AI Adoption · Field note

What AI adoption actually looks like inside a Philippine SME

Forget the demos. Real AI adoption starts with one workflow, one team, and a clear answer to what it's replacing. A field account of getting AI into daily operations — and what made it stick.

JS John Soriano / / 3 min read

Most AI projects don’t fail in the model. They fail in the gap between an impressive demo and a Tuesday afternoon when someone has actual work to finish. I’ve watched smart teams buy the demo, roll it out, and quietly stop using it within a month.

The businesses where it stuck did something less exciting: they picked one workflow, one team, and made the tool earn its place in a process people already cared about.

The demo problem

A good demo shows the ceiling — the most impressive thing the tool can do under perfect conditions. Adoption lives at the floor: the boring, repeated task that eats an hour every day. When you sell the ceiling and deliver the floor, people feel let down even when the floor is exactly what they needed.

So the first conversation isn’t “what can AI do?” It’s “where does work actually slow down, and who feels it?”

Start with one workflow

In this case it was a support team drowning in repetitive intake. Every request arrived as free text, got read by a person, and got manually routed. The volume wasn’t huge, but the context-switching was brutal and the backlog was always one bad day from chaos.

We didn’t “add AI to support.” We replaced one step: reading an incoming request and proposing a category, a priority, and a draft reply. A person still approved everything. Nothing went out unread.

Request Triage & draft Human approves Sent
The change was one step, not the whole workflow. The human stayed in the loop.

What made it stick

The difference between the rollouts that lasted and the ones that didn’t came down to a few unglamorous things:

  • It replaced a named task. Everyone could point to the hour it gave back.
  • The human stayed in control. Approval was one click, and the team trusted that nothing went out unseen.
  • It lived where they already worked. No new tab, no second login — it showed up inside the tool they used all day.
  • We measured the boring number. Not “AI usage.” Response time, and the size of the backlog.

Within a few weeks, response time dropped by more than half and the backlog stopped being a daily fire. Not because the model was clever — because it removed one specific bottleneck and stayed out of the way.

What I’d do differently

I’d spend even more time up front watching how people actually work, and less time in the tool’s settings. Almost every adoption problem I’ve seen traces back to automating a workflow nobody fully understood yet. The model is rarely the hard part. The process is.

If you’re a Philippine business looking at AI right now, the opportunity isn’t to copy what a US startup demoed last week. It’s to find the one workflow that quietly costs you the most, and make a tool earn its place there first.


JS
John Soriano
Technical Founder · Product Engineer

I help founders and companies design and implement AI, software, and operational systems that create real business value. Founder of XataTech.

Let's talk

Looking at AI for your business?

Tell me the workflow that quietly costs you the most. I'll help you figure out whether AI actually belongs there — and what it would take to make it stick.