How is the “value to you” figured out — and how do I make it more accurate?

By default the value is the labor it saves: how many times the process runs a month × how long each run would take a person × a loaded hourly rate for your industry. Both of the first two come from YOUR real runs where we have them — how often it actually ran, and how many records a run actually touches — so the figure moves as the process does real work, and we never claim more runs than we have seen. You see that as “Value to You” + hours saved + an opportunity-cost line (what those hours free your team to do instead — e.g. working pipeline, filling roles, resolving accounts). Most processes are worth MORE than the hours, though — a renewal process protects revenue, a collections process avoids write-offs. So once a process is live we ASK you directly, right under its value card: what is it worth beyond the time it saves? The question names the hours we already count, so you don’t count them twice. You (or IT) give a rough annual figure and pick the kind (revenue it protects/accelerates, risk or cost it avoids, decisions it makes trustworthy) — or click “Not now” and we won’t ask again for that process. You can add or change it any time via “Add/Update the business value this delivers”. We credit a deliberately conservative share of that on top of the labor value, and you always keep the majority — you never see us take a cut, only your net value and the price. And you never have to take the number on trust: click “Value to You” on the process card and it opens its own working — every input, each line marked MEASURED (counted from your real runs), ESTIMATED, or “you told us”, adding up to exactly the figure shown. It also names what is missing: if we have not measured how many records a run touches, it says so and that the figure is therefore an UNDER-count.

What is “What it found” on a process — and how is it different from the value?

It is the one number on that card we MEASURED rather than worked out. Everything else there is labour: how often the process runs, how long each run would take a person, and a rate — plus whatever you told us the outcome is worth beyond the hours. “What it found” is different: it counts the records the process actually read, and how many of those turned out to be wrong. That is the work that WASN’T being done — records nobody knew were bad. It reads like “1 in 4 records was wrong · 448 checked · 112 corrected · 90 approved by a person and written back”. Every figure traces to a real run: “checked” is what the connector really returned, “corrected” is what the agent really changed, and “approved and written back” only counts changes a person approved AND that reached your system of record — an approval on its own is not proof the record changed. It appears once the process has surveyed a real segment on at least three production runs, and only when the process looked at more records than it changed (an agent that only reads the record it is about to fix would otherwise show a meaningless 100%). Until then it is simply absent — we would rather show nothing than a number we cannot stand behind. It never affects your price.

How do AI agents improve operational efficiency without adding risk?

Q: How do AI agents improve operational efficiency without adding risk?

By taking the routine, rules-bound work that runs across systems (counting, matching, routing, drafting, chasing) off people, while leaving every decision that changes a real record with a person who approves it in seconds instead of doing it in minutes. The efficiency comes from hours moved; the safety comes from the gate on writes, the tests before go-live and the audit trail after.

Where the hours actually go

Operations teams lose their days to work that is understood, written down and still done by hand: reading from one screen and typing into another, reconciling two lists, sending the same digest every morning. That work is repetitive enough to automate and consequential enough that nobody wanted to automate it blindly. A governed agent is what makes it safe to hand over.

tacitrun measures the value as time returned: every approved decision records the minutes a person did not spend, so the platform can show what a process is worth rather than count agent runs.

The controls that keep the risk out

Evaluation cases derived from the process before an agent is submitted; shadow runs on real work before it is promoted; IT approval as the go-live gate; every production write held for a person; figures the agent could not ground refused; and a trace of every run. None of these can be skipped by an agent that is confident, and each one leaves evidence a person can read.

The terms, as the product defines them

Value delivered

The work your agents handled this month, in money — after what you pay.

We price a process at a fraction of what it delivers, so value delivered is what you keep: the labour the agents took off your team (hours handled × a loaded rate for your industry) plus any outcome value you have affirmed, minus the price of the process. It is the same figure your billing page uses — Performance and billing read one source, so they cannot tell you two different things. Before an agent has enough live runs, the figure is an estimate from how the process is set up, and it is labelled as one.

Shadow mode

The domain agent runs alongside humans without acting, to prove itself.

In shadow, the domain agent proposes what it would do on real cases while a human makes the actual call. Tacit measures how often they agree. A domain agent must reach a high agreement bar (default 90%) before it can be promoted to act on its own.

Decision queue

Where domain agents send decisions for a human to approve.

When a domain agent isn't fully autonomous (or hits a high-risk action), it routes the decision to the queue. A human accepts, modifies, or rejects it. Those corrections feed back in as learning signals.

Questions people ask about this

How is the “value to you” figured out — and how do I make it more accurate?: By default the value is the labor it saves: how many times the process runs a month × how long each run would take a person × a loaded hourly rate for your industry. Both of the first two come from YOUR real runs where we have them — how often it actually ran, and how many records a run actually touches — so the figure moves as the process does real work, and we never claim more runs than we have seen. You see that as “Value to You” + hours saved + an opportunity-cost line (what those hours free your team to do instead — e.g. working pipeline, filling roles, resolving accounts). Most processes are worth MORE than the hours, though — a renewal process protects revenue, a collections process avoids write-offs. So once a process is live we ASK you directly, right under its value card: what is it worth beyond the time it saves? The question names the hours we already count, so you don’t count them twice. You (or IT) give a rough annual figure and pick the kind (revenue it protects/accelerates, risk or cost it avoids, decisions it makes trustworthy) — or click “Not now” and we won’t ask again for that process. You can add or change it any time via “Add/Update the business value this delivers”. We credit a deliberately conservative share of that on top of the labor value, and you always keep the majority — you never see us take a cut, only your net value and the price. And you never have to take the number on trust: click “Value to You” on the process card and it opens its own working — every input, each line marked MEASURED (counted from your real runs), ESTIMATED, or “you told us”, adding up to exactly the figure shown. It also names what is missing: if we have not measured how many records a run touches, it says so and that the figure is therefore an UNDER-count.
What is “What it found” on a process — and how is it different from the value?: It is the one number on that card we MEASURED rather than worked out. Everything else there is labour: how often the process runs, how long each run would take a person, and a rate — plus whatever you told us the outcome is worth beyond the hours. “What it found” is different: it counts the records the process actually read, and how many of those turned out to be wrong. That is the work that WASN’T being done — records nobody knew were bad. It reads like “1 in 4 records was wrong · 448 checked · 112 corrected · 90 approved by a person and written back”. Every figure traces to a real run: “checked” is what the connector really returned, “corrected” is what the agent really changed, and “approved and written back” only counts changes a person approved AND that reached your system of record — an approval on its own is not proof the record changed. It appears once the process has surveyed a real segment on at least three production runs, and only when the process looked at more records than it changed (an agent that only reads the record it is about to fix would otherwise show a meaningless 100%). Until then it is simply absent — we would rather show nothing than a number we cannot stand behind. It never affects your price.

How do AI agents improve operational efficiency without adding risk?

Where the hours actually go

The controls that keep the risk out

The terms, as the product defines them

Questions people ask about this

Related