Opportunity Brief · Confidential

Marcolin Mexico
AI Opportunity Brief

Following our discovery call on Tuesday, May 12. This is a brief, not a proposal — three pains we heard, two paths to move forward.

ForAntonio Castiglione
GM Marcolin Mexico
FromGiuseppe Belpiede
Raffaello Starace
DateMay 14, 2026
TL;DR

The next step.

Following our call on Tuesday, please find this brief (not a proposal).

Ask 1

Get your feedback on this document, and high level whether you think Path A or Path B could work better.

Based on that:

  1. Get intros to your 3 direct reports (Commercial Director, Logistics + Customer Service Manager, Finance + Credit Manager) — we'll run the interviews ourselves and/or our AI agent.
  2. Discuss data-access — SAP and Salesforce: who owns it at the Mexico level, what's possible locally without going through global IT.
Where we landed

A rare fit.

Fit. Based on our focus, Marcolin Mexico is a perfect match across all four dimensions we look at (ERP underused, back-office complexity, transaction volume, multi-party coordination), which is rare and top quartile in terms of AI opportunity.

Three operational pains you named. Each one could become a project. We are intentionally not recommending ONE yet — more on why below.

Two paths forward

Two ways we could start.

Path A

Deep-dive on one pain area

"I already know which pain I want to attack — let's get to a real proposal fast."

2-3 weeks of focused investigation on a single pain area: direct-report interviews on that area, data audit, end-to-end flow walkthrough, scoping conversation. Output is a real proposal with scope and price for that one project.

If you go this route, our gut at this point — purely from the conversation on Tuesday — leans toward Pain 2 (proactive customer service) as the most tractable starting point. Smaller scope, fastest revenue unlock, you already endorsed the direction. But that's a gut read, not a recommendation. We'd validate it in the deep-dive before committing.

Path B

AI literacy workshops + project co-definition

"I want to start moving now, even though we haven't picked the project yet — and I want my team learning AI in the process."

A series of workshops with your team (you decide if top or mid management), to get your team up to speed with AI. Both live and async classes / material, spread over 2 months, 1-2 hours per week. The goal is to find 1-2 projects.

Three layers in each session:

  1. AI literacy (theoretical).
  2. Finding practical AI solutions.
  3. Project co-definition.

Output of Path B. A team that has hands-on competence with AI tools, plus a co-owned shortlist of the next 2-3 things to build. Then we pick the first one together and scope it properly.

Why this might be the right wedge for Marcolin specifically. You mentioned on the call that 1-2 of your 45 people have ever used Claude or ChatGPT. The literacy gap will be a drag on any project we build later, no matter how good the project is. Workshops close that gap and pay for themselves in the process by clarifying the project list.

The three pains

What we heard.

Each pain below is structured the same way: what you described, why AI is interesting here, and the questions we still need answered before we can scope.

Pain 1 Credit collection

What we heard

4 people manage cobranza for 1,600 active clients. Over half are in overdue. Credit-merit decisions are inconsistent — good clients denied on superficial flags, bad clients extended credit they shouldn't get.

Why AI is interesting here

Credit decisions and collection prioritization are data-rich and rule-based — exactly the territory where AI moves the needle. The data inputs likely already exist in SAP.

Open questions we still need answered before we can scope

  • What credit data actually lives in SAP today — payment history, exposure, terms per client?
  • What's the current scoring model — formal rule set, or individual judgment?
  • Is the Mexico credit policy autonomous, or are limits set by Marcolin global / VSP?
  • What's the annual write-off rate today?
  • What percentage of new clients are denied that "should" have been approved — and how would you know?
  • Where does the cobranza team actually work day-to-day — SAP screens, Excel exports, phone notes?
Pain 2 Customer service → proactive lead generation

What we heard

Customer service today is reactive — handling defects, delays, and incoming questions. Your view: it should be proactive, reaching out to dormant clients (4+ months without a purchase) with a personalized restock prompt before they slip away. Today this happens manually when it happens at all.

Why AI is interesting here

SAP has the purchase history. Each license category has an expected rotation. WhatsApp is the proven channel in Mexican B2B. The loop "dormant client detected → personalized WhatsApp prompt → hot lead flagged back to commercial" is a well-defined automation problem.

Open questions we still need answered before we can scope

  • Is the "4 months no purchase" rule already encoded somewhere, or is it fully tribal?
  • Does Salesforce already track contact history / activity log per client?
  • What's the current SAP → commercial team data pipeline?
  • Per-client "expected restock" model — does it exist anywhere, or to be defined from scratch?
  • WhatsApp Business API: does Marcolin Mexico already hold one, or does it need to be procured?
  • Is the customer-service team measured on outbound activity today, or purely on inbound SLAs?
Pain 3 Sales agent routing

What we heard

16 field agents covering ~1,550 independent optical shops. No systematic value-based routing. Agents revisit comfort clients vs. scout new ones. Appointments are inconsistent — many visits are cold drop-ins. No visibility on ROI per route.

Why AI is interesting here

Adaptive routing — where the algorithm learns from visit and order outcomes and re-weights which clients to prioritize — is the AI-specific angle. Standard VRP (vehicle routing) software solves "shortest path between fixed stops". The harder and more valuable problem at Marcolin is "which stops, in what order, at what frequency" — and that's not what VRP solves.

Open questions we still need answered before we can scope

  • Are agent visits tracked anywhere — Salesforce iPad app, manual log, none?
  • Where do orders captured during visits land — same Salesforce session, or a separate flow?
  • Is "value per route" a measurable concept today (revenue tracked per visit / per agent)?
  • Are agents employees or commission-based reps? (affects how much we can change behavior)
  • What's the current planning rhythm — weekly route plans, daily, ad hoc?
  • Which is the bigger constraint today: route design, or agent adherence to plan?
Why we're holding scope

Three reasons we're not recommending one yet.

  1. We haven't met your direct reports. What you see as GM and what they see in the daily grind will overlap maybe 60-70%. The 30% they hold is usually where the actual project lives.
  2. We don't yet know what data lives where. SAP and Salesforce are both in play, but Salesforce is at ~5% utilization and Mexico has no IT team. The first scoping move is mapping what data is actually accessible vs. what people think is there.
  3. One pain might be the symptom of another. "Customer service is reactive" might collapse into "the SAP→commercial pipeline doesn't exist" — which would change which workflow we attack first.

Premature scoping equals wrong wedge plus mispriced project plus a slow start. We'd rather invest two weeks in getting it right than rush a quote we'd have to amend twice.

About AdapttoAI

Who you're talking to.

We're Giuseppe and Raffaello — two ex-founders building AI engagements for mid-market industrial and distribution companies in LatAm and EMEA.

Today: active engagements with a lighting distributor going live on AI-powered quote generation, an industrial pod manufacturer automating their RFQ-to-descrittivo workflow, and a Mexican auto-parts distributor on WhatsApp lead capture. Same shape of problems as Marcolin's, different verticals.

We have no interest in selling you a million-dollar transformation. We have strong interest in finding the one or two workflows where AI moves the needle for Marcolin Mexico, scoped to your budget, shipped fast, and measured honestly.