Kaddiya

EVERYONE ELSE SELLS AUTONOMY

"Digital labor" that acts without waiting for human confirmation. "Self-executing workflows." "The first autonomous software engineer."

Their pitch is how much happens without you. Gartner expects over 40% of these projects to be cancelled by end of 2027 — cost, unclear value, inadequate control.

KADDIYA IS BUILT THE OTHER WAY ROUND

There is no autonomous write path in the build. It drafts a reply; a human presses Send. It proposes a change; a human presses Create. Every write is a click on the exact payload, running on that human's own credentials.

Not a policy. Not a setting someone can flip. A property of the build — and one your instance can enforce at its own edge as well.

ServiceNow only. That's the point, not the gap.

The generalists connect to everything and know none of it deeply. They can't tell you what an update set captures, why an ACL denied a row, or where sc_req_item sits in the task hierarchy — those things exist in one platform. Kaddiya is built by ServiceNow specialists, for ServiceNow, and nothing else. Depth in one platform beats breadth across twenty.

FOR EVERY SEAT ON THE INSTANCE

One agent. More capacity at every level.

SERVICE DESK

The queue becomes a review pass

Ask what's on your plate. Kaddiya pulls each case, reads the whole journal, finds the last time someone fixed the same thing, and drafts the reply — Claude does the reading, you keep the judgment and press Send.

my queue · prior fixes · drafts behind a button

DEVELOPERS

A senior colleague who read all the docs

Schema truth instead of guessing, encoded queries you can lift straight into your own work, the update set checked before anything is proposed, and complete scripts written for your review — grounded in the documentation for the release you actually run.

sn_schema · update sets · full scripts · docs cited

LEADS & ADMINS

Answers about the instance, not anecdotes

Breaches by group, ageing P2s, who holds which role, why an ACL denied the row. Questions that used to be a report request become a sentence — with every query it ran on screen to check.

every query shown · every commit named

What each seat actually gets

HOW IT WORKS

Illustration: three colleagues around a laptop, beside a stack of four steps — Analyze, done; Draft, done; Propose, in progress; You approve, still open, labelled send or create.

01

It signs in as you

OAuth on your own instance. Your roles, your groups, ServiceNow's own ACLs decide what it sees. We wrote no authorization logic.

scope: read · identity: yours

02

It drafts, never sends

Replies, update sets and configuration changes arrive as proposals behind a button. There is no code path that writes without that button.

writes: human click only

03

You approve, it's logged

You press Send or Create. The commit runs on your credentials and is logged against you — the person who approved it.

audit: every action, named

WHAT IT DOESN'T DO

It doesn't work while you're asleep, and that's deliberate.

There are no background jobs, no scheduled runs, and nothing sitting on a queue overnight — it works while a human is watching it work.

Unattended operation would mean holding a durable token to your production instance long after you'd closed the tab. We didn't want to hold one, so we didn't build the thing that needs it.

What we holdWHILE YOU'RE SIGNED IN

Your tokensession only

After sign-outdiscarded

Background jobsnone

Offline agentsnone

Your model keyyours, not ours

The anatomy of one turn, the full tool catalog, and why this one is a property of the build

SHOW THE RECEIPTS

Transparency you can read, not a claim you have to trust.

A REAL TOOL CARD

QUERY · incident 0.38 s · 14 records

assignment_group.name=Network Ops^active=true^ORDERBYDESCpriority

QUERY · sys_user_grmember running · 0.3 s

group.name=Network Ops^user.active=true

ServiceNow's own encoded-query syntax, on screen while it runs. If you can read your instance, you can read what the agent asked it.

A REAL AUDIT LINE

comments · INC0048291 · 2026-08-30 14:32:08 · approved and sent by dana.whitfield

A REAL COST FIGURE

≈ $0.011

one answered question, reply drafted

claude-sonnet-5 · in 1,842 · out 486 · cached 12,288

Every receipt, taken apart

IT KNOWS YOUR INSTANCE

Docs for your release. Memory of your quirks.

Documentation pinned to your release

Kaddiya reads ServiceNow's official docs for the release family your instance is running — supported for the current family and the two before it — and cites the topic it used. Platform questions get the answer for your version, not a recollection of someone else's.

release: detected from the instance · answers cite the topic

A notebook of your instance's gotchas

The first time it learns that a state was renamed, a table hides behind an ACL, or VPN cases route to a special group, it writes the note down for a person to keep or discard. Kept notes start every later conversation. The notebook lives in Kaddiya — writing to ServiceNow still requires you.

kept by a human · local to Kaddiya · never written to the instance

WHAT IT COSTS

Platform AI is metered in credits with no published price. Here is a number in cents.

01Tokens and dollars shown under every answer, per turn, while you work — not in a quarterly true-up.

02Your own Anthropic key, billed to you at Anthropic's published rates — or your own gateway, if your platform team would rather the traffic left through theirs. What connects today.

03Your platform vendor can't reprice it at renewal, because they don't sell it.

FOR THE SECURITY TEAM

You've stopped AI pilots before. Here's what's different, stated plainly.

No autonomous write path

Anything you can do on your instance, it can draft. Nothing it can do without you. The agent can propose a write; it cannot perform one. Every commit happens on a signed-in human's click, on that human's credentials, through a named endpoint, from a list that fits on one page. Your instance can enforce the same limit at its own edge.

ServiceNow's ACLs enforce access

OAuth as the signed-in user. It sees exactly what that user sees — no service account, no elevated role, no authorization logic of ours to audit.

Full audit trail

Every query is shown on screen as it runs and retained. Every commit is logged against the person who approved it, in the instance's own audit tables.

Your key, or your gateway

The console speaks one protocol — the Anthropic Messages API — and calls the endpoint you point it at: Anthropic directly on your own key, or your own gateway, so instance data leaves through your DLP and logging and we hold no model credential at all — except on the hosted free plan, whose trial allowance runs on a labelled key we hold and meter until you connect your own. The shapes that connect today, including the ones that don't yet.

THE QUESTIONS THAT ALWAYS COME

How do we stop it, immediately, without you?

Set the OAuth record inactive. Every token dies with it, from your side, without asking us or waiting for us.

Where does our instance data go?

To the model endpoint you configured, and nowhere else. Point it at your own gateway and it leaves through your DLP and logging. We hold no model credential, apart from the metered trial key the hosted free plan runs on.

Are you SOC 2 certified?

Not yet, and we won't imply otherwise. in progress — ask us for current status rather than trusting a badge on a website.

The blast radius, the risks we'd flag ourselves, and every questionnaire answer

WHAT SETUP ACTUALLY IS

Nothing is installed on your instance.

No scoped app, no update set, no widget, no stored script, no service account. One OAuth client record your admin creates — the same mechanism ServiceNow's own mobile apps sign in through — and the console runs on your side of the wire.

See exactly what moves where