Replies ready before you even open your inbox

Project Overview

Who is Dryguard?

Dryguard is a specialist contractor in damp-proofing and waterproofing, based in Flemish Brabant. They treat rising damp with injection work, do basement waterproofing and tanking, install salt membranes, and handle the follow-up treatment. Their clients are homeowners, but also construction companies, architects and property managers across Flanders, Brussels and part of Wallonia. It's an SME with a few dozen employees.

Every day, a lot of emails come in with the same kind of questions. Price requests, technical questions about rising damp, scheduling, warranty, hygroscopic salts. Keeping up with that inbox eats into the time of exactly the people who are needed most on site and with clients.

The challenge

"How do we make sure clients get a faster reply, without us writing every email from scratch, and without every colleague using a different tone?"

The team already used AI here and there for one-off tasks, but not built into the way they worked day to day. The idea for a mail agent came up through the ongoing CRM project with Flowkify.

Project Delivery

The approach

We started with interviews with management and the operational team, and analyzed the existing email flows. From there we picked three priority mailboxes to start with, including one test mailbox. That let us adjust course without any risk to client communication.

Next, we securely brought together seven years of sent emails and extracted Dryguard's writing style from them. The result is a profile of their own voice: the Flemish construction terminology, when they use formal versus informal address, their go-to phrases, and the difference between their external and internal tone.

What we built

1 · The knowledge base

Dryguard's know-how, in one place.

• Six themes: rising damp, treatment and injection, salts and damage, follow-up treatment and finishing, pricing with warranties and subsidies, and practical client info.
• It's all kept in simple files that the team maintains themselves.
• The agent searches that knowledge base, hosted in the EU.

The benefit: knowledge that used to live in a few people's heads now sits in one place, and it also helps onboard new colleagues.

2 · The mail agent

An AI agent that drafts replies per mailbox.

• It fetches new emails.
• It decides, per email, whether a reply is needed, and leaves ads and noise alone.
• It looks up the right info in the knowledge base.
• It prepares a draft in Gmail, in Dryguard's voice, in Dutch or French depending on the sender's language.
• An employee reads the draft, adjusts it if needed, and sends it themselves.

So the agent never sends anything on its own. That was a deliberate choice: a person always stays responsible for what goes out. The agent goes through the inbox three times a day, so there are drafts ready every time someone opens their mail.

The benefit: nobody starts a reply from a blank screen anymore.

3 · The feedback loop

Every day we compare the agent's draft with the email that actually went out, and summarize what was changed. That shows us, in black and white, where the agent still gets it wrong, so we can fine-tune the knowledge base and the agent. No custom AI model is needed for that.

The benefit: the agent gets better based on real emails, not guesswork.

The technology behind it

The solution runs entirely on Google Cloud, since Dryguard already works with Google Workspace and Gmail. In practice that means: Google Gemini as the language model, the Google Agent Development Kit for the agent, Vertex AI Search for the knowledge base, and Cloud Run with Cloud Scheduler for timing. We didn't train a custom AI model — we brought existing building blocks together and tuned them to Dryguard's industry, voice and day-to-day practice.

Results

The mail agent is running in production on three mailboxes, with drafts ready three times a day. The whole project was live within one to two months.

For recurring questions, the estimated time saving is 30 to 50 percent per reply. Instead of writing from a blank screen, people now read and correct. Leads and clients get a faster reply, which matters in an industry where people request several quotes at once. The communication also sounds the same no matter who sends it. And the technical knowledge lives in a knowledge base the team manages itself.

The work doesn't stop there. Based on how the team uses the agent, we're making the drafts more visible in the inbox and working on processing emails as they arrive. After that comes the expansion to more mailboxes. Flowkify continues to support Dryguard as an AI partner.

Want to know what an AI agent could do with your inbox too? Get in touch with Flowkify for a tailored AI scan.