enrichment endpoint for each contact, passing their name, company, and title. Apollo returned the data, and Claude pushed it directly into the matching Attio records through the MCP connector. No export files, no import wizards, no field mapping configuration. The enrichment phase took roughly 40 minutes.
Executing outreach. With the CRM fully populated, we moved to execution. We sent 31 LinkedIn connection requests with personalized notes. The messaging varied by vertical. Law firm outreach emphasized regulatory compliance risks and client confidentiality. Accounting practice outreach focused on business continuity during tax season. As responses came in, pipeline stages were updated through Claude and the Attio connector without switching tabs or manually entering data. About 30 minutes for this phase.
The Results
The complete workflow took approximately three hours. Total monthly cost for the tool stack: $200.
- 33 prospects identified, enriched, and loaded into the CRM
- 32 companies created with verified domains and LinkedIn pages
- 30 of 33 contacts enriched with LinkedIn profile URLs
- 24 of 33 contacts enriched with verified email addresses
- 31 connection requests sent with vertical-specific personalized messaging
- 5 connections accepted within the first few hours
- 1 substantive sales conversation initiated, with a law firm principal asking about in-house versus outsourced IT
- Zero paid integration tools purchased
- Zero lines of code written
Two contacts were flagged for manual follow-up where Apollo's data didn't match. That's a normal enrichment gap that would exist with any tool.
The workflow also proved repeatable. Adding 30 new prospects takes roughly 45 minutes using the same process. As we expand prospecting into neighbouring regions, the playbook scales without additional setup or tool purchases.
One structural advantage stood out beyond the speed. Because the entire workflow ran inside a single Claude conversation, context carried forward through every step. Claude remembered which companies were law firms versus accounting practices, which prospects had already been enriched, and what messaging decisions had been made earlier in the session. That kind of continuity normally requires a dedicated RevOps person maintaining notes across multiple disconnected tools.
Doing This Without Creating a Data Problem
Here is the part I would be a hypocrite to skip, given what we do for a living. The moment you wire an AI assistant to an enrichment API and a CRM, you have built a small data pipeline that handles personal information about real people, and that pipeline deserves the same scrutiny we would apply to any client's. Before we ran a single enrichment call, we worked through the same questions we ask clients to answer about their own tools.
The first is what data actually flows where. In our case, names, titles, and companies went to Apollo, verified contact details came back, and that data landed in Attio. We checked each vendor's terms for how that information is retained and whether it gets used to train anything, because "convenient" and "compliant" are not the same word. The second is access. The API keys that let Claude reach Apollo and Attio are credentials, and a leaked key is a breach, so they live in a secrets manager rather than pasted into a prompt or a config file someone can stumble across. The third is the boundary on what goes into the model at all. A prospect's name and company is reasonable. A client's confidential data is not, and the line between "this is fine to send" and "this should never leave our tenant" has to be decided before someone is moving fast on a Friday afternoon, not after.
None of that slowed the build down in any meaningful way. It is maybe twenty minutes of thinking up front, and it is the difference between a clever workflow and a quiet liability. It is also exactly the discipline we bring to a client's environment when we run an AI Governance and Workflow Readiness Review: not a list of tools to ban, but a clear answer to what data is allowed to go where, who holds the keys, and how you would know if something went wrong. The firms that get burned by AI are rarely the ones who thought about this. They are the ones who never did.
Key Takeaway
Most small businesses stall on outbound prospecting because the setup cost feels wildly disproportionate to the expected return. Evaluating tools, paying for integrations, configuring data flows between platforms. It can easily consume weeks of effort and hundreds of dollars in monthly subscriptions before a single prospect hears from you. So it stays on the to-do list. Indefinitely.
MCP-connected AI tools compress that timeline. We applied the same thinking when we replaced $12,000 in annual SaaS costs by building four internal tools with Claude Code on existing infrastructure. By using Claude as the orchestration layer between Apollo and Attio, we eliminated every integration point. Claude reads from the enrichment API and writes to the CRM. That is the full integration. No middleware, no field mapping, no webhook debugging.
For business owners who have been putting off CRM setup and outbound pipeline because it feels overwhelming: the barrier has shifted. The tools are either free or close to it. The integrations happen through conversation. The bottleneck is no longer budget or technical complexity. It's deciding to start.
The one thing I would not skip, especially for a firm that handles client data, is the half hour of thinking about where that data goes before you let an AI tool start moving it around. That habit is what separates a smart shortcut from a future incident, and it is cheap insurance on a workflow that otherwise costs almost nothing to run. Move fast, by all means. Just know where your data is while you do it.