About

Built because good questions kept becoming tickets

Every organisation running SAP has the same quiet cost: the answers are already in the system, and getting them out takes a specialist, a queue and a week. AnrilX exists to close that gap without loosening a single control.

What we are building

AnrilX is an agentic AI platform for SAP. You ask a question in plain language; it finds the data, computes the figure, shows you where it came from, and, when something needs to change, puts the exact change in front of a person before anything is sent.

The interesting engineering is not the conversation. It is everything underneath it: indexing an entire SAP OData surface into something searchable, compiling a question into a typed query that a model cannot corrupt, executing that query with no model in the path at all, and printing enough provenance that a controller can argue with the result.

How we work

We build against a real SAP system, we measure the things we claim, and we write down the measurements, including the ones that contradicted what we expected. Several of the core design decisions in this platform are the worse-scoring option on a narrow metric, kept because the better-scoring one failed on cases that mattered more.

We would rather tell a prospect what is not built than let them find out in a pilot.

What we believe

Four positions we build from

A number without its source is not an answer

It is an assertion. At scale, assertions are worse than a slow report; nobody can check them and everybody acts on them.

Structure beats instruction

Telling a model not to do something is a hope. Removing its ability to express the thing is a guarantee. We reach for the second every time it is available.

Measure it, then decide

Most of the decisions in this platform overturned the obvious version once we measured it. We write down what we measured, including the results that were inconvenient.

The failure path is the product

Anything can look good when it works. What separates a tool you can rely on is what it says when the system is unreachable, the read was incomplete, or the rule did not apply.

The company

AnrilX is a KO Innovation product, built by a small engineering team that works directly with the SAP teams deploying it. If you want to talk to the people who built it, you will be talking to the people who built it.

Working with us

What we will not do

Commitments about how we work. The product's own limits are on the platform and security pages.

We do not take a requirement and disappear for a quarter. You see it working in the first session, on a question you bring, or you find out in thirty minutes that it cannot be.
We do not put a salesperson between you and the people who built it. If you want to interrogate the architecture, you will be talking to whoever wrote it.
We do not ship a number we have not measured. Where a figure is approximate, the product says so in the answer rather than in a footnote.
We do not hold your data. AnrilX runs on-premise inside your landscape: your SAP data, your documents and your policy stay there.

Questions people ask us

Who is behind AnrilX?

A small engineering team working directly with the SAP teams deploying it, building AnrilX as a product of KO Innovation. There is no layer between the people who use it and the people who write it; a question about why something works the way it does reaches whoever made that decision.

What does working with you actually look like?

A first session against a live system with a question you bring. If that answers, a bounded pilot in your own landscape with your own data and your own reporting standards loaded. Deployment is a containerised stack you run; there is no hosted tenant to migrate onto.

What do you need from us to get started?

OData connectivity to your S/4HANA system and a person who knows which questions matter. If you want answers on your own reporting basis, the standards you already have written, as they are written, not rewritten into a rules format.

How do you decide what to build next?

From what gets asked and what fails. Most of the architecture in this product exists because something measurable went wrong first, and several core decisions are the worse-scoring option on a narrow metric, kept because the better-scoring one failed on cases that mattered more.

Can we talk to a reference customer?

Ask us. We would rather introduce you to somebody running it than hand you a quote on a web page.

Bring the question your reports cannot answer

Thirty minutes against a live SAP system we provide: no access to yours, nothing to set up. If it cannot answer, you find that out in half an hour rather than three months into a pilot.

Book a demo Talk to us