In July, the IT and Business Process Association of the Philippines (IBPAP) rewrote its 2028 roadmap. The workforce target fell from 2.5 million to a best-case 2.14 million, and the revenue goal dropped from US$59 billion. e27 published its analysis on 11 November 2026 under the headline “1.89 million BPO workers, 53,500 training seats”. Its argument is that the AI numbers for the Philippines don’t add up.
I could only read the feed summary, because the full article wouldn’t load for me. So I’m not going to describe what those 53,500 seats cover or how the new revenue goal is built. The facts I have are enough for the point I want to make.
An industry body cutting its own target says something
Industry bodies rarely talk down their own industries. e27 makes that point too. A revised roadmap with a lower headcount shows people who know the business adjusting their forecast.
The common reaction is that this is a skills problem. Train more people, retrain the existing ones, and the gap closes. That is the excitement around it, and it assumes the number of seats the industry needs stays roughly where it was.
Does the headcount need to come back?
I start with whether there is a real need. Is the work that 1.89 million people do still needed in the same form? A large part of it is transactional. Where is my order, how do I reset this, can I change my booking, what does this charge on my bill mean. These are questions with answers in a system somewhere.
That is the part that changed. A competent AI customer service setup can now handle most of these enquiries. My view is that it’s 80 to 90% of what comes in. That is my opinion of what competent automation can do, not a platform-wide figure for any product. It also covers customised, multi-step workflows now, not only FAQ replies.
What didn’t change is that some conversations still need a person. A customer disputing a large charge, a family choosing an expensive purchase, someone who is upset. Those moments matter more, not less, once the routine ones are gone.
So I read the roadmap revision as a sign that transactional support work is going to AI. More training seats for the same job description would be the wrong response to it.
What I’d do if I ran a service business
Take a clinic, a shop or a marketing team that gets its customer messages on WhatsApp and chat. The default move is to hire another agent when the volume grows. I wouldn’t do that.
I would automate the transactional layer as far as it will go. Order status, bookings, rescheduling, product questions and routine requests all run without a person. Where a request goes past a limit you set, such as a refund above a certain amount, it goes to your team with the customer and the order details attached.
Then I would use the people differently. There are three jobs.
- Supervision. Someone reads the conversations the AI handled, finds where it got stuck and fixes the brief.
- Customer success. Someone follows up, wins back customers who went quiet and finds the next sale.
- The human premium. Someone is deliberately there for the moments where a human experience is what sets you apart. Big-ticket purchases are the clearest case, online or in a shop.
I think “we only use humans” is a weak position. Human-only service doesn’t scale economically, and a competitor using AI will outgrow you. Adding more seats to keep up with machine speed is not a strategy either.
Where this touches what I build
This is the shape I’m building AeroChat for. You describe the workflow in your own words, and it runs inside your customer conversations, checking, changing, booking or handing over to your team. For one bakery merchant it handles 87 to 94% of conversations with no person, through festive peaks. That figure is for that one merchant and not a rate across all customers.
What a business should do
Look at one week of your customer messages and sort them into transactional and everything else. The first pile is the work to automate. Set the limits on refunds, discounts and delivery dates yourself, and make sure anything beyond them reaches a person with full context.
Before you hire for support, decide which of your people will supervise the automation and which will spend their time on customer success and the high-value conversations. Hire for those jobs. If you need seats for the transactional layer, that is the part the Philippines’ own industry body is now planning around.
Source: e27: 1.89 million BPO workers, 53,500 training seats: The Philippines’s AI maths doesn’t add up