AI consulting for enterprises

We work out what can actually be solved with AI in your company, with which technology, what it costs to build and what it costs to keep running every month. In two weeks, in writing.

2,000,000 active users served · 10 countries in operation

Built on

An engineering team that has already been on the other side.

Ninety seconds: who we are, how we work and what you get at the end.

Two weeks, and a decision you can defend

Day 0

Fifteen-minute call

What has to be decided and what information exists today. If the problem does not need a diagnostic, we say so on that call and nothing is billed.

Days 1 to 3

Sessions with the people who execute

Two or three sessions with the people who run the process every day, not with whoever documented it. That is where the real process comes out, with its exceptions and its shortcuts.

Week 1

Inventory and baseline

We count documents, sources and query volume. The evaluation set is built and the starting point measured, before anything is touched.

Week 2

Document and working session

The decisions written down with their rationale, and two hours to go through them with the technical team and with whoever signs the budget.

Day 0

Formats

Four ways in, depending on where the company stands

All of them end in a document with decisions. What changes is where you start and how long it takes.

01 2 weeks

Opportunity assessment

For when it is still not clear what to do with AI. Out come the cases worth doing, in what order and with how much return.

Interviews with whoever runs the process
Documents, sources and volume counted
Cases ranked by return and effort

Quoting
Tier-1 support
Reconciliation
Board report

Candidates by return

02 3–5 days

Architecture review

You have a design or an implementation and want a second opinion before committing quarters of work.

The risks in the current design
The ruled-out routes, with their cost
What to require from the team building it

Documents → Fine-tune → Own model$$$ · 6 months
Documents → Index → RAG$ · 6 weeks

Two routes, one case

03 1 week

Diagnostic of an existing system

For when something is already in production and answering badly. Measured against real questions, with the failures ranked.

Reference set with real questions
Quality measured against that reference
Failures ranked by impact, with their fix

Fails to retrieve41%
Answers with no source27%
Cites an expired version19%
Makes the answer up13%

Failures by frequency

04 Minimum 3 months

Ongoing advisory

For when the internal team builds. We review architecture, evaluation and cost on a fixed cadence.

Architecture review every three weeks
The evaluation set kept current
Cost per query watched month to month

Arch.
Eval.
Cost
Wrap

Twelve weeks

Before building

Six decisions closed before anything is built

Each one is settled with your documents and your numbers, and written down with its rationale.

Whether AI is worth it here

We look at the process and the real query volume. When most of it comes out of a query against the system you already have, we say so: cheaper, and no monthly bill.
What gets built

RAG, fine-tuning, prompting or an agent. The cheap option and the expensive one are ten times apart, and we pick with your documents and your volume on the table.
Whether your information works as it is

We count the documents, how many carry a valid-from date and how many are duplicates. If it has to be cleaned up first, we work out how much that lowers the monthly cost.
What it costs to run

The monthly bill at your volume, not the build. It is worked out per query and per model, and written down before you sign.
What it is measured with

We build the evaluation set from your own questions and their correct answers. It is what tells you a change improved the system, and you keep it even if you change supplier.
What happens the day it fails

What the user sees, which threshold fires the alert and who receives it. Defined now, not after the first complaint.

This is where the data comes in

An AI system is worth what its sources are worth. These are the standard connectors; anything with an API or a database connects the same way, and what has no API is handled by file.

SAP

Enterprise ERP

ERP

Oracle

ERP and database

ERP

NetSuite

Cloud ERP

ERP

Salesforce

CRM and service

CRM

HubSpot

CRM and marketing

CRM

PostgreSQL

Database and pgvector

Databases

Fourteen sectors, and in each one the measurement changes

Insurance

Policy wordings, exclusions, claims

Says what a policy covers, quoting the clause and the version in force that day. If the wording changed in March, it answers with March’s.

AI for Insurance →

An example

— Am I covered if my car is stolen from a public car park?

Yes, with a 10% deductible. 2026 policy wording, clause 4.3.

A RAG system in production: three years, two million users

We built and operated the assistant for a loyalty platform serving more than two million active users across ten countries.

We built the full pipeline: document ingestion and normalization, chunking, embedding generation, vector store on Azure AI Search and Pinecone, and retrieval with grounded generation on LangChain.

We held it above 99% availability for three years.

RAG systems →

Frequently asked questions

What exactly do you deliver?

A document with decisions, not trends: use cases prioritized by return and effort, the recommended architecture with its justification, development and monthly operating cost estimates, the metrics that will define success, and the risks with their mitigation. Plus a working session to discuss it with your team.

Is this useful if we do not yet know what we want to do with AI?

That is exactly what it is for. The opportunity assessment starts from your processes and your information, not from a catalog of technologies. It produces three or four concrete candidates, ranked by estimated return and effort, with a recommendation on where to start.

What if the conclusion is that we do not need AI?

We tell you, and it happens more often than you would expect. Many problems framed as "we need AI" are better solved with a well-built query, an integration or a redesigned form.

Do you need access to our systems?

For the opportunity assessment, no: interviews with the people who run the processes and a sample of the information are enough. For the diagnostic of an existing system we do need to see the data it was fed and a sample of real conversations or runs. An NDA is signed first, and we work on anonymized samples when personal data is involved.

How long does it take?

The architecture review, three to five days. The diagnostic of an existing system, one week. The opportunity assessment, two weeks. None of the three is a long process: the goal is for you to decide quickly, not to accumulate documentation.

Can you work alongside our current vendor?

Yes. In many cases the role is to review and validate what another team is building, with concrete technical criteria rather than opinions. We deliver findings ranked by impact so you can hold them accountable, and we can verify afterwards that they were implemented.

Do you work with internal teams?

Yes, through the ongoing advisory format: your team builds and we review architecture, evaluation, cost and technical decisions on a fixed cadence. It is the format that leaves the most capability behind, and the one we recommend when you already have technical people.

Fifteen minutes. A concrete answer.

It gets fixed, it gets built, or it is not worth it. And if the diagnostic does not reach three actionable findings, it is not charged.