A customer story worth studying closely
Zapier recently published a customer story by Rob Ayre, Customer Advocacy Marketing Manager at Zapier, titled How a recruiting agency uses Claude and Zapier MCP to find the candidates who aren't looking. It follows Curtis Key, CEO of Blue Castle Agency, a recruitment firm that works only in real estate and property management.
Curtis is not new to automation. He has been building it for about 20 years, first with basic code and scripts, then in Salesforce at a real estate brokerage that grew from 8 to 105 agents. When hiring became that brokerage's hardest problem, he founded Blue Castle.
The story is a useful blueprint for anyone building AI agents on Zapier, Make or n8n. It also points to a question many builders skip: once your agent works, how do people actually use it? That is where Chat4Agent comes in.
What Blue Castle built
Blue Castle's starting point is simple. The best candidates aren't looking for a job. They are already employed, and a recruiter has to find them and earn a conversation. Most of the finding is repetitive, so Blue Castle handed it to Claude.
Here is how the system works, according to the source:
- Sourcing runs several times a day. Claude picks active jobs from the applicant tracking system and reads everything the company knows about each one: success profiles for the company and the role, plus notes on the segment (multifamily, associations, affordable housing or luxury). Then it uses Blue Castle's sourcing tools to build a candidate list.
- A second Claude pass handles quality control. Later in the day, another run checks the list with the same context, the step a human used to do.
- Submissions are drafted on Blue Castle's own server. When a recruiter is ready to submit a candidate, a process reads their notes and the Zoom transcript, checks them against the job and success profile, scores the candidate, researches the companies on the resume and writes the submission.
- Zapier MCP connects everything. Claude can reason about a candidate, but it can't reach into an applicant tracking system, a sourcing platform or a Google Sheet by itself. Zapier MCP gives it access to the tools Blue Castle already used, without a custom integration for each one.
The results
The numbers reported in the story speak for themselves:
- Human sourcers' candidates are approved 78 to 83 percent of the time. Claude's are approved about 94 percent of the time.
- Curtis estimates that sourcing and quality-control automations save the team 240+ hours per week at current volume.
- The submission process saves each recruiter an additional 4+ hours a week.
- Sourcing output is up about four times.
Recruiters now start their day with a list that's already built and checked, candidate meetings already booked and a submission ready to go. Curtis sums up the philosophy this way:
"It all comes down to context. The technology is all there. We just have to provide the understanding—Zapier MCP helps make that possible."
And on scaling:
"This is all infinitely scalable. When we need more sourcing capacity, we don't need to start an internal hiring process. We just turn the knob."
The layer most builders forget: the front door
Blue Castle's setup works because Curtis designed it as what he calls an end-to-end experience for recruiters: every step makes sense, and a recruiter can't get lost in the process. That phrase deserves attention, because it is the part many agent projects get wrong.
Most AI agents built on automation platforms live inside a scenario, a workflow or a Zap. They are powerful, but they have no face. The people who should benefit from them, recruiters, sales reps, support staff or a client's team, often have no natural way to interact with them. They depend on whoever built the automation to trigger it, tweak it or explain what came out.
Blue Castle has 20 years of automation experience and its own server for custom processes. Most teams don't. For them, the gap between "the agent works" and "my team uses the agent every day" is where projects stall.
What happens when an agent has no usable interface
If you have built an agent and stopped at the backend, you have probably seen some of these symptoms:
- The agent is used by one person: its builder. Everyone else finds it too technical or doesn't know it exists.
- Requests come through side channels. Colleagues send messages asking you to "run the thing," and you become the bottleneck the agent was supposed to remove.
- Clients don't see the value. If you are an agency, a workflow diagram is hard to sell. A client wants something they can open, type into and get an answer from.
- Custom front-end work eats the budget. Building a chat interface from scratch means design, authentication, hosting and maintenance, before you have even improved the agent itself.
The cost is not just time. An agent nobody uses produces no hours saved, no approval rate gains and no output multiplier. Blue Castle's 240+ hours a week only exist because recruiters actually work with what the system produces.
How Chat4Agent closes the gap
Chat4Agent is built for exactly this layer. It gives AI agents built on Make.com, n8n, Zapier and similar platforms a chat interface, without requiring technical skills.
You keep the agent you already built: your workflows, your context, your connected tools. Chat4Agent adds a professional, customizable chat interface on top of it, ready to deploy. Your team or your clients get a single, clear place to talk to the agent, instead of a workflow they don't understand.
Think about the Blue Castle model through that lens. The heavy lifting (sourcing, quality checks, scoring, writing) happens in the automation. What a recruiter needs is an easy way to ask questions, request a check or get a draft. A chat interface is the most natural way to do that, because it matches how people already work: they ask, and they get an answer.
Before and after: what changes for your users
The following scenarios are illustrations of how a chat interface changes the experience, not claims about Blue Castle's own setup.
Before: an operations agent built in n8n can summarize a client file, but only the builder knows how to trigger it with the right inputs. Colleagues email the builder, wait, then copy the result into their own tools.
After: the same agent sits behind a Chat4Agent interface. A colleague opens it, asks for the summary in plain language and gets the answer directly. The builder goes back to improving the agent.
Before: an agency delivers a Make.com scenario to a client and spends meetings explaining how to use it.
After: the agency delivers a branded chat interface the client's team can use from day one. The agent becomes a product, not a diagram.
Built for three kinds of builders
No-coders
If you are an entrepreneur, marketer or operator who built an agent with no-code tools, you shouldn't have to learn front-end development to share it. Chat4Agent lets you put your agent in front of real users without writing code, so the value you built in your automation platform actually reaches the people it was meant for.
Developers
You could build a chat UI yourself. The real question is whether you should. Every hour spent on interface plumbing is an hour not spent on context, prompts and integrations, which is where Curtis says the value really lies. Chat4Agent lets you skip the interface work and move straight to production.
Agencies
Blue Castle is itself an agency that used automation to multiply output without multiplying headcount. If you sell AI agent services, Chat4Agent helps you do something similar: deliver a polished, customizable interface for each client without rebuilding one every time. That makes your offer easier to sell, easier to demo and easier to repeat.
Context and judgment, with a clear way in
One of the most interesting parts of the story is where Blue Castle is heading next. The team has started building Next-Gen Zaps with Claude over MCP. One runs at 2 a.m. on weekdays, builds prep sheets for the day's candidate screening calls and posts a morning note in Slack. Another logs each closed-won deal to a sheet for tracking active clients.
Curtis's next question is where Claude is repeating work that a Zap would do the same way every time. Moving those steps into Zaps leaves Claude the judgment calls and should bring the cost down.
This is a mature way to think about AI agents:
- Deterministic steps go into automations. Same input, same output, every time.
- Judgment goes to the AI. Reading context, scoring, writing, deciding.
- People get a clear way to interact with both. This is the part a chat interface handles.
Chat4Agent fits into the third layer. It doesn't replace your Zaps, scenarios or workflows. It makes them accessible, so the judgment calls your agent is good at can be requested by anyone on the team, at the moment they need them.
Curtis also believes the best technology starts with a vision, and that creating tech for its own sake leads to fragmented stacks and inefficiencies. A chat interface that sits on top of the tools you already use supports that vision: one entry point instead of yet another disconnected app.
Why the interface is what makes the agent count
Blue Castle's story is ultimately about people. As Curtis puts it:
"Recruiters should be spending their time doing what they do best—connecting with people."
The agents exist to free up human time. But that only happens when humans can reach the agent without friction. A powerful backend with no usable front end is like a well-stocked warehouse with no door.
If you are building agents for your team or your clients, ask yourself three questions:
- Can someone other than me use this agent today, without my help?
- Would a client understand its value in the first minute of a demo?
- Am I spending time on interface work instead of improving the agent?
If any answer makes you hesitate, the interface is your next priority.
Give your agent a face today
You have already done the hard part: building an agent with real context and real connections to your tools. Now let your team and your clients use it.
Try Chat4Agent for free and give your Make, n8n or Zapier agent a professional chat interface in minutes, with no code to write. Your agent already does the work; with Chat4Agent, everyone can finally use it.
Source: How Blue Castle Agency uses Claude and Zapier MCP to source candidates | Zapier