Back to the cases

Training

Twelve weeks to decide on artificial intelligence with judgment of their own

A company that resells software and network equipment to industrial clients in northwestern Mexico wanted to offer artificial intelligence, and first needed to understand it well enough to decide with judgment of its own. We gave it a twelve-week workshop, with people from almost every area, that ended with prototypes running on its own data.

The goal

To move artificial intelligence from the ground of technology to the ground of business value, and come out knowing how to decide what matters: public or private models, on their own machines or in the cloud, fixed workflows or agents that make decisions.

How it worked before

• They already paid for automation tools included in their office licenses, without making use of them.
• They had little to offer their clients in artificial intelligence, and technology decisions followed whatever the vendor of the month proposed.

What we did

People from sales, marketing, development and project management came to the workshop, on purpose: artificial intelligence is decided better when those who sell, those who communicate, those who build and those who deliver understand it the same way. It was four modules of three weeks each, with one in-person session per week and remote follow-up, between November 2025 and February 2026:

• Infrastructure and fundamentals: what runs where and what it costs, on their own machines or in the cloud.
• Tools and work environment: language models, databases that search by meaning, workflow orchestration, and how to give instructions to a model.
• Building prototypes: a workflow that takes the electronic invoices arriving by email, reads their tax data and files them on its own; a database that links each quote with its purchase and sales invoices and calculates its margin; and an agent that answers questions about invoices, quotes and margins by email.
• Challenges and business cases: connecting the automation platform with the office tools they already paid for, and a comparison of three paths (all with their license vendor, all with open tools, or a mix) with its decision matrix.

The invoice workflow was built in three versions, one of them almost entirely visual, so that someone who does not program could maintain it. And the last challenge was left on purpose for them to solve on their own.

The impact, measured with CREO

I measure the impact of every project with four questions, which I call CREO: did the company become more Competitive? Did it improve its Return, meaning its profitability? Is it more Efficient? Does it meet its Obligations better, the tax, legal and contractual ones? Not every project hits all four, and here are the ones that apply.

Competitiveness

They came out with something to offer their clients, and out of that came a partnership to resell an electronic invoice lookup service.

Return

The comparison of the three paths let them see what each would cost per month before committing, starting with making use of what they already paid for.

Efficiency

They saw working, with their own invoices and quotes, what they used to do by hand: filing receipts and calculating the margin on each sale.

Obligations

The prototypes read the current version of Mexico’s electronic invoice, by type of document, which is the foundation of any automated tax control.

The result

• The twelve weeks were delivered in full and paid in full.
• The relationship continued after the workshop: the client went from student to partner, with joint projects for other clients.

More with less

Artificial intelligence is learned better with your own data than with catalog examples. With their invoices and quotes as raw material, each area came out with judgment to decide what to buy, what to build and what to leave as it is, which is what keeps money from going to the wrong tool.