How I decide
The money first, the tool second
Everybody is selling artificial intelligence. I start somewhere else: where you are losing money, which process is holding back your growth, and which decisions you are making on incomplete information. Up to that point, zero technology.
After that, every piece of the work gets sorted into four, and that sorting decides what it gets built with:
- With a person
- It gets fixed by changing who does what, or in what order. Zero technology, and it is where I start.
- With explicit rules
- Written once, in deterministic code, meaning the same inputs always produce the same result and it can be audited line by line. Anything that touches money lives here.
- With machine learning
- Once there is enough history for a prediction to beat a rule written by hand.
- With generative artificial intelligence
- Reserved for what has no structure: free text, a phone call, or reading a truck odometer off a photo.
The hard rule across all my projects: the system flags, and the final call is always
a person's. Knowing what to keep out of artificial intelligence is half the work.
Who you would be working with
28 years moving the systems of large companies
- Among the first
- nationwide SAP implementations carried out in Mexico, delivered on time and on budget.
- Two failed
- attempts by outside consultancies were behind the SAP implementation at Comercial de Carnes Frías del Norte, the Baja California supermarket and wholesale distribution chain, when I came in. The third one landed with the in-house technology team that was already there, which I had the privilege of leading: what got reorganized was the work.
- 15 years
- from 2005 to 2020, as Chief Information Officer of Grupo Hermosillo, one of the strongest construction groups in Mexico, building since 1963: sixteen companies across four markets, and a team of nineteen people I built from scratch.
- 900 million
- dollars in operations, and that team carried all of its technology on a budget under two million.
- 40%
- saved by migrating off our own data center and into the cloud: the infrastructure went from being sized for the yearly peak, and idle the rest of the time, to following demand. That cost plan carried the group through the 2008 and 2009 recession.
- Six months
- is what it took me to go from CIO to CEO of Opulence Group, in Dubai, and partner at three. The group then included Ali Zaid Al Quraishi & Brothers, the Saudi family group of more than two hundred companies, and Unimech, in Dhaka.
- Two passports
- and two native languages, Spanish and English. I am a US citizen and a Mexican citizen, I live on the border, and I have worked on both sides of it and well beyond: in Saudi Arabia I was received better as a US citizen, and in the Emirates better as a Mexican.
- CETYS
- BS in Computer Science, Mexicali campus, 1992 to 1996.
Built and running
What I am building right now
Each one replaced something that used to be done by hand. I describe clients by their
industry, because putting their name on this is their call.
A system of their own, instead of one that costs millions
Industrial construction · in progress
The usual alternatives: the system you buy costs millions, plus twelve months of work just to stand up the base version, and the one you rent costs tens of thousands a year and never quite becomes yours.
We are building theirs in phases, and each phase delivers value before the next one starts: every project's balance in real time, cost estimates and contracts without double entry, the vendor portal, and an engine that produces new budgets by learning from the real costs of their earlier jobs, which you brief by describing the project out loud.
Running on top of that are the specialists wired into their database, the data loads that feed themselves, a dashboard you can talk to, and the weekly scheduling that used to be assembled by hand.
A quote that gets built by talking
Cross-border logistics, Mexico and the United States
Before: every quote was put together by hand, driver assignment lived on a whiteboard, and permits and risks lived in the head of whoever was on shift.
Today you describe the trip to an assistant, which classifies the type of service and fills in the parameters on its own, and their people fine-tune it by hand where needed. Around that, the whole operation online and in two languages: dashboard, assignments, permits, distance-band rates, and a living register of twenty-seven risks.
Process redesign and automation, pushing together
Freight transport · in progress
An outside firm is redesigning their processes while we build the system that holds them up. Redesigning without building leaves the manual in a drawer, and building without redesigning automates the mess.
Before a line gets written, every piece is sorted: what belongs to a person, what is explicit rules, what is machine learning, and what is generative artificial intelligence. Anything that touches money goes into deterministic code, and the AI is reserved for what has no structure, like reading a unit's odometer off a photo. The hard rule of the project: the system flags, and the final call is always a person's.
Their own development team, able to build it without me
Hauling new trucks across Mexico and the United States
This client moves the manufacturers' new trucks across Mexico and the United States, and is among the largest haulers of its kind in North America. It also owns a development company, with its own programmers, so build capacity was the one thing it was not short of. What was missing was deciding which part of all this artificial intelligence would actually serve the operation, instead of buying whatever the vendor of the month was advertising.
I started by running a workshop on automation and artificial intelligence for those programmers, and out of it came what was worth building. Part of it we built together, part of it they built, and part of it I built with my team. Running on their own server today is an assistant that answers questions about the operation from their database, with an evaluator that grades every answer before it goes out. The team that was already there came out able to keep going without me.
Twelve weeks so an executive team can judge AI on its own terms
Executive education
Before: technology decisions made on whatever the vendor of the month said.
Four in-person modules with remote follow-up, from infrastructure to business cases. Delivered in full and paid in full.
And there is more, along the line of serving customers better with fewer people: the voice
agent that answers the phone at a gas distributor and takes the routine orders, handing a
person what genuinely needs a person; and the customer portal at an electrical components
wholesaler, unused for eight years, that today lets every customer look up their price, their
inventory and their balance, and pull their own invoices.
I want one for my company
Before you write to me
Who this is for
It is for a company that is already running: it has customers, it has people, and it has
enough volume that the information, the processes and the systems are getting in the way of
growth. It is usually the owner, the CEO or the COO who comes looking, and I will talk to
anyone who has both the need and the authority to do something about it.
It stops being for you in two cases, and I would rather say so than waste your time. If what
you need is a website, a phone app or an online store, other people do that better and
cheaper, and I will gladly pass you a name. And if you already know exactly what you want
built and you are looking for someone to type it, same thing: what I charge for is precisely
the part where we work out what is worth building.
That said: working it out is where I start, and never where I stop. I build it
too, with my team or by leading the people in your own IT department
if you already have one. Both shapes are running today: one with my team at the industrial
construction group above, and the other with the programmers at the hauler.
Where it starts
Half an hour, and we go from there
- Half an hour to find where your operation binds
- Free. You tell me how you work and I tell you where I think the problem is and whether I can help. If I see that I am not the one who can help you, I say so right there, and I tell you who can.
- Operations review, two weeks, 35,000 MXN
-
I interview your people, watch the operation run, and go through your data. At the end you
get a document with five things:
- The three or four places where you are losing money or time, with what each one costs and where that figure came from.
- Which one of them is capping all the others, because fixing the rest first moves your result very little.
- What gets fixed by moving people, what by rules, and what has to be built.
- In what order it is worth doing, and the cheapest test to know whether it works before building it.
- How much effort and how much money each piece takes.
35,000 MXN plus tax, agreed before we start. And if at the end you decide
to keep going with me, it is credited back in full against the first month,
so the review ended up costing you nothing.
That document is yours even if somebody else does the work, or even if you
decide to leave it there. You can use it to have your own people do it, or to request quotes
from a vendor and grade what comes back.
I work with a small team of people who know their trade, deliberately: small
enough that your project has me in it, and not just my name on the cover. That is why we carry
three reviews at a time; if all three are taken, I tell you the date the next
slot opens.
Book the half hour
Frequently asked questions
What people ask before hiring me
If yours is here, the answer is complete. If it is missing, write to me and I will answer it.
How long does an SAP implementation take at a Mexican company, and what drives the cost?
I have led two SAP implementations. One of them was among the first done at a national level in Mexico, for a construction group of sixteen companies, and it finished on time and on budget. Time and cost depend far less on the software than on how similar the processes already are across the companies involved. What stretches these projects is discovering halfway through that every company does the same thing differently, and only then starting to negotiate it.
Can an SAP implementation that already failed be rescued?
Yes, and I have done it. The implementation at a Baja California supermarket and wholesale distribution chain had already failed twice with outside consulting firms when I came in. The third one landed with the in house technology team that was already there, without replacing a single person: what got reorganized was the work. Before restarting you need to know why it failed, because it almost always fails over business decisions left pending, and almost never over the software.
What is a fractional CTO, or technology director by the hour, and when is it worth it?
It means hiring the judgment of a technology director for a few hours a month instead of the full time position. It is worth it once a company has systems, vendors selling to it, and technology decisions worth millions, and lacks someone on its own side to judge what it is being sold. The job is translating between the executive team and the technical one, reviewing quotes and architectures, and saying what is worth building and what is worth leaving alone.
Should we buy a system, rent one, or have our own built?
It depends on how closely your operation resembles the rest of your industry. Buying one of the big platforms runs into the millions, plus a year of work just to reach the base version, and renting costs tens of thousands a year and never quite becomes yours. Building your own earns its keep when the thing that sets you apart in the market is exactly what no off the shelf system does. It can be done in phases, each one delivering something usable before the next begins.
Where does artificial intelligence belong in an operation, and where should it be kept out?
Before building anything, I sort each piece into four: what a person does, what plain explicit rules solve, what calls for machine learning, and what calls for generative artificial intelligence. Anything that touches money goes into deterministic, auditable code, meaning it can be reviewed line by line and returns the same result every time. Generative AI is reserved for what has no structure, such as reading a truck odometer from a photo. And the hard rule: the system flags, and a person draws the conclusion.
What can a midsize company automate without switching systems or buying new licenses?
Almost everything that travels today by email, by WhatsApp and by spreadsheet between one system and another. The data typed in twice, the report someone assembles by hand every Monday, the bank file reconciled against the system, the alert that something ran out of range. That gets connected on top of what you already own, so the operation keeps running while it is built, and the money comes back before anyone has to argue about replacing the system.
What does technology consulting cost, and how is it billed?
The first half hour is free: you tell me how you work and I tell you where I think your problem is, and if I am the wrong person for it, I say so right there and point you to who is. From there, the review of your operation takes two weeks and costs 35,000 MXN plus tax, credited back in full against the first month if you decide to have me build it. Longer projects run on a monthly retainer, with the scope agreed before we start.
How do I know whether my company is ready to automate?
One question answers it, and most people answer without stopping to think: if twice the work showed up tomorrow, where would it jam first. That is the bottleneck, and it usually turns out to be different from the symptom you had in mind. If the answer is a person staying late, a file somebody builds by hand, or a permit reviewed one by one, there is already something to start with. What you do need is an operation that already runs and has enough volume for the processes to get in its way.
Who is Gustavo Carreño?
A technology and artificial intelligence consultant based in Mexicali, Baja California, with 28 years in the field. I was Director of Information Technology at Grupo Hermosillo for fifteen years, from 2005 to 2020, and then ran the technology and later the whole company at Opulence Group in Dubai. Today I work with a small team on AI applied to operations, process automation, fractional technology leadership and SAP consulting.
Do you work with companies outside Mexicali?
Yes. I live in Mexicali and also work in Tijuana, across the rest of Mexico, and with US companies operating on both sides of the border. I hold two passports and two native languages, Spanish and English, so meetings and documents run in whichever one is needed. Much of the work happens remotely, and the interviews for the operations review happen at your plant or your office.
