AI & Automation

e-Conomy SEA 2026: 75% Would Let an AI Agent Buy for Them. Is Your Catalogue Machine-Readable?

November 13, 2026 · 5 min read · Etzal Group Team

The e-Conomy SEA 2026 report from Google, Temasek and Bain projects Indonesia's digital economy at US$119 billion in 2026, up 16% from US$102 billion. Katadata reported the Indonesian figures on 11 November 2026. The regional report was released by Google on 21 October 2026 with one finding that draws attention: 75% of users in Southeast Asia are willing to let an AI agent compare prices and complete a transaction on its own.

For mid-sized distributors and manufacturers, the useful question is not whether AI agents will replace salespeople. It is simpler: if a system wanted to read your products, prices and stock today, could it?

The numbers in the report

The e-Conomy SEA 2026 Indonesia figures, according to Katadata:

  • Digital economy gross merchandise value (GMV) of US$119 billion in 2026, or Rp2,157 trillion at an exchange rate of Rp18,130, with projections of US$227 billion in 2030 and US$446 billion in 2035.
  • E-commerce at US$85 billion, up 16%, about 71% of total GMV.
  • Digital payment transaction value of US$595 billion, up 13% from US$524 billion in 2025.

For the region, according to Google's post:

  • Generative AI is used daily by 243 million people in Southeast Asia, 58% of the working-age population.
  • Regional GMV is projected at US$368 billion in 2026, up 17%.
  • 72% of users rely on everyday digital automation features such as price tracking.
  • 75% are willing to let an AI agent compare prices and complete transactions independently.
  • What users expect of agent features: privacy safeguards (44%), transparency about what the AI can and cannot do (36%), and the ability to cancel a transaction made by an agent (30%).
  • The report says data centre capacity needs to expand about three times in five years and projects regional AI use rising 60 times in five years.

Reading those numbers honestly

Three limits need stating.

These are projections, not actual results. Every economic value for 2026 onward is the authors' estimate.

The 75% and 72% figures are regional. Google's page gives no Indonesia-specific willingness figure.

They describe consumers. The survey covers people shopping for themselves, not a company's purchasing department. B2B buying has layered approvals, contract prices and payment terms. No figure in our sources measures corporate buyers' willingness to hand orders to an agent.

Why should a distributor care, then? Because the people in purchasing are the same consumers after hours. The habit of automated price comparison will follow them to work, first as a research aid, long before it becomes automatic ordering. That is our view, not a finding of the report.

What "machine-readable" means

From here on, the guidance is ours.

Many mid-sized distributors keep the catalogue as a PDF, the price list as a spreadsheet sent by chat, and stock levels in the warehouse staff's heads. People can work with that. Systems cannot.

ItemHard for a machineMachine-readable
ProductsPDF brochure, photos with textTable with code, name, unit, one attribute per column
PricesA different spreadsheet per salespersonOne price source with written discount rules
StockAsked by phoneA figure a system can request, with update time
OrdersFree-text messages, retypedStandard format that goes straight into the ERP

Three things to tidy up

1. Structured product data

Each item has one code, one standard name, a clear unit, and attributes in separate columns: size, material, packaging, quantity per carton. This work pays off even if no AI agent ever arrives. New salespeople learn faster, order errors fall, and search on your own site improves.

The trade-off is time. Cleaning several thousand items takes weeks of work from people who know the products, not only from IT.

2. Prices and stock through an API

Give customers' systems an official way to request price and availability. Start with the large customers already asking. There is a real business decision here: B2B prices often differ by customer, and opening them too widely lets competitors read them too. The answer is authenticated access, where each customer sees only its own contract prices.

Stock has a trap as well. A displayed stock figure that is wrong does more harm than showing nothing. Fix stock accuracy in the warehouse first.

3. Order rules that allow cancellation

According to the report, 30% of users expect to be able to cancel a transaction made by an agent, and 36% expect transparency about the AI's limits. Translated to B2B:

  • Orders from a system arrive as drafts until confirmed, except for agreed customers and items.
  • There is a value and quantity limit per automated order.
  • There is a written cancellation window, aligned with the warehouse picking schedule.
  • Every order records its origin: a person or a system.

Privacy as the entry requirement

The expectation users named most often is privacy safeguards, at 44%. Data you open to other systems must be sorted: product data is open, price data is restricted per customer, and contact persons' personal data does not flow without a clear basis.

A sensible order of work

  1. Take your hundred best-selling items and check the data for completeness: code, unit, attributes, photo.
  2. Set one price source and stop the circulation of personal price spreadsheets.
  3. Measure stock accuracy with a sample count in the warehouse.
  4. Ask your five largest customers how they want to order two years from now.
  5. Only then decide whether to build an API, a customer portal, or both.

This order deliberately puts technology last. An agent reading bad data only produces mistakes faster.

Etzal Group works on product data and order integration through its enterprise software and AI consulting services.

Sources

Frequently asked questions

Is the 75% figure about B2B buyers in Indonesia?

No. It comes from the regional Southeast Asia report and describes consumer users, not corporate buyers and not Indonesia specifically.

What is the projection for Indonesia's digital economy in 2026?

Katadata, citing the report, gives projected GMV of US$119 billion, or Rp2,157 trillion at Rp18,130 to the dollar, up 16% from US$102 billion. It is a projection, not an actual result.

What does a machine-readable catalogue mean?

Product, price and stock data are available in structured form that another system can fetch through an API or a standard file, not only as a PDF or brochure image.

Author

Etzal Group Team

Etzal Group Team, helping businesses grow with technology. Contact us.

Related service

WhatsApp AI Chatbot

We build, train, and run a WhatsApp AI agent that answers customers 24/7, qualifies leads, and follows up.