UpFix: one AI Messaging Agent, four industries served
"The AI messaging has been quite strong. I stopped checking on it."
That's the highest praise a founder gives a system. Vitaly Sandyrev, Co-Founder and CEO of UpFix, stopped checking his AI Messaging Agent's work, because he felt it stopped needing to be checked.
UpFix has been repairing electronics from Buford, Georgia since 2006. Not one category of electronics: automotive systems, household appliances, industrial machinery, and fitness equipment. More than 200,000 customers and over a million parts serviced later, the pitch is unchanged. Don't buy the replacement. Send us the original and we'll fix it.
UpFix has run its customer communications on Aircall since 2024.
The challenge
A single support line at UpFix handles a customer whose Volkswagen touchscreen has gone dark, a customer whose oven control board failed, a body shop chasing an ABS module, and a customer whose treadmill console won't power on. Four industries, one queue, and no two questions alike.
The customers themselves raise the stakes. A large share of them arrive at UpFix because they've already been told “no.” The part is discontinued, or on indefinite backorder, or the dealer quoted a replacement that costs more than the appliance. As Co-Founder Ernest Martynyuk describes it on the UpFix site, the company exists "to bring rare, back-ordered, or obsolete parts back to life, fast."
That makes UpFix the last option for many customers, and last-option customers are invested. They ask detailed questions, and they ask questions at all hours.
This is exactly the volume and variety that breaks scripted automation. A decision tree can cover the top five questions on one product line. It cannot cover four industries.
The turning point
The unlock wasn't a decision to automate. It was noticing that the hard work was already done.
Over years of answering a wide range of questions, UpFix had written answers down in a variety of places: website FAQs, blog articles, markdown files, and custom company information built inside Aircall. The team had also already set up an Aircall AI Voice Agent for after-hours coverage, pointed at that same material.
So the question wasn't whether Aircall's AI could learn a catalog this wide. The voice agent had shown it could. The question was why messaging should start from scratch.
It didn't have to. Aircall's AI Messaging Agents connect to the same knowledge sources that AI Voice Agents already use. So, UpFix pointed them at the material it had spent years building.
The solution
UpFix now runs an always-on AI Messaging Agent on its main customer line, alongside an AI Voice Agent that covers after-hours calls.
Because the AI Messaging Agent reads the same FAQs, blogs, markdown files, and company information as the AI Voice Agent, there's one source of truth across channels. When UpFix documents a new repair service, both AI Agents learn it. Nothing gets updated in one place and forgotten in another.
For a business that keeps adding repair categories, that matters more than it would for a single-product company.
Answers that arrive while the customer is still holding the phone
The messaging agent runs 24 hours a day. A customer asking about an order at 1am gets an answer at 1am. Order status, shipping questions, and the routine questions that make up most of the queue no longer wait for a rep to reach them.
Actions that reach into HubSpot
Through AI Actions, the agent looks up order status and knows what was said in previous interactions.
Answering in multiple languages
UpFix primarily supports English and Spanish, the two languages its customer base uses most. But the agent isn't restricted to those languages. Vitaly found that out when a customer switched to Japanese mid-conversation.
"He started texting in Japanese, and it switched and started replying in Japanese."
For a business shipping repairs nationwide, that flexibility is practical rather than novel. Aircall’s AI Messaging Agent can be configured for over 85 languages.
The results
UpFix serves customers in four industries across every US time zone.
UpFix switched on its AI Messaging Agent in late May 2026. In its first two months it sent 1,064 replies to just over 400 customers, and it’s messaged with actual customers every single day since.
43% of replies went out outside 9am to 6pm, Monday to Friday
15% went out on a Saturday or Sunday
19% went out between 8pm and 8am
52% of customers received more than one reply
Having more than 40% of replies ocurring outside standard working hours tracks with the nature of UpFix’s business. An oven fails on a Sunday. A car won't start before a Monday commute. Customer questions arrive when the appliance breaks, not always when phones are staffed.
Over half of customers received more than one reply. Across a catalog spanning ABS modules to treadmill consoles, the agent isn't returning one-line acknowledgements. It's handling nuanced technical conversations.
Looking ahead
UpFix is expanding on two fronts.
On voice, the AI Voice Agent is set to graduate beyond after-hours-only and into business hours as a backup to the team rather than a replacement for it, with a handoff to a person whenever the AI can't finish the job.
On messaging, the team is eager to dig into performance insights to improve agent performance further. When the agent can't answer something, the dashboard will show you gaps in your documentation along with configuration recommendations.
For a company whose catalog keeps growing, a report that says "here's what your customers asked that your documentation doesn't cover" will be invaluable.
Advice to others
Vitaly's advice is to stop grading AI on whether it knows everything and start using it to find out what you never wrote down. Every unanswered question is a piece of missing documentation you can go fix, which makes the agent better and makes the next customer's answer better too.
It's the same philosophy that built the company. As he puts it:
"I believe in fixing what's broken, not just throwing things away."
Look how well this worked out for UpFix. Now let's take your customer communications to the next level.
