How we built an AI intake agent in 4 days — and what it actually does

The intake agent is almost always the first one we ship. Not because it is the most technically interesting — it is not — but because it eliminates the most hours for the least risk, and it produces visible results in the first week of operation.

Here is exactly what we built for a Manila-based professional services firm, how long it took, and what it does.

What the intake process looked like before

The firm receives forty to sixty inbound inquiries per month — a mix of forms, emails, WhatsApp messages, and the occasional cold call forwarded by reception. Before each inquiry could be routed to the right team member, a coordinator had to:

1. Acknowledge the inquiry (average response time: 4 hours)
2. Send a qualification questionnaire (separate email, manually composed each time)
3. Wait for the questionnaire back (average turnaround: 48–72 hours, often with a follow-up)
4. Classify the inquiry by service type and client profile (judgment call, 10–15 minutes per inquiry)
5. Route to the appropriate team member with a summary
6. Log in the CRM

Total coordinator time per inquiry: approximately 45 minutes across the full cycle. Forty inquiries per month equals thirty coordinator-hours — or roughly one full working day per week spent on intake alone.

What we built and how long it took

The intake agent runs on Make.com for orchestration, with a Claude API call for the qualification and routing logic, and writes to Airtable as the CRM. The build took four days, including QA.

Here is the flow:

1. Trigger. Any new inquiry — web form, email to a monitored inbox, or WhatsApp via a connected number — hits the Make.com webhook.
2. Acknowledge. An instant response goes out within ninety seconds. The message is warm and specific to the inquiry source — not a canned autoresponder.
3. Qualify. If the inquiry is thin on detail, the agent sends three targeted follow-up questions (not a form — a message) and waits for the reply. The questions were written with the team and reflect their actual qualification criteria.
4. Classify and score. The Claude API call reads the full inquiry thread and produces: a service-type classification, a priority score (1–5), a one-paragraph summary for the team member who will take the next call, and a recommended next step.
5. Route and log. The Airtable record is created with the full classification, the summary, and the priority score. A Slack notification goes to the relevant team member with everything they need before the first call.

The coordinator now reviews the Airtable queue once in the morning and once in the afternoon. The qualification work has moved from her calendar to the agent. Her judgment — knowing which clients need a faster response or a softer touch — now applies to the escalations, not the routine.

What it produced in the first month

Response time dropped from four hours average to ninety seconds. The coordinator's intake workload dropped by approximately twenty-two hours per month. The firm's conversion rate from inquiry to proposal improved — not because the agent is better at sales, but because responses are faster and the team member walks into every first call already knowing the client profile.

The agent costs approximately ₱12,000–₱18,000 per month to run (API calls, Make.com, Airtable). The coordinator's freed time, redirected to billable client work, is worth roughly eight times that at the firm's standard billing rate.

What this took from the owner

Two working sessions in week one (about four hours total) to map the qualification criteria and approve the message templates. A sign-off review at the end of week two. After that, the agent runs without owner involvement.

The four-day build timeline assumes the owner can show up to those two sessions and answer questions about their own intake logic. That is almost never the bottleneck — owners know how their business works. What they have not had is someone to translate that knowledge into a running system.

The intake agent is not the destination

We ship the intake agent first because it pays for the engagement in the first month and builds the team's confidence that agents are real, not theoretical. From there, the roadmap typically moves to ops reporting, then knowledge base, then client delivery — each one building on the infrastructure the previous one established.

The Discovery Workshop is where we build that roadmap for your specific operation — ninety minutes, your AIQ in hand, a prioritized list of which agent comes first and why.