Back to logic overview

01 / Fact derivation

Facts become
new conclusions.

Turn separate facts into connected, checkable reasoning. See calculations, conditions and multi-step dependencies produce new conclusions.

Start with this example
START WITH THE FIGURESRevenue and costs
WORK IT THROUGHHow much profit remains?
SEE THE CHANGEMore revenue, less profit
1Revenue grows2Profit falls3Follow the derivation

Your AI can show how it knows.

EXAMPLE 01 / AI ANALYTICS

Growth can hide a different story.

Revenue rises 25%. Combine the revenue and cost records to discover what happened to profit and margin.

Change the current cost. Watch the same evidence structure support a different conclusion.

Example · follow the evidence and reasoning
What the records tell us

Same company · Q4 · USD

PeriodRevenueCost
Q4 2024$1,000,000$600,000
Q4 2025$1,250,000$875,000

Same company, quarter and currency. Source records are available below.

Current cost scenario

Current profit: $375,000 · margin: 30%.

01

Profit

Subtract cost from revenue for each period.

$1,250,000 − $875,000 = $375,000

Previous period$400,000
→
Current period$375,000
02

Profit margin

Divide profit by revenue to compare how much each dollar earns.

$375,000 ÷ $1,250,000 = 30%

Previous period40%
→
Current period30%

The result

Revenue rose. Profit fell.

Revenue grew by $250,000, while cost grew by $275,000. Profit fell $25,000 and margin fell 10 percentage points.

Profit change−6.25%
Inspect the evidence+
S1

Revenue ledger · rows 12–13

Q4 2024 | revenue_usd = 1,000,000
Q4 2025 | revenue_usd = 1,250,000
S2

Cost ledger · rows 08–09

Q4 2024 | cost_usd = 600,000
Q4 2025 | cost_usd = 875,000
S3

Reporting scope · paragraph 2

Both periods cover the same company and the same quarter. Revenue and cost are measured in USD.
See the decision conditions+

profit(period) = revenue(period) − cost(period) margin(period) = profit(period) ÷ revenue(period) profit_fell ← current_profit < previous_profit

A useful answer connects the underlying records, derives the intermediate facts, and keeps the entire path available for inspection.

LOGIC ACROSS AI WORKFLOWS

Different questions. Explicit logic.

The same idea applies to an agent action or a chain of dependent services.

EXAMPLE 02 / AGENT ACTION

Can the agent share this report?

All three conditions must be satisfied before sharing.

✓Owner permission
○Destination approvalNot approved
✓No restricted content
One condition is not satisfied.

Let an agent explain the conditions behind a proposed action.

See the decision conditions+

Sharing is allowed when the owner has authorized it, the recipient has approved it, and the content has no restrictions.

EXAMPLE 03 / MULTI-STEP REASONING

Which services are affected?

Key store unavailable
  1. Identity serviceNeeds the key storeAffected
  2. CheckoutNeeds the identity serviceAffected
  3. Public status pageUses a separate static cacheNot affected

Follow a dependency through several steps to surface an impact that no single record states.

See the decision conditions+

When one service is disrupted, check the other services that rely on it.

EVIDENCE BEFORE ANSWERS

Same information. A decision you can trust.

One overlooked detail can turn an approval into a mistake. Change the situation below to see how CitProof catches missing information, return restrictions and changed records.

Example · AI refund assistant

The order information

Purchased
12 days ago
Package
Unopened
Return restriction
Not stated in the record

Refund policy: within 30 days, unopened, and eligible for returns.

A direct answer

Example response
“It has only been 12 days and the item is unopened. Approve the refund.”

It approves before checking whether returns are permitted.

The answer mentions only

Purchased · 12 days agoPackage · Unopened

Checked with CitProof

  • Within the window

    Day 12 is inside the 30-day limit

  • Packaging is eligible

    The order confirms the item is unopened

  • Return eligibility still needs checking

    This missing detail can change the decision

Premature approval caught

Confirm this detail first

Only timing and packaging are confirmed. CitProof identifies what was overlooked and what to check next.

Next: check whether this order allows returns
01

Show the evidence

Connect each important decision to the source information behind it.

02

Explain the reason

Make the conditions visible so users can follow the decision.

03

Keep answers current

Identify answers that need another look when records or policies change.

Give users a reason to trust your AI

Give users more than an answer: the facts it rests on, the restrictions that could change it, and anything still waiting to be confirmed.

A conclusion with its workings.

Evidence → Reasoning → Reusable knowledge

01 →

Source-backed facts

Preserve the passage, value and scope behind each premise.

02 →

Explicit derivation

Bring the source facts and business conditions together, then explain each step leading to the answer.

03 ✓

Check the answer again

Revisit the evidence and replay the relevant derivation with an independent checker.

CITPROOF

Build reasoning into your AI.

Bring us a workflow where answers, actions or discoveries need a stronger basis. Explore how evidence chains and logic verification can fit your product.

Talk to the team