Which Task to Hand AI First? Pick by Impact and Complexity

Your company wants to adopt AI, but which task should you hand over first? Start with the biggest, flashiest problem and the list of prerequisites never ends. Months later, there is still nothing people actually use.

This post is based on what the daglo team heard at OpenAI Builder Bootcamp Seoul, held in Seoul on October 27, 2026. It covers the solution scoping session led by an OpenAI engineer: how to pick the right problem for AI and set its scope.

Choose your first AI project by impact and complexity

Solution scoping is the process of choosing a problem worth solving with AI and defining its success criteria, scope, and prerequisites. The basic lens is impact versus complexity: start where impact is high and complexity is low.

Impact is not only revenue. It also counts cost and time saved. The speaker's example: if a project saves customers or staff 1,000 hours a month, it is worth investing in even if it never earns money directly. Low complexity means three things are already in place:

• Process readiness: The work can be written down as a standard operating procedure (SOP).
• Technical readiness: The data exists and the systems can be connected.
• Organizational readiness: The people receiving the output will actually use it and get value from it.

Questions to ask in stakeholder interviews

Before you choose a project, you need to understand how people work today. The speaker suggested building trust with the people who do the work, then mapping the workflow together on a whiteboard while asking questions like these:

• What steps do you follow today, and in which systems?
• Which requests come in most often, and which cause the biggest problems?
• How do you measure turnaround time and quality?
• What have you tried before, and why didn't it work?
• What would the ideal future look like?

Tracing past attempts often reveals that the real cause lies somewhere else. Slow responses may come from missing procedures or systems rather than a lack of training, and not every problem needs AI to solve it.

Setting priorities: a customer support example

The speaker used a fictional e-commerce company whose support team was buried in tickets. There were three places AI could help: sorting incoming requests, investigation and resolution by specialist agents, and root cause analysis with feedback.

Sorting comes first in the workflow, but the suggested build order was root cause analysis and feedback, then sorting, then specialist resolution. The order of a process and the order you build it in are not the same thing. While the quick wins are underway, the team prepares the procedures, data, and evaluation material the slower specialist work needs.

Metrics should not be invented by engineers either. Draft measurable candidates, such as queue size or resolution time by request type, then confirm with the business owner that they actually matter.

Success means finished work, not good answers

The session's core message was to define success as finished work, not a good answer. However convincing the AI's response, the job is not done until the required action has actually happened.

• If a rule can do it, skip AI: Use AI where human judgment is needed.
• Optimize cost later: Validate a prototype that real users can use with a strong model first, then optimize against your evaluation set once the scope is clear.
• Show skeptics a demo: Let people feel how many hours a day they could save.

The easiest first project: meeting summaries and report drafts

The speaker named repetitive, easy-to-build work like meeting summaries and report drafts as tasks you can start right away. The value is obvious and the prep is light, which makes them a good first win for AI adoption.

Record a meeting or a stakeholder interview with daglo, and the transcript and summary are ready right away. If you report in the same format every week, save that format in My Templates. daglo remembers your instructions and summarizes every new recording in that format automatically. Ask BoardChat "List only the workflow bottlenecks mentioned in this interview," and your scoping notes come together fast.

When you interview several people at once, Speaker ID saves you from figuring out who said what. It remembers voices and fills in real names instead of "Speaker 1" from the next recording on, and daglo is the first service in Korea to offer it to every user. Learn more about Speaker ID

For the key takeaways from every session, see 7 Lessons Before You Build an AI Agent. To grow AI use from individuals to the whole team, read AI at work: three stages.

#AIAdoption #SolutionScoping #WorkflowAutomation #OpenAI #BuilderBootcamp #MeetingSummary #ReportDrafts #daglo

👉 Start auto meeting summaries