About 25% of Enchanted Wares' qualified leads were landing in a spam folder. We built a custom AI agent to monitor and reroute them — and the client reports a 10% increase in conversions.
Enchanted Wares was losing about 25% of its qualified leads to a spam folder. We built a custom AI agent that watches the affected inbox, identifies real leads inside it, and reroutes them to the right place. The client reports a 10% increase in conversions, attributed directly to the agent. Engagement: June–July 2026, under $10,000.
That's the whole case study in four sentences. The rest is how it happened — and it's the answer we give when someone asks the only question worth asking an AI agent development company: can you point at an agent you shipped and a number that moved because of it?
This is the first client case study in our portfolio, and it sits inside our AI automation work.
Enchanted Wares is a B2C recreation company in Tempe, Arizona — "an end-to-end entertainment solution," in the words of owner Madeline Sisco — running a team of twelve.
Its growth depends on inbound inquiry: people find the company, message it, and book. That makes the inbox the front door. Anything blocking the front door isn't an IT annoyance; it's a revenue problem wearing an IT costume.
They found us through an online search plus a referral, and chose us on three grounds they named directly: pricing that fit the budget, culture fit, and value for the cost.
About 25% of qualified leads were messaging the wrong inbox, and those messages landed in a spam folder.
The workaround was the part that hurt. Somebody had to manually check every single spam email, just in case one of them was a customer. That's two costs stacked on each other:
Note what this problem is not. It isn't a chatbot problem, a content problem, or a CRM problem. It's a routing problem, and routing problems are where AI agents earn their keep quietly.
A custom AI agent that continuously monitors the affected inbox, distinguishes qualified leads from everything else in it, and reroutes them to the correct inbox.
Alongside it, the agent is integrated with social media for lead capture and redirects, so the capture path is covered where this audience actually starts the conversation rather than only at the end of it — the same integrate-with-what-you-already-use approach we take on most automation work.
It also ships with ongoing support if it ever breaks. That was part of the deal, not an upsell after handoff. An agent sitting in the middle of a company's lead flow is infrastructure; infrastructure without a support path is a liability you haven't been billed for yet.
The scope was deliberately narrow: monitor, classify, route. No attempt to rebuild the inbox, replace the tooling around it, or turn a two-month engagement into a platform migration. The smallest thing that fixes the actual problem is usually the thing that ships — and shipping is what produces the number in the next section.
Conversions increased by 10%, and the client attributes that increase directly to the new AI agent.
That is the client's number, from the client's own review — we didn't compute it, and we're not going to dress it up with an ROI model built on assumptions they never made. A 10% conversion lift on a business whose leads were partly disappearing into a spam folder is a plausible size for the fix, which is roughly the most we can honestly say about the mechanism.
The second result doesn't show up in a conversion figure at all: the reason the agent was built was to take the spam-folder chore off a person. Automation that only produces revenue is worth buying. Automation that produces revenue and removes a job nobody should have been doing is worth telling other people about.
The engagement scored 5.0 across quality, schedule, cost, and willingness to refer — verifiable in the client's review on our Clutch profile, which is the point of citing it rather than paraphrasing it here.
"How approachable yet professional Larsen Code Labs was was impressive."
— Madeline Sisco, Owner, Enchanted Wares
Work started as soon as the deposit cleared, and the build was delivered ahead of schedule. Communication ran through virtual meetings and email or messaging, with daily updates on progress. The client also had direct phone access for anything urgent during ongoing support.
Two things the client said about working with us that we'd rather quote than characterize:
"We worked with 2–5 teammates from Larsen Code Labs."
"At the price I paid, it was an enormous value."
For anyone evaluating us — or evaluating anybody else — here's the honest read on what a project like this does and doesn't demonstrate.
What it proves: we can take a business problem stated in plain language ("our leads are going to spam and someone has to dig them out"), turn it into a narrow agent with a defined job, integrate it with the systems where the leads originate, deliver it inside two months and under $10,000, and support it afterward. There's a named client, a named owner, a published review, and a number they attached to it themselves.
What it doesn't prove: that your inbox, your lead patterns, or your definition of "qualified" look anything like theirs. They almost certainly don't. Requirements discovery is most of the work in an engagement like this, and it's why the first conversation is a conversation and not a quote.
Have a version of this problem? Leads going somewhere they shouldn't, or a manual chore that exists purely to catch what your tools miss — tell us what it costs you now, and we'll tell you honestly whether an agent is the fix. Start a conversation, or see the rest of what we build with AI.