Sidekick built an AI you reach by text message

The most-used app on a factory floor is not a factory app. It is the text message thread every worker already has open. Sidekick, a company from Y Combinator's Summer 2026 batch, decided to meet workers there instead of asking them to download one more thing.

The pitch is short: a worker texts a question, and an answer comes back - pulled from the company's own manuals, checklists, and safety documents. No app to install. No password to reset. It works on the cheapest phone in the building, and it answers in whatever language the question was asked in.

That constraint sounds small. It is the whole idea. Roughly 80% of American workers do not sit at a desk, and most workplace software was built for the other 20%. The people running machines, fixing cars, and stocking shelves have mostly been handed tools designed for someone with a laptop and a quiet minute to log in. They rarely have either.

The OriginA request from the founder's father

Sidekick did not start as a market analysis. It started as a favor. Founder and CEO Justin So has a father who owns a manufacturing facility, and the ask was practical: build something to help with onboarding new workers, training them, and keeping the plant's knowledge from walking out the door.

So had the background to try. He holds a Masters in Computer Science from the University of Pennsylvania and had worked in finance at the chip-design firm Arm before turning to the problem full time. What he saw on his father's floor is familiar to anyone who has run an operation: the answer to almost every question already exists. It just lives in the wrong places.

The knowledge existed. It always existed. It just lived in one veteran's head, a paper binder, and three systems that don't talk to each other.

When a machine jams or a new hire freezes, the fix is usually known by someone - a supervisor, a manual in a drawer, a training video no one can find. The gap is not knowledge. The gap is reach. Sidekick's bet is that closing that gap is worth more than any new information it could add.

There is a cost to that gap, and it shows up in the parts of the day nobody logs. A line that sits idle for twenty minutes because the one person who knows the restart sequence is on break. A part ordered wrong because the spec lived in a binder two rooms away. A supervisor's phone buzzing after hours with a question that has been answered a hundred times before. None of these are dramatic. Added up across a plant across a year, they are the real tax on not being able to reach an answer in the moment you need it.

How It WorksFour steps, one text thread

A manager uploads the company's documents and video walkthroughs. Sidekick turns those into a searchable knowledge base and step-by-step tutorials. From there, the worker never sees any of that machinery. They just text.

Two details make it fit the environment rather than fight it. A worker with dirty or gloved hands can send a voice memo instead of typing. A worker who does not know what a part is called can send a photo of it, and Sidekick identifies the equipment. These are not showpiece features. They are answers to the specific reasons software fails on a loud, dirty floor.

A representative exchange. Sidekick answers in the language the worker used.

The DifferenceWhy not just build an app?

The frontline-software market is not empty. Maintenance platforms like MaintainX, UpKeep, and Limble handle work orders. Communication apps like Beekeeper and WorkJam try to reach deskless teams. Most share one requirement that quietly sinks adoption: install the app, create a login, learn the system.

Sidekick removed all three. That is the difference, and it is easy to underrate until you have watched a rollout stall because half the floor never got past the download screen. Texting is the rare interface that needs no training. A worker who has never opened a business app has sent thousands of texts. The learning curve is a phone number.

The language point matters here too. A plant floor in the United States is often multilingual, and asking a worker to translate their own question before a system will understand it is its own kind of friction. Sidekick takes the question in Spanish, Vietnamese, or Tagalog and answers in kind, from documents that may only exist in English. That is not a headline feature. It is the difference between a tool a whole floor uses and one only the English speakers bother with.

There is a second advantage that compounds. Every time Sidekick escalates a question and a manager answers, that answer is captured and reused. The knowledge base is not something a busy supervisor has to sit down and write. It builds itself as a byproduct of people doing their jobs. Over months, that turns a company's scattered know-how into something searchable.

Documentation as a byproduct of the work, not a chore stacked on top of it.

The ProductMore than a chatbot

Around the core SMS assistant, Sidekick has assembled the pieces of a frontline operations platform - all reachable through the same text interface.

The BusinessWho pays, and why it might stick

Sidekick sells to the company, not the worker. The buyers are operators with frontline teams - manufacturing plants, auto dealerships, retailers, and the facilities and logistics groups that keep them running. The sales motion today is founder-led, with Justin So booking demos directly. Pricing is not published, which is normal at this stage.

The traction is the part worth watching. A single-founder company that is live in more than 20 plants with over 600 workers is not a demo waiting for users. It is a product people already reach for, and the fact that it requires no behavior change is exactly why it spreads inside a building. One worker texts a number. The person next to them asks how. That is the whole onboarding.

The MarketThe layer nobody built

Sidekick describes itself as the SMS layer for frontline operations. That framing is useful. It is not trying to be the system of record that stores everything, and it is not trying to be another dashboard managers log into. It is trying to be the thin, universal way a worker asks a question and gets an answer - sitting on top of whatever documents and systems a company already has.

Whether that layer becomes a large business depends on questions no one can answer yet: how deep the workflows go, how well the escalation loop holds up across very different industries, and whether incumbents copy the no-app approach once it is proven. What is clear is the shape of the opportunity. A very large share of the workforce has been served badly by software, and the tool that reaches them best is the one already in their pocket.

It is also worth noting what Sidekick is choosing not to do. It is not asking factories to rip out the systems they have. It is not selling a hardware rollout or a screen for every workstation. The wager is that the last mile of frontline software - the part where a human actually needs something - is a text away, and that owning that last mile is a durable position even as the systems underneath it change.

For now, Sidekick is a small team with a specific idea and early evidence that the idea works on real floors. Whether it grows into the default way frontline teams reach their own knowledge is unproven. But the starting point is unusually concrete: a founder who watched the problem up close, a product already in daily use, and a design decision - meet workers where they already are - that gets harder to argue with the longer you look at it.