From Tester to AI Operator: what I'm bringing to Bydgoszcz

If AI can write code, what is left for testers? The case I am taking to the Sii x ŁuczniczQA meetup in Bydgoszcz: code is cheap, judgment is the bottleneck.

From Tester to AI Operator: what I'm bringing to Bydgoszcz

If AI can write code, what is left for testers? Next Wednesday, 25 November, I'm taking that question to the Sii x ŁuczniczQA meetup in Bydgoszcz. The talk is in Polish, but the argument travels, so here it is in English.

What the job market is already saying

Layoff data from the US, Europe and Poland points in one direction. Executional work and junior roles go first. Judgment holds. That is uncomfortable for a profession that has spent years being measured by test cases executed. It is also an opening, because judgment is exactly what good testers build over a career.

Slide: Who goes first: juniors and scripted execution. Minus 13% employment of 22 to 25 year olds in AI-exposed jobs in the US; junior test-automation offers in Poland down from 46 to 6; automation's share of tester offers in Poland down from 64% to 49%.
Experienced workers in the same roles held steady. The Polish figures come from one job board: a signal, not the whole market.

The experiment I ran on myself

I wanted evidence, not opinions. So I built and shipped BowSmith, a production iOS app, without knowing Swift. What I did have was QA experience, test frameworks and the basics of architecture. I ran a team of AI agents as the project's product owner, backed by roughly 64,000 automated tests, CI pipelines and quality gates.

Slide: I didn't write the code. I ran the team that did. 0 years of coding background, about 21 months from first experiments to the App Store, 63,994 unit and integration tests, 49 international alpha testers.
Zero years of coding background, about 21 months to the App Store, 63,994 unit and integration tests.

The result surprised me less than the lesson. Code turned out to be cheap. The bottleneck was evaluating it.

Slide: Generation was cheap. Judgement was the bottleneck. At the start, 6 to 7 of every 10 hours went to judging the output; better specs and gates brought it down to about 2.
At the start, 6 to 7 of every 10 hours went to judging the output. Better specs and gates brought it down to about 2.

What agents don't have

Agents are fast. What they lack is a sense of risk, of impact and of business context. They will happily produce a thousand green checks for the wrong thing. Deciding what is worth checking, and when a result is good enough to trust, stayed with me the whole way. That is the tester's craft, moved one level up.

Slide: Execution is at risk. Information is not. At risk: running scripted regression, writing boilerplate automation, entry-level test tasks, reporting pass and fail counts. Holds: risk and impact analysis, requirements and acceptance criteria, designing the quality gates, release decisions.

Where to grow from here

In the talk I map a career path from tester to AI operator. Three parts:

  1. The advantages QA specialists already have and agents lack: risk, impact and business context.
  2. The areas worth investing in now: architecture and spec-driven development.
  3. Practical tools from my chapter “Quality at Machine Speed” in the AI Transformation anthology, which keep AI-speed delivery trustworthy.
Slide: A career path, not a cliff edge. 1 Tester, 2 Quality Engineer, 3 AI Quality Operator, 4 Quality Orchestrator.
Which rung are you on, and what's your next step?

Our experience already covers the new bottleneck. We just have to reach for it.

The evening

Wojciech Żygas, Test Architect and author of the first training in Poland on evaluating and testing AI systems, opens at 17:00. His talk tackles how to deliver an AI feature that actually works, including the uncomfortable fact that it can stop working even if it worked on delivery day. I follow at 18:15. Pizza and networking from 19:15.

Sii office, pl. Kościeleckich 3, Bydgoszcz. Registration: sii.pl/wydarzenia/sii-x-luczniczqa-2.

If you can't make it, I'll publish a recap here afterwards. And if you are making the same shift in your own team, I'd like to hear where it landed for you.