Show us a process where humans are doing work software should be doing

Is your AI pilot ready to go into production?

Use this self-assessment to see how ready your system is to operate inside the company.

The guide also includes eight real systems already operating in companies in Mexico and Chile, with their metrics, controls, and production status.

  • 12 questions to evaluate a pilot or a vendor's work.
  • 8 real systems in operation, with their industry, scale, and results.
  • Our process for going from an initial test to a system running with real users.

Free.

Get the guide

PDF language

Verification

We use your email to send the guide and, at most, two messages afterwards. You can unsubscribe from any of them. We do not share your data.

Guide cover
Self-assessment, page 2

What's inside

A 12-question self-assessment

The questions cover four areas: results, control, owners, and operations. When you finish you will have a quick read of what is already in place and what is still missing before you expand the system.

Eight systems that already operate

We include cases from marketplaces, retail, higher education, car rental, visa services, and live entertainment.

Some results from these projects:

  • 190,325 items processed in one season, at 97.9% success.
  • More than 20,000 students using AI-generated content inside their LMS.
  • A public MCP server used from 17 countries.

The cases are real Kemeny Studio projects. See the cases.

How we go from pilot to production

We show the nine stages we use on our projects, from understanding the current process to leaving the system running and measured. Each stage has a concrete condition for moving on. We also include the working cadence and the people who typically take part.

What we review technically

Integrations, data access, identity, permissions, models, evaluations, observability, human intervention, and operations. We review what the system needs to run inside the company in a stable way, with clear owners.

Three of the twelve questions

  1. 01

    Does the AI reach data through tools with a defined scope, or with direct credentials to the database?

  2. 02

    Is someone on the business side responsible for the system day to day, or only IT and the vendor?

  3. 03

    Do you have a measurement of the process from before the AI?

In these projects

8

systems already running, or that already ran

6

enterprise customers

14+

months on the longest project still running

2

markets: Mexico and Chile

Who it is for

Business and technology leaders who have already tried AI and now have to leave it running. Operations managers with a pilot that never quite enters production. IT people who have to approve a vendor. Directors who will decide the next budget. The guide assumes you have already used generative AI.

Who wrote it

Francisco Kemeny

Francisco Kemeny, CEO of Kemeny Studio

AI-native product and engineering studio in Mexico City and Santiago. We build AI systems for business processes and run them in production. OpenAI Select Partner. NVIDIA Inception. Microsoft for Startups. Start-Up Chile Growth.

We're now an OpenAI Select Partner

Is this sales material?

It is a description of how we evaluate and take AI systems into production, with eight real projects. You can use it to review an internal project, evaluate a vendor, or understand how we work at Kemeny Studio.

Download the guide and do the self-assessment

It takes about ten minutes. When you finish, the score shows what is already in place and what is still missing. Below 9, there is work before the system can run more widely. From 9 up, it can already operate, and the next step is choosing which process to take further.

If you already know the process, book a 30-minute review, or start with the Fit Check.