Adopting Agentic Testing

Inside this ebook

You'll learn how successful rollouts actually run.

• Why the “replication trap” – recreating old manual processes step by step – is the single most common adoption failure
• The “new colleague” mental model that calibrates expectations and supervision
• How to scope a POC: walk the app, pick one or two flows, write success criteria down before kickoff
• The first-ten-tests discipline – why active is a promotion, not a default
• The three pace traps that look like productivity: volume, edge cases, and integrations too early
• How to build a business case that survives CFO scrutiny – parallel runs, payback math, headcount reallocation
• The 90-day onboarding plan: team training → foundation build → regression suite → PR testing
• How the QA role shifts from author to director – and what skills matter after adoption

• 

  Daniel Mauno Pettersson

  CEO

  Tech leader and hands-on engineer for the last 20 years. Previously CTO at memmo, Billogram, Dooer, CEO at Agigen.

  LinkedIn
• 

  Patrick Lef

  CPO

  Engineer-turned product leader, detail-oriented. Previously CPO/CTO, memmo, Collabs, Besedo, Videofy.

  LinkedIn
• 

  Vilhelm von Ehrenheim

  CAIO

  Built EQT's Motherbrain from nothing to a market leader. Lead Data Scientist at Klarna.

  LinkedIn

Table of contents

Seven parts. One sequential blueprint from POC to production.

1. —

   Foreword

   Why Volume 1's argument raised one question everywhere: how do we roll this out without breaking the release cycle?

2. 01

   The mindset that breaks adoption

   The replication trap, the “new colleague” frame, and the three pace traps that look like productivity but aren't.

3. 02

   Scoping – picking the right POC

   The walk-the-app call, credible test-volume estimates, one or two flows not coverage, success criteria written down, and when to say no.

4. 03

   Kickoff – the first week

   The Phase 0 infrastructure checklist, crawl before you write, and the first-ten-tests discipline that determines the next hundred.

5. 04

   Mid-POC – the discipline of patience

   Why active tests compound and passive tests drift, the 30-minute workshop pattern, and talking to the agent before escalating to a human.

6. 05

   Turning a POC into a commitment

   Building the wrap for the actual decision-maker, co-building the business case, and the three honest outcomes.

7. 06

   Onboarding – the first 90 days

   The week-by-week plan from team training to PR-triggered testing, plus the KPIs that survive executive review.

8. 07

   The role shift

   From clicker to QA manager: what changes in the day-to-day, and why quality becomes a shared cultural property.

By the numbers

What rollouts actually deliver.

• A full regression of 45 tests ran in roughly eight minutes – versus three days manually.
• Customers have eliminated 320+ manual testing hours per month in the first quarter.
• Teams have gone from sub-70% test reliability in week two to 95%+ by month three.
• One healthcare team's most expensive regression – an eight-person, multi-day exercise – was reproduced by the agent in an hour.

Quick answers

Terms this ebook uses, defined.

• What is an agentic testing POC?

  An agentic testing POC is a short, narrowly scoped trial – typically one or two critical user journeys over one to two weeks – that answers two separate questions: does the agent work on our product (proof of concept), and does it create value at our release cadence (proof of value). Conflating the two is the most common way POCs end ambiguously.

• What is the “first ten tests” discipline?

  The practice of building the first ten to twenty tests slowly and by hand-guidance, reviewing every step, running each test three times for stability, and only then marking it active. Because the agent learns from its active test set, a careful first ten makes the next hundred mostly trivial – and a sloppy first ten compounds.

• What does “active” mean for an agent-built test?

  Active is a promotion, not a default. Marking a test active tells the agent “imitate this” – so only tests with sensible natural-language steps, a sane recording, and three stable runs should earn it. Weak tests stay in draft.

FAQ

Questions teams ask before rolling out.

How long does an agentic testing POC take?

Successful POCs typically run one to two weeks of active work: a scoping call, a Phase 0 infrastructure setup, a first week producing ten to fifteen stable active tests, and a wrap-up built for the actual decision-maker. Teams that compress the first week end up redoing it; teams that stretch past a week are usually blocked on access or InfoSec items that belong in Phase 0.

How many user flows should an AI testing POC cover?

An AI testing POC should cover one or two critical user journeys – not the whole application. A POC answers whether the agent works on your product and whether it creates value at your cadence; a few well-built flows answer both, while a coverage attempt answers neither.

What's the most common reason agentic testing adoption fails?

The replication trap – recreating an existing manual click sequence step by step instead of adapting to goal-based testing. Teams that do this get a brittle imitation of the old process and conclude the tool isn't ready. The unit of work in agentic testing is the goal a user is trying to achieve, not a recorded click path.

How do you measure the success of an agentic QA rollout?

Five metrics consistently survive executive review: coverage (automated user journeys), pass-rate reliability (% non-flaky runs, target 95%+), execution time versus the manual baseline, bugs caught pre-release, and manual hours eliminated per month.

Do I need to read Past the Bottleneck before Adopting Agentic Testing?

No. Past the Bottleneck (Volume 1) makes the case for why scripted QA breaks at AI-era velocity; Adopting Agentic Testing (Volume 2) is the operational playbook for rolling agentic QA out. They stand alone, but together they cover the why and the how.

Keep reading

The rest of the field guide series.