Home › Case studies › Valfonta
Case study · Industry · Customer service
Valfonta case study: they stopped answering the same question a hundred times
Their technical team spent all day answering the same questions. Today an AI assistant answers them in seconds, even in the middle of the night, and only what truly needs a person reaches one.
At Valfonta, customers' technical questions were eating up the team's time, and those that arrived after hours went unanswered until the next day. We added an AI-powered assistant to their website that has learned from the company's own documentation: it answers instantly, at any hour, and alerts a person when an inquiry calls for it.
24/7
on duty, nights included
−55%
repeat questions to the team
<1 min
response time
+30%
inquiries captured after hours
* Metrics and testimonial pending confirmation with the client.
The context
Valfonta sells technical products, and that has a consequence: before buying, customers ask questions. Dimensions, materials, compatibility, lead times. Legitimate questions, almost always the same ones, that ended up on the desks of the company's most knowledgeable people.
People who buy technical products don't buy on impulse. They have a construction project, a machine, or a renovation underway, and they need to be sure the part fits before signing anything. That's why they ask: they're not haggling, they're looking for peace of mind. And until they have it, they don't move forward.
What's more, they almost never ask in just one place. They open three or four websites of similar suppliers and send the same question to all of them. From there, the decision comes down to something very simple: whoever answers first and best has the edge, because they show they know their stuff.
The result was twofold: a technical team stuck in their inboxes, and a pile of inquiries that came in during the evening or on weekends and went unanswered for many hours. In a sale where the first to respond wins, those are opportunities going cold.
The challenge
On closer inspection, the problem wasn't a lack of information: it was that all that information depended on someone looking it up and writing it out again. Each answer took a few minutes, but those minutes were multiplied by dozens of emails a day.
- The same questions, over and over, taking hours away from the people who add the most value.
- After hours, nobody answered: and customers don't wait.
- All the information existed, but it was scattered across documents and product sheets that someone had to go and dig up.
- Inquiries that were lost simply because they came in too late.
The biggest cost was the one that never showed up on any balance sheet: interruptions. An engineer working on a complex quote who stops three times to answer a question about a dimension doesn't lose three minutes; they lose their train of thought. And getting it back takes far longer than the email does.
Burnout was a factor too: repeating the same diameter for the tenth time motivates no one. And meanwhile, Saturday morning's email sat there waiting for Monday, from a customer who by then had already requested a quote somewhere else.
The solution: an assistant that knows its stuff
We didn't install one of those generic assistants that answer anything. We set up one that only talks about what it knows: Valfonta's products and documentation. The difference seems small, but it changes everything, because it determines whether customers trust the answer.
1 · Organizing what they already had
We gathered and organized the technical documentation, the product sheets, and the most frequently asked questions. That's the material the assistant learns from: the company's own verified information, nothing made up.
It was the most time-consuming part and also the one that delivered the most value on its own: reviewing everything uncovered incomplete product sheets and data that existed only in two people's heads.
2 · Putting it where the customer is
The assistant lives on the website, in plain sight. Customers ask in their own words, the way they'd ask a sales rep, and get an answer right away.
Nobody has to fill out a form or look up a phone number. Someone visiting at 11 p.m. on their phone types the question exactly as it comes to mind and moves ahead with the purchase without leaving the page.
3 · Knowing when to bring in a person
It handles the repetitive questions on its own. When it detects an important inquiry (a quote, a special case, a large customer), it hands it off to the team with full context, so no one has to start from scratch.
The sales rep doesn't get an empty alert: they get what the customer was looking for, what they were told, and where things stand. They can call already knowing what the conversation is about, which is the exact opposite of opening with “How can I help you?”
4 · Making sure no contact slips through
Every conversation is logged alongside the customer's record. What used to be an anonymous question at 11 p.m. is now an identified contact you can follow up with.
That makes something as simple as calling the next day and saying “I saw you were asking about this part number” possible. Customers feel heard, and the company stops throwing away interest it had already paid to generate.
5 · Teaching it to say “I don't know”
An assistant that answers everything is a liability. We gave it clear limits: if a question goes beyond what's documented, it doesn't fill the gap with guesswork. It says so and offers to connect the customer with a person.
It may look like a limitation, but it's exactly the opposite. Admitting a limit builds more trust than a brilliant but wrong answer, especially when that answer determines whether a part fits an installation.
6 · Improving it with what customers ask
Every month, the team reviews what was asked, what was answered well, and where the assistant had to hand off the inquiry. That's used to fill in the documentation, so next time it answers on its own.
That review is valuable in itself: it's the market's real list of questions, written by the customers, and it helps improve the product sheets and even the sales team's pitch.
What changed day to day
Before, a typical morning started with an inbox full of emails that had piled up since the previous afternoon. Many of them were short questions: a dimension, a material, a lead time. Answering them ate up the first hour and a half of the day, the most focused time, the best for thinking.
Now that hour and a half no longer exists as a task. Repeat questions are answered the moment they're asked, many of them in the middle of the night. What the team finds when they arrive is a handful of alerts for inquiries that really do need a person, each with the previous conversation right in front of them.
The change shows most in the mood: the team is no longer putting out fires but choosing whom to call first. And when a tough question comes in, they can give it the time it deserves.
Results
Customers get an answer instantly, at any hour. The technical team has won back a good share of its time and spends it on what truly requires a person. And the company no longer loses the inquiries that arrive when the office is closed.
Repeat questions reaching the team have been cut by more than half. They haven't disappeared, and that was never the goal: the ones that still come in are the ones that matter, the ones a professional needs to handle because there's a nuance, a volume, or a decision involved.
Business hours stopped being a barrier. A significant share of inquiries come in after hours, when customers finally have some quiet time to look at suppliers. That window used to be empty; today it's selling time like any other.
And there's an effect that's hard to measure: peace of mind. Knowing that nothing goes unanswered and that every contact is logged takes a small but constant worry off everyone's shoulders.
Why this case matters
- The knowledge you already have in documents can work for you around the clock instead of sitting in a folder waiting for someone to open it.
- Answering fast sells: in technical sales, the first to respond has the edge, even if they're not the cheapest option on the list.
- Well-scoped AI takes work off the team's plate without overreaching in its answers: it knows its subject, and when it doesn't know, it alerts a person.
What it includes
It isn't an off-the-shelf product you just plug in: it's a set of pieces custom-built for each company, starting with the documentation that already exists and ending with measurement. Here's what we put in place at Valfonta.
AI assistant on the websiteIt learns from their documentation24/7 answersAlerts the team when it mattersEvery contact loggedMeasuring results
“There used to be questions that went unanswered until the next day. Now a customer asks at 11 p.m. and gets the answer right away. And when it's the kind of inquiry that needs a person, it reaches us with full context: we know what they need before we pick up the phone.”
— Valfonta management
Frequently asked questions
What if the assistant makes up an answer?
It's the first question everyone asks, and rightly so. That's why we don't use a generic assistant: it answers only from the company's documentation, and its scope is tightly limited. If something falls outside that scope, it doesn't improvise: it alerts a person.
In technical sales, a wrong spec isn't a minor slip: it can end with a part that doesn't fit and an unhappy customer. That's why we'd rather it answer fewer things and get all of them right.
Don't customers prefer to talk to a person?
They prefer to get an answer. For a dimension or a lead time, an immediate answer beats waiting until the next day. And when the inquiry is important, the assistant passes the baton to a person; it doesn't get in the way.
In fact, something interesting happens: many people ask things they would never put in an email because they seem too basic. Those small questions are often the ones that stall a purchase.
Is it a lot of work to maintain?
Very little. It's updated when products or documentation change, just as you'd update a catalog. The heavy lifting is the initial setup: organizing the information you already have.
From then on, all it takes is a periodic review of what customers are asking: a little time each month, and the team usually enjoys it, because it generates ideas for the website.
How long does it take to go live?
It depends mostly on the state of your documentation. If your product sheets and catalogs are reasonably organized, things move quickly. If scattered information has to be pieced back together, that phase drives the rest of the schedule.
What we do aim for is getting it up and running early with a limited scope, instead of waiting months for everything to be perfect. Better for it to answer the ten most common questions well and grow from there.
Does it work if my products are highly specialized?
The more specialized the product, the better it works. An assistant that talks about everything adds nothing; one that knows a specific technical catalog inside out solves exactly what no one else can.
The one requirement is having the information written down somewhere: product sheets, manuals, compatibility charts. If it only lives in two people's heads, the first step is to get it out of there, and that's worth doing anyway.
Can it handle multiple languages?
Yes. It understands the question in whatever language it's written in and replies in that same language, backed by the same documentation. For an exporting company, that means serving international customers without depending on who happens to be in the office that day.
It's a good idea to check how the technical terms read in each language, because that's where credibility is won or lost. It's a short review, and it only needs to be done once.