AI can read your data. Is your process ready?

What changed this week

For years, many businesses treated artificial intelligence as a tool for drafting copy or summarizing documents. This week offered another signal that the market is moving beyond that stage: AI is entering operating data and customer-service history directly.

On September 10, OpenAI introduced a Data agent that can connect to approved sources, investigate why a metric changed, and build shareable dashboards through plain-language questions. The more important part is not the interface. It can use the business definitions, data relationships, and permissions an organization already relies on.

A week earlier, on September 3, the company added a Zendesk integration that can review tickets and customer history, find relevant knowledge, and prepare replies while applying existing permissions. It also introduced a specialized financial-services experience with citations, access controls, encryption, and audit-log exports. These are different products, but they point in the same direction: useful AI is becoming specialized, connected, and governed.

The dashboard is no longer the bottleneck

If someone can ask, “Why did sales slow down this week?” and receive an analysis with visualizations, the time required to produce a report falls. That does not automatically turn the result into a sound decision.

The system still needs to know what counts as an active opportunity, when a customer conversation is considered handled, which date defines a sale, and which source wins when two systems disagree. If every salesperson interprets pipeline stages differently, AI can produce a fast answer that is still wrong for the business.

AI does not repair an ambiguous process by itself. It makes that process faster, more visible, and—when the data is inconsistent—harder to challenge.

Customer history becomes infrastructure

The support integration shows another shift: the value is not merely generating a response, but preparing it with the right context. For a company selling through WhatsApp, Instagram, email, and web forms, this creates a practical question: where is the complete conversation recorded?

If context is split across personal phones, spreadsheets, and inboxes, no assistant will have a reliable view. Before automating replies, define the minimum information that travels with each request, who may access it, when a case must be escalated, and which actions require human approval.

The best first pilot is not to let AI answer everything. Use it to prepare drafts for one frequent request type, measure the human corrections, and record the exceptions. That way, learning improves the process instead of hiding its weaknesses.

Economic pressure changes the priority

In Mexico, annual inflation rose to 3.26% in August from 3.12% in July, while annual core inflation eased to 3.88% from 3.95%, according to official data reported on September 9. The picture is mixed: some pressures are cooling, but costs are not moving uniformly.

For an SMB owner, that favors targeted improvements over broad cuts. Reducing rework, responding to prospects faster, preventing missed follow-ups, and controlling discount leakage may protect margin more effectively than buying another platform without changing how the team operates.

What to do in the next seven days

• Choose one repeated decision. For example: which opportunities need follow-up today, or which service cases require escalation.
• Define five metrics. Write down each metric’s name, formula, source, cadence, and owner. If two people interpret it differently, it is not ready.
• Map the real flow. Identify where information enters, who updates it, which exceptions appear, and what action should happen next.
• Run an assisted pilot. Let AI analyze or prepare a draft, while preserving human approval and logging every correction.
• Review the outcome as a process. Measure time saved, errors, missed cases, and decision quality—not merely how many tasks the tool completed.

The advantage is an operation people understand

Technology will continue to make it easier to query data, build dashboards, and prepare responses. That raises the standard for everyone. The durable advantage will be an operation where people know what to record, systems share definitions, and every automation has clear limits and owners.

Merluca helps turn fragmented sales and service processes into clear, measurable workflows that are ready for automation. If your reporting still depends on copying and pasting data, or follow-ups live across multiple inboxes, the next step is not another tool. It is organizing the process the tool must support.