# Scale AI in the company: governance and stack first

Source: https://somosgentedigital.com/en-us/blog/ampliar-ia-empresa-gobernanza-bases

Date: 2026-09-30

If almost no one on the team feels ready for AI, skip a generic course. Start with governance, clean data, and support for your web, portal, or ERP.

![SGD Aura crew: fox and cat astronauts with ecru star](https://images.prismic.io/somosgentedigital-alfa/oQ3wRW6-MD6PhX8q_sgd-cover-canonico-aura-16x9.png?auto=format,compress)

The question almost always arrives the same way: “We want more AI in the company… but almost no one on the team feels ready.” That makes sense. **Artificial intelligence (AI)** — systems that help automate or suggest from data — is no longer a lab experiment; it is in chat, in spreadsheets, and in sales proposals. The problem is not curiosity. It is scaling use **without clear rules, without reliable data, and without an owner of the stack** (the web, portal, and ERP you run every day).

At Somos Gente Digital we do not sell an “AI course for everyone” or a magic tool package. We are a B2B agency / tech partner: the blog is a channel; the work is web, portals, and ERP that people can operate. Here the order is different: **governance foundations** + **accompaniment on the stack** you already use (or should use) every day.

## What the RSM LatAm report says (Sep 16, 2026)

RSM Latin America published findings on September 16, 2026 from its report AI Readiness in Latin America 2026, based on a survey of more than 400 business leaders in the region. What is verifiable in the public release:

1. Around **40%** of organizations are already implementing AI.
1. Among those that already adopted it, **68%** plan to expand use in the next 12 months.
1. Only **18%** say they have formal AI governance frameworks.
1. Only **17%** have AI policies for employees.
1. Just **4%** report a high level of training maturity; only **8%** of employees feel fully prepared to work with AI.
1. Efficiency motivates **67%**; planned investment concentrates in IT, operations, and marketing, with relatively less priority on HR and security — exactly where readiness and risk usually live.

Source: [RSM Latin America — 68% plan to expand AI use, but governance gaps remain.](https://www.rsm.global/latinamerica/es/news/68-de-las-empresas-latinoamericanas-planea-ampliar-el-uso-de-la-inteligencia-artificial-pero-hay-lagunas-de-gobernanza)

The headline sounds like “more AI.” The fine print says something else: **high ambition, weak foundations**.

## The risk is not “lack of interest”: it is scaling without foundations

Eileen Turkot (RSM International) puts it clearly in that same release: the main risk is not lack of interest, but expanding scope **before you have the capabilities** to turn that interest into measurable, sustainable value. Without clear architecture, governance, and metrics, licenses and consulting become sunk cost or technical debt.

That shows up at an owner or ops table in LatAm: someone bought a chat subscription, someone else pasted a bot on the site, marketing asks for “AI content,” and nobody knows what data left the company or who answers if the model hallucinates a price or a deadline. Scaling use in that scenario is not progress; it is **scatter**.

## What AI governance is (in plain language)

**AI governance** is the set of rules, roles, and controls that make AI use deliberate — not improvised. It is not a compliance PDF sitting in Drive. In practice it answers simple questions:

1. **What is allowed and what is not?** Permitted cases (internal summaries, reviewed drafts) vs. prohibited ones (customer data in personal tools, commercial promises without human review).
1. **Who decides?** A clear owner per area — not “everyone experiments and nobody owns it.”
1. **What data can go in?** Inventory, prices, orders, PII (personally identifiable information): what leaves the company and to which vendor.
1. **How do you measure value?** Time saved, errors avoided, tickets reduced — not only “we use AI.”
1. **What happens when it fails?** Human escalation, incident logging, rollback.

RSM links stronger governance with more mature risk assessments. Governance is not a brake: it is a **foundation to scale with confidence**.

## Why a generic course is not enough (and a magic package is not either)

A “prompting for everyone” workshop can lower anxiety. It does not fix:

1. Catalog or inventory living in three different spreadsheets.
1. A B2B portal showing prices out of sync with the ERP.
1. Nobody knowing whether the site assistant invented a shipping condition.

An “AI package” that promises magic usually pushes tools **before ordering the stack**. At SGD we have seen that with other technologies: good UI at the wrong maturity moment. With AI it is the same: the tool is not the product; the product is **operating without surprises**.

That is why we do not sell a generic course or magic. We sell **stack accompaniment** — web, [B2B / wholesale portal](/es-co/portal-b2b-mayorista), [ERP–ecommerce B2B integration](/es-co/integracion-erp-ecommerce-b2b) — so governance rules have somewhere to run: access, data, sync, and flows the team already touches.

## Practical foundations before you scale use

Before “more licenses,” a short pass helps:

1. **Inventory of current uses.** What tools exist? Who pays for them? What data goes in?
1. **A minimum employee policy.** What can be pasted into an external chat; what needs human review; what stays in company systems.
1. **Owner per use case.** A named person responsible — not “the innovation area.”
1. **Clean data where it hurts.** If ERP and store disagree, AI amplifies the error; it does not fix it. That is where [ERP–ecommerce integration](/es-co/integracion-erp-ecommerce-b2b) comes in.
1. **Bounded pilots.** One process, one metric, one deadline. Scale only if the pilot shows value and control.
1. **Training tied to real work.** Not an abstract course: “this is how we use AI in our portal / our order flow,” with guardrails already defined.

## Where the stack comes in: web, B2B portal, and ERP

Governance without stack is paper. Stack without governance is chaos. The meeting point in B2B LatAm is often:

1. **Web / site:** what gets published, what an assistant generates, who reviews before a customer reads it.
1. [B2B / wholesale portal](/es-co/portal-b2b-mayorista): list prices, quotas, recurring orders — sensitive data; AI only helps here if the portal is already an operable source of truth.
1. [ERP ↔ ecommerce](/es-co/integracion-erp-ecommerce-b2b): sync of inventory, prices, and orders. Without reliable sync, any “smart recommendation” is built on sand.

SGD accompanies that stack. We do not replace your HR policy or your lawyer; we order the digital ground where AI will live.

## B2B LatAm checklist

For owners, ops, and area leads who want to scale AI without improvising:

1. **Name current use:** list tools, owners, and data that leave.
1. **Write the minimum policy** (one page): allowed / prohibited / human review.
1. **Pick a pilot** with a metric (time, errors, tickets) and a cut-off date.
1. **Audit sync:** do ERP, portal, and store say the same thing this week?
1. **Define what never goes into external chats:** customers, internal prices, contracts.
1. **Separate training from stack:** the workshop does not replace access, roles, and clean data.
1. **Decide the next accompaniment step:** do you need to order the [portal](/es-co/portal-b2b-mayorista) or [ERP–ecommerce](/es-co/integracion-erp-ecommerce-b2b) before “more AI”?

## FAQ

### What is AI governance?

Rules, roles, and controls to use AI deliberately: what is allowed, who decides, what data goes in, how value is measured, and what happens when it fails.

### Why does it matter if only 8% of the team feels ready?

Because scaling use without readiness or policies multiplies errors and data leaks. The RSM LatAm report (Sep 2026) shows ambition (68% want to expand) with weak foundations (18% with formal governance).

### Does a company-wide AI course fix this?

Almost never by itself. It helps lower fear; it does not fix dirty data, broken sync, or missing owners. Useful training sticks to real cases and a minimum policy.

### Does SGD sell AI packages or generic courses?

No. We are a B2B agency / tech partner: we accompany governance applied to the stack (web, portal, ERP), not magic or “AI in a box.”

### When does it make sense to look at the portal or the ERP?

When the pain of scaling AI collides with inconsistent prices, inventory, or orders. Without an operable source of truth, AI amplifies the mess. Start by ordering the channel; then run the AI pilot.

### Is this the same as writing content with AI?

No. The post [Don't use AI to write](/en-us/blog/dont-use-ai-to-write) is about not using AI to replace thinking in content. This one is about scaling AI use in the company with governance and stack.

If the next step is ordering the channel where AI should live — not buying another license — see the [B2B / wholesale portal](/es-co/portal-b2b-mayorista), [ERP–ecommerce B2B integration](/es-co/integracion-erp-ecommerce-b2b), or write us via [Contact](/en-us/contact-us).

## Want to scale AI without improvising the stack?

[Contact us](/en-us/contact-us)