Skip to content
XEVRION
Menu
Improve

Which rules are too strict, or too loose?

Let agents earn trust from real decisions. Test a new rule on your own history before it goes live.

Improve

Agent trust scores

A 0 to 100 score and a four-step ladder, built from what people actually did with its proposals.

  • Earned, not guessedScores come from real approvals and rejections, with the maths to show how sure we are.
  • Four stepsFrom every action reviewed, up to trusted for routine work.
  • Per agent and per actionAn agent can be trusted for refunds and not for deletions.
How Agent trust scores works →
app.xevrion.co.uk / agents / refund-agentExample data
Trust score78of 100
Approved as proposed91%Last 200 requests
Edited6%Changed before approval
Rejected3%Turned down

Four steps

People decide each move
  1. Always asks a person
  2. Asks one person
  3. Rule passes low risk
  4. Trusted for this action

By action

  • refund.issue82
  • ticket.close64
  • fee.waive31

Improve

Policy what-if

Replay past requests against a new policy and see reviews saved, hours saved, and what it would have missed.

  • Your data, not a guessThe replay uses the requests you actually had.
  • Priced in plain termsReviews saved, hours, cost, and the cases a looser rule would have wrongly let through.
  • Saves nothingIt returns a draft policy. You decide whether to use it.
How Policy what-if works →
app.xevrion.co.uk / policies / refunds / what-ifExample data
Reviews saved312Of 1,480 last month
Hours saved26At 5 min each
Would have missed0Wrongly passed

Reviews per week

Now and draft v13
W1W2W3W4v13

Draft v13

Change
Auto-approve up to £75
Replayed
1,480 past requests
Result
Safe to publish

Improve

Counterfactual replay

Re-run any past request under another policy version and get a signed answer, next to what people actually decided.

  • One request, two rulesSee the old route and the new route side by side.
  • Signed answerKeep the result as evidence of what the rule would have done.
  • Compared to peopleShows how the humans really decided.
How Counterfactual replay works →
app.xevrion.co.uk / decided / r_01HX9 / replayExample data

Under refunds v12

What happened
Rule
Over £500
Route
Two reviewers
People decided
Approved

Under refunds v13

What would happen
Rule
Over £750
Route
One reviewer
Signed answer
Valid

Improve

Compliance evidence packs

See how your review practice lines up with common frameworks, with a ranked list of what to fix first.

  • Computed, not typedFour-eyes on high risk, independence, timeliness, SSO, log export, signing and more, from real activity.
  • Mapped to frameworksEU AI Act, NIST AI RMF, ISO/IEC 42001, SOC 2 and GDPR.
  • Signed and printableA signed JSON pack and a report you can print.
How Compliance evidence packs works →
app.xevrion.co.uk / proof / complianceExample data

Your practice, by framework

Computed
RowEU AI ActNIST AI RMFUK GDPR
Human oversight
Record keeping
Incident response
Change control

Fix first

Ranked
  • 6 actions had no checkHigh
  • 2 agents have no ownerMedium
  • Policy tests not run on releaseLow

Improve

Insurer pack

Plain signals an insurer asks about, with the figures behind each one, signed and sealed to the recipient you choose.

  • What they askHuman review of high-risk work, reviewer independence, limits, two-person control, verified trails.
  • Numbers behind eachEvery signal shows its evidence.
  • Sealed to one readerOptionally lock it to the recipient's own key.
How Insurer pack works →
app.xevrion.co.uk / proof / insurerExample data

What they ask

With the figures
  • Human review of high-risk actions100% of 1,204Yes
  • Spend limits per agent6 of 6 agentsYes
  • Tamper-evident recordsSigned log, 48,211Yes
  • Tested before releaseLast run 26 OctYes

Sealed to

Reader
An insurer you choose
Opens with
Their key only
Signed
Valid
Expires
In 30 days

Improve

Anonymous benchmarks

Opt in to compare against other organisations. Only totals are shared, and only for groups of five or more.

  • Opt inOff until you switch it on.
  • Totals onlyNo request content ever leaves your organisation.
  • Groups of five or moreA comparison appears only when the group is big enough to stay anonymous.
How Anonymous benchmarks works →
app.xevrion.co.uk / settings / benchmarksExample data

Median answer time

You and your peer group
YouP25MedianP75

What is shared

Opted in
Yes
Shared
Totals only
Peer group
11 organisations
Smallest group
5

Know what your agents did.

Start free. Guard your first agent in a few minutes, and keep a record anyone can check.

  • Free plan, no card
  • Signed, chained decisions
  • Checked offline, without us