AI in Peruvian companies: how to choose the first use case when the budget isn’t growing

AI in Peruvian companies: how to choose the first use case when the budget isn’t growing
AI in Peruvian companies: how to choose the first use case when the budget isn't growing

If the AI budget isn’t growing, the first use case should meet five conditions: a high-volume process, clear rules, data that already lives in a system, a measurable outcome, and low risk if the AI gets it wrong. Starting with a single, well-scoped workflow, measuring it and then expanding is safer than launching several pilots at once.

Two snapshots of the same market

Recent studies on AI in Peruvian companies seem to contradict each other. In reality, they are measuring different companies:

StudyWhat it found
BCRP survey (Central Reserve Bank of Peru, released in February 2026)49% of the companies surveyed use AI and 51% do not. Among those that use it, 66% rely on generative tools such as ChatGPT or Copilot to look things up. 56% will not increase their AI spending in 2026.
EY Peru, survey of 250 directors and shareholdersThere is interest in adopting AI, but many organizations have not yet turned that intention into concrete action. About one in five does not use AI or does not know exactly which tools it uses.
IDC with Lenovo98% of companies in Peru will increase their AI budget, and 60% are already piloting or systematically adopting it.

The InterConnecta view: part of the market is already investing, while another part, probably most mid-sized companies, will work with the same budget as last year. For that second group, the question is not whether to adopt AI, but where to put it first so that it pays for itself.

Trust is also part of the picture. The CEO of Salesforce Peru told El Peruano that confidence in data security and privacy remains the biggest challenge for 2026, followed by the availability of reliable data.

Five criteria for choosing the first use case

This is our recommendation. A good first use case meets all five criteria; if it doesn’t meet at least four, look for another one.

CriterionQuestion to validate it
1. VolumeIs it a task that repeats dozens or hundreds of times a week?
2. Clear rulesCan you explain in a few sentences how to handle it well?
3. Data in a systemIs the information already in the CRM, the ERP or the help desk, rather than in personal spreadsheets?
4. Measurable outcomeCan you measure today how long it takes or how much it costs, so you can compare later?
5. Low riskIf the AI makes a mistake, will a person catch it before it affects the customer or the finances?

Examples of first use cases that usually meet the criteria

These are illustrative examples, not results from a specific client:

  • Sales: classify and prioritize leads coming in through the website or WhatsApp before assigning them to the sales team.
  • Customer service: answer frequent questions and order-status inquiries, and hand complex cases to a person with full context.
  • Finance: summarize overdue invoices by customer and prepare draft reminders for someone to review before sending.
  • Management: answer recurring questions about sales or pipeline directly from CRM data, without building manual reports.

None of these cases requires a large project. All of them depend on having the data organized in a system such as Zoho CRM, Zoho Desk or Zoho Books.

InterConnecta experience

When we redesigned Tami, our website and WhatsApp assistant, we didn’t start by automating everything. We started with a clear flow, a specific intent and a targeted CRM integration: when the conversation needed attention, the assistant asked for name, email, phone number and a description of the request, and logged that information as a lead in the CRM, along with the conversation history.

We tested, adjusted and simplified. Only then did we expand. The lesson we apply in every project today: the technology wasn’t the most important part; the decision to simplify was.

How to measure whether the first use case worked

  1. Before you start, record three numbers: weekly volume, average time per task, and cost or effort per task.
  2. During the first few weeks, review a sample of what the AI did to measure quality, not just speed.
  3. At the end of the pilot, compare against the baseline and decide: expand, adjust or discard. A pilot discarded on the basis of data is also a good outcome.

Frequently asked questions

Where should a company in Peru start with AI if the budget is limited?

With a single high-volume process that has clear rules, data already in a system, a measurable outcome and low risk if the AI gets it wrong. Measure the baseline, test, and expand only once the result is proven.

Does using ChatGPT to look things up count as adopting AI?

It’s a first step, but it has a limit: the tool doesn’t know the company’s data or rules. The operational impact appears when AI works connected to business systems, such as the CRM or the help desk.

What if the company’s data isn’t organized?

Choose as the first use case a process whose data is already in a system, and clean up customer and sales databases in parallel. AI does not make up for incomplete or duplicate data.

How do you protect data security when using AI?

Choose solutions that work inside your systems, respect each user’s permissions, log every action and do not use your data to train models.

What should your first AI use case be?

In a diagnostic session, we review your processes, your Zoho data and your systems, and help you choose a use case you can measure.

Book a session