Agoda gives people tokens to experiment and still wants the draft read

A usage figure is easy to put in a town hall. The percentage is up, the licences are open, and the week of work can look unchanged. Someone still has to decide whether the paragraph that came back is the one a candidate, a colleague or a customer should see.

Paolo Inga, chief organisation and people officer at Agoda, describes an inside-out start for a workforce of around 8,000 people. Employees learn and experiment with AI first, on their own productivity. More than 90 percent use generative AI, and the company has built over 120 use cases across software development, content generation, product, finance and HR.

In late October his team brought more than 300 people from the People function to Bangkok for a hackathon on productivity. Presentations, spreadsheets and aligning boxes, work that once took considerable time, can now start from a simple prompt. He wants that time spent on meaningful work and on better decisions, and he puts the decision ahead of the time saved.

Employees receive AI tokens so they can experiment. When a team member used up her allocation on a meaningful project, Inga congratulated her, because the spend showed she was using the tools to improve the work. He has also received an eight-page document of objectives that had clearly not been reviewed or proofread. Large volumes of AI content with no critical thinking applied, he says, fall short of adoption. Efficiency depends on someone assessing the output and refining it, and people remain responsible for their outcomes, careers and performance.

He treats behavioural change as the measure of the learning. A school PDF listing 15 parent-teacher meetings would have taken him an hour to enter in Outlook. He asked Claude, and the agenda was updated in seconds. He describes the value as time back for coaching, colleagues or family. Psychological safety, permission to experiment, feedback and practice sit with managers, HR and the leadership team.

Early in a career the first draft is the easy one to keep. When he started work 25 years ago, preparing presentations and gathering information from senior colleagues was how a junior learned. AI can produce that material almost at once, so the work moves to evaluating it, improving it and tailoring it to the audience. More than 80 interns come for 10 weeks each summer. In hiring he looks for curiosity and critical thinking, while AI matches CVs, schedules interviews and helps judge who suits an AI-driven workplace.

Workforce data, on his account, is for seeing reskilling needs before a gap becomes a business problem. He sets aside any tally of emails sent. What he weights more heavily than a technical move, from recruitment to rewards or from coding to database administration, is the will to learn something new and to unlearn an old method. Agoda GROW shows people what they need to learn in order to progress. More than 2,000 people in technology have a path for depth or for managing a team, and he thinks they may climb faster while the ladder stays much as it is.

1. Hold the 90 percent to a file a manager can open. More than 90 percent of around 8,000 people use generative tools, across over 120 use cases. The use case that matters is the one a team can name.

2. Count a spent token allocation as a success only after the draft is read. He congratulated a colleague who used hers up on a meaningful project. An eight-page set of objectives that nobody proofread is the case he holds up as falling short.

3. Give the internship a draft to mark. More than 80 people come for 10 weeks. The skill he describes has moved to evaluating a first version, improving it and fitting it to an audience.

4. Ask Agoda GROW which behaviour changed. The programme tells people what to learn in order to progress, and more than 2,000 people in technology have a path. Behavioural change is the measure he set.

A critical thinking session is a practical place to take one draft a team already sends, name the person who has to mark it and write down what reviewed means before the token allocation runs out.

Sources:

• People Matters, 5 November 2026