A company runs on context.
Its software throws it away.
Every tool a company buys holds a piece of what the company knows, in a shape no other tool can read.
Kaso is the argument that those pieces belong in one system. Sales, marketing, operations, development, research and training in one stack, with shared context, coordinated agents and usage-based pricing. Use one product, or connect them across your company.
Every tool was bought
to solve one problem well.
None of them was bought to remember anything for the others. That is not a failure of any one of them. It is what happens when the unit of purchase is a feature and the unit of work is a company.
A record that lives in two places is read from one.
Two lists of the same thing drift, and nothing reports the drift: an account marked closed-won in one system and still being sequenced in another, both correct by their own records. Wherever a fact has more than one home, the screen built to read it reports a fraction of what happened and looks right doing it. The fix is never a reconciliation job. It is one list, read by both.
The seams are where the money goes.
Four vendors can answer the same question about the same person, at four prices, with four ideas of what an answer is. Somebody has to decide the order, check for free before spending, cache the result, and record every attempt including the misses. Do that in one place and the spend is attributable to the thing it bought. Do it in six, and the bill arrives with no way to read it.
Work that crosses a boundary stops at it.
A research finding informs a campaign. A customer conversation shapes a feature. An internal tool improves the next handoff. Every one of those sentences describes context leaving one system and arriving in another by hand — retyped, re-explained, or lost between the two.
A number a person cannot interrogate is a number they will ignore.The rule every score in Kaso is built to
The context stops being
something you carry.
Put the work and the record in one system and a set of problems stop being anyone's job. The account history is already beside the call. The call is already beside the follow-up. Nobody moves it.
What sits on top of that record is not one assistant. It is a set of single-purpose agents, because the work is not one shape: scoring has to be cheap across thousands of rows, drafting has to be reviewable one message at a time, and the agent that answers questions needs tools the others must not be given.
Each inherits the permissions of the person who triggered it and cannot widen its own scope, because the workspace is not something the model gets to choose. A workspace's records are reachable only from inside it, enforced in the database rather than in the code that queries it, so it holds even where somebody forgets the filter.
Text arriving from outside the workspace is treated as data, and anything in it that reads as an instruction is flagged rather than followed. A meeting fact carries the sentence it came from — the line, the speaker, the quote — and a claim no sentence supports is left empty, because that absence is the finding. None of it is a feature list. It is what one record makes cheap and six make expensive.
- 01An agent inherits your permissions, never more
- 02Every answer opens onto the records behind it
- 03“Not found” is an answer, not a gap to fill
- 04Writes wait at the review points you set
- 05A count that is a floor says that it is a floor
- 06Every run can be opened, read and run again
A seat prices access.
The work is what costs.
Usage-based pricing is not a packaging decision here. It falls out of the first two sections: when the system does the work, the work has a price the moment it runs.
A lookup costs money. A model call costs money. An hour of meeting assistance costs money. A licence that ignores that has to guess, and it guesses in one of two directions: it overcharges the team that barely uses it, or it quietly rations the team that would use it most. Charging for work means recording the work. Every paid lookup is written against the run that asked for it and the person who asked. Money is held in whole minor units, never a float, because a rounding error near a ledger is a forecast defect waiting to happen. Metering runs whether or not the answer was any good — the provider charged either way, and the run that matters most to see is the one that went wrong.
That record does two jobs, and the second is the one that matters to an operator: the same ledger that bills you is the ledger that stops a runaway. A workflow's ceiling is set outside the workflow, and the workflow cannot raise it.
What this page does not carry is a rate. Rates are quoted before anyone adopts anything, and an illustrative bill is not a price list.
The list of things
we will not do.
A claim is only worth reading if something could have gone in its place and did not. This is that list, and it is meant to be checked.
Sales and marketing first.
Then the rest of the company.
Kaso starts with sales and marketing and expands into research, training, operations and development. Context that travels between them is what the whole thing is pointed at.
The end state is unglamorous, and that is the point: one place the company's work happens, one record of what happened, and a bill you can read.
See what your team
can do with Kaso.
Bring your workflows. We’ll show you where to start.
Request accessOne product or a connected platform. Built around you.