AI consulting for English-speaking owners and operators running businesses in Mexico.

Find out where AI pays in your business — before you spend on it.

A 2-week Sprint produces your 90-day AI roadmap: a real baseline of your operation, a ranked list of where AI pays and where it doesn't, and a compliance review. If the answer is "almost nowhere," you get that in writing and you stop there.

Then we build what the roadmap ranked first, and we stay on to run it. Your operation's knowledge ends up in a system that keeps it, instead of in the heads of people who eventually leave.

Book a call See what we've actually built

The model is swappable; the loop isn't.

The problem

Four things that are probably true in your operation right now

Your operation runs on knowledge nobody wrote down. How the intake really works, which supplier to call, what to say to a difficult family. It lives in three or four people's heads. It walks out the door with them.
Calls and WhatsApp messages go unanswered outside business hours. The family shopping for a residence, or the patient with a Sunday-night question, does not wait for Monday. They call the next place on the list.
The same manual steps repeat every day. Intake, scheduling, follow-up, reminders. Each one depends on a person remembering, so each one fails the week that person is out.
Most AI tools sold here were built somewhere else. The demo is in English, the compliance assumptions are American, and odds are nobody on the sales call has heard of LFPDPPP, NOM-024 or how PROFECO reads a marketing claim. You are the one who finds out later.

Services

Three ways to work, in sequence

Each one stands on its own. Most engagements start at the first and stop wherever the value stops.

Sprint

A 2-week Sprint that produces your 90-day AI roadmap.

We map your workflows, your data, and your regulatory exposure. You get one document that says where AI pays in your operation, where it doesn't, and in what order to build. It is worth having whether or not you buy the Build.

Deliverables

A real baseline of your operation — the numbers you're actually running today, measured, not estimated from a conversation.
A prioritized roadmap — where AI pays, where it doesn't, and the build order, with a 90-day horizon.
A compliance review — your exposure under LFPDPPP, PROFECO and SAT, plus NOM-024 where health data is involved.

Duration:: 2 weeks. The roadmap covers the next 90 days.
Fee:: fixed fee, credited toward the Build if you continue.

Build

We implement the items the roadmap ranked first.

Voice reception, WhatsApp flows, knowledge systems, SEO — whichever ones earned the top of your roadmap. Built on tools that fit a real monthly tooling budget and wired into the calendar and records you already use. We do the technical work.

Deliverables

The systems the roadmap ranked first, built and connected to your existing operation.
Mexican compliance implemented in the build, not patched on afterwards.
A supervised go-live with monitoring, not a handover email.

Duration:: 2–4 weeks. Expect 3–4 hours of your team's time.

Retainer

We operate the system, tune it against real data, and keep it current.

Month to month. We run what we built, add capabilities as your operation changes, and keep the compliance side current as the rules move. Your institutional knowledge accumulates inside frameworks we maintain — which is the part that keeps paying after the build is done.

Deliverables

Monthly operation and tuning against your real numbers.
New capabilities as the operation changes.
Regulatory upkeep.
Your knowledge compounding inside a system you keep.

Duration:: month to month. No long-term contract.

Knowledge

When your best thinking isn't retrievable six months later

Your team reads, researches, argues, decides. Then the reasoning evaporates. Notes scatter across four apps, the good insight from March can't be found in September, and the same question gets researched twice by two different people. The reading isn't the cost. The cost is that judgment never accumulates — last quarter's conclusion can't inform this quarter's decision, because it isn't retrievable, linked, or trusted.

Knowledge management as a service

We build your operation the layer that keeps what it learns. Not a wiki nobody updates — a system with rules that make it hold:

Evidence captured verbatim, then frozen. Sources are stored exactly as received and never edited afterwards. That is what lets you act on a claim without re-verifying it.
Synthesis on top, linked, not filed. Each source becomes linked pages. The value comes from the connections between them, not from the folder tree.
Status flags that force honesty. Verified vs unverified is visible on every claim, so the system tells you what is safe to act on. For a business where a wrong claim is a liability, that distinction is the product.
An append-only history. Nothing is rewritten. That is the difference between a knowledge base and a pile of notes.
An agent as the interface. Adding a source costs minutes, not a context-switch into a note-taking ritual — which is why it actually gets used.

Your operation keeps it, and it gets more useful the longer it runs.

Who this is for

Professional firms where judgment is the product — accounting, legal, advisory — and any operation where the same research keeps getting redone.

Proof

Built, not promised

The clearest system we can point to is our own, because we control the data end to end and every number is pulled from the repository itself rather than from a recollection.

A compounding second brain (our own build)

A markdown-only, git-tracked knowledge system where an AI agent does the capture, the synthesis, and the querying. Started from zero. Here is what it held eight weeks later.

A compounding second brain (our own build)
Measure	Start (2026-05-15)	After (2026-07-08)
Synthesized, interlinked pages	0	373
Sources captured verbatim	0	219
Commits (auditable change history)	0	270
Elapsed	—	~55 days (8 weeks)
Sustained throughput	—	~7 artifacts/day

All five figures: measured, pulled from the repository's own git history and file counts on 2026-07-08 (as of July 2026). These are throughput numbers — how much the system captured and connected — not a return on investment. Any money figure we ever show you is calculated from your baseline, labelled estimated, and never presented as a guarantee.

Other systems we've designed, built, or now operate

A private school in CDMX — an end-to-end admissions funnel: paid social to an AI WhatsApp assistant to a CRM handoff.
A senior-living residence — paid acquisition, a WhatsApp assistant built from real coded conversations, local search presence, and an AI roadmap, run on a continuing retainer.
A CDMX chiropractic clinic — a brand and authority build.
A youth sports academy — campaign architecture and claims review on the copy that goes out.
An HVAC contractor in Vancouver — email-security repair, search and answer-engine baselining, and ongoing site operations.

Client names and client numbers stay off this page by policy. If you want specifics for a business like yours, ask on the call and we'll walk through what's shareable.

Why Delaren

What's hard to copy here

Not the tools. Anyone can buy the same voice platform or the same automation account. Three things around them are the reason this compounds.

1. We handle the Mexican regulatory reality — in Spanish — so you don't have to.

SAT, PROFECO, LFPDPPP, and NOM-024 where health data is involved. Your staff, your suppliers, your accountant, and the regulator all work in Spanish. You may not. Every system we build is designed to survive a local review, not an English-language demo — and the compliance thinking goes in at design time, because retrofitting it is where these projects usually die.

We do this work with real filings behind us: a Mexican trademark filed end to end through IMPI, a full LFPDPPP 2025 privacy stack in production on our own platform, tax operations running across Mexico, the US and Canada.

2. Judgment that carries forward.

Every engagement leaves reusable frameworks and verified data behind. The next client starts where the last one finished. That is why the diagnosis takes two weeks: most of the questions have been asked before, and the answers are already encoded.

3. You found us. We don't cold-call.

No cold email, no purchased lists, no outbound. Clients arrive through published work, referrals, and open-source tooling. That filters who we take on, and it means we already have context before the first meeting. It also means we can tell you honestly when the answer is no — nobody here has a quota.

The AI model can be swapped in a day. The system around it can't.

Who we serve

Who this is for

English-speaking owners and operators running businesses in Mexico. You make the decisions, your operation runs in Spanish, and the gap between those two facts costs you time every week.

Healthcare first

The spearhead is healthcare in Mexico City: clinics and specialty practices, senior-living residences, pharmacies. Condesa, Roma Norte, Polanco, Coyoacán and around. We start there because the regulatory barrier is the hardest part to copy — sensitive personal data under LFPDPPP, clinical-records rules under NOM-024, a COFEPRIS-adjacent context. The barrier is the advantage, not the obstacle.

The vertical isn't the limit

The practice is vertical-flexible. Two offers travel outside healthcare without changing shape:

Knowledge management — the compounding layer described above, for any operation where knowledge is the asset.
SEO and visibility in search and answer engines — content and positioning, built on the same encoded judgment and verified data.

What stays constant is the loop, not the sector.

Vancouver and Canada

We take Vancouver and Canadian work through referral and inbound. There's an existing client there and the bilingual Mexico–Canada operation is real. One thing is different there: the Mexican regulatory edge is the core of what we do, and it doesn't apply north of the border. What transfers is the method — diagnosis before build, systems over tools, knowledge that stays.

Process

What the engagement looks like

What the engagement looks like
Stage	What happens	How long	Your team's time
Sprint	Baseline, prioritized roadmap, compliance review	2 weeks	A few working sessions
Build	We implement what the roadmap ranked first	2–4 weeks	3–4 hours
Go-live	Supervised launch with monitoring	—	—
Retainer	Monthly operation, tuning, new capabilities, regulatory upkeep	Month to month	—

The Build number is the one people ask about: expect 3–4 hours of your team's time, total. We do the technical work.

FAQ

Questions we get

"How does pricing work?"

The Sprint is a fixed fee, agreed before it starts, and it's credited toward the Build if you continue. The Build is quoted against the scope your own roadmap defines, so we're not pricing a guess. The Retainer is monthly with no long-term contract. We don't publish numbers, because a number on a page is a number for a business we haven't met — you'll have real ones in writing after a short call.

"We already tried an AI tool and it didn't do anything."

Common, and usually the same story: somebody installed a generic assistant, translated the prompts, and left. The tool wasn't the problem so much as the absence of a diagnosis before it and an operator after it. We don't sell the tool. We sell the diagnosis of where it pays, the build that fits your operation, and staying on so your knowledge accumulates. The tool underneath is the interchangeable part.

"I'm not sure AI applies to my business at all."

That's what the Sprint is for. Instead of selling you an entry product, we map your operation and tell you honestly where AI pays and where it doesn't. If the answer is "almost nowhere," you get that in writing, you keep the baseline and the compliance review, and you skip the Build. That's a good outcome for both of us.

"Our patient and client data is sensitive. How is that handled?"

It should be your first question. LFPDPPP and NOM-024 are designed into the build, not patched in at the end, and the Sprint maps your regulatory exposure before anything is connected to anything. If a proposed automation can't be built compliantly, we say so during the Sprint rather than after the invoice.

"I need to take this to my partners before deciding."

Expected, especially for a residence or a practice with a board. For that conversation we'll send you the Sprint scope in writing, an example of what the roadmap document looks like, and a plain explanation of how we present any financial projection — always calculated from your numbers and always labelled as an estimate, never as a promise. Then we'll agree a date to pick it back up.

"Do you work in English?"

Yes — with you. The systems themselves are built to speak Mexican Spanish to your patients, families, and staff, because that's who they serve. You get the strategy, the roadmap, and the reporting in English. That split is the point.

About

Who you'd be working with

Alfredo del Arenal. Three decades in enterprise IT before this — starting at Andersen Consulting in 1994 — which is the reason you won't hear much hype on a call with me. I've watched enough technology cycles arrive as inevitable and leave as line items. The skepticism is the service. Most of what gets sold as AI transformation is a demo with a subscription attached.

I build, I don't only advise. Strategy, positioning, the build itself, the launch, and the measurement afterwards — same person through the whole sequence. That's unusual in consulting and it's the reason a diagnosis takes two weeks: I'm not writing a recommendation for someone else to implement and misread.

The practice is one operator working with a fleet of AI agents — a custom skill library, connected data and tooling servers, and automated checks that run whether or not I'm at the desk. It compresses work that would normally need a team into work one person can actually ship. It also means you deal with me, not with an account manager relaying to a delivery team.

The regulatory side is not theoretical. A trademark filed end to end through IMPI, a full LFPDPPP 2025 privacy stack live in production on our own platform, tax operations across three countries with real filings behind them.

One more thing, since it shapes everything here: whatever I build runs at very high contrast, never carries meaning in colour alone, and removes steps rather than adding them. Accessibility is how I work, not a feature I bill for. Why I design this way

AI should simplify work and amplify the people doing it. Not replace them, and not extract from them. If a project only pencils out by cutting people, it's not a project I want.

Alfredo del Arenal

Founder, Delaren Consulting · Mexico City

Independent AI strategy consultant. Three decades in enterprise IT, starting at Andersen Consulting (later Accenture) in 1994, including multi-country technical architecture and accessible-technology programs at the Canadian National Institute for the Blind.

Contact

Book a call

Thirty minutes, no deck. You describe how the operation actually runs. I ask questions — most of them about where things break, not about what you'd like to automate. By the end you'll know whether a Sprint makes sense, and if it doesn't, I'll say so and we'll both have our afternoon back.

If it does, you leave with the Sprint scope and the fee in writing.

Book a call

Prefer to write first? The booking page has a message field — tell me what's breaking and I'll come to the call having already looked.