About
Agents are smart. They just don't know your business.
That sentence is the whole studio. Here is why we think it is the problem worth working on now.
The question changed.
Two years ago the question was whether a model could do the task at all. Draft the reply, sort the tickets, write the quote. Mostly, now, it can. The demos are easy and everyone has seen one.
What separates an agent that works from one that embarrasses you is no longer the model. It is whether the agent knows how this company does the task. What you refuse to do. Who gets called first. What "urgent" means here and not somewhere else. Which customer gets the owner's number and which gets the form.
None of that is in your documents, because nobody ever needed it to be. The people who held it were in the room.
tacit, adjective.
Understood or implied without being stated. Of knowledge: held by people, absent from documents.
Now there is something in the room that reads documents and nothing else. So the job is to write the knowledge down, in a form an agent can follow and a person can check. That is what we do. It is a writing job as much as an engineering one, and most of the time is spent with your best person, not with a model.
Why not just automate things?
Because automations encode the task and not the judgement. They work until the first case that needs a human's sense of the business, and then someone quietly goes back to doing it by hand. We have watched this happen enough times to build a studio around the alternative.
It is also why we don't compete on building automations. Agents are getting good at building those themselves; the floor rises every month. The part that does not get cheaper is knowing what to tell them. That is the part we sell.
Who is behind it.
Harsh Jaiswal
AI engineer. Builds and ships agents for a living, and started Tacit because the gap between what an agent can do and what it knows about any particular business kept turning out to be the whole problem.
How we'd like to be judged.
Not by a case-study page. By whether your team can change a rule in a skill six months from now without calling us, and whether the agent that follows it sounds like your company and not like a template. If it doesn't, we have not done the job.