Agenda

Welcome to the ADLC: coding agents are writing more code every day, and the SDLC built for human-paced development can't keep up. As Experience 2026 kicks off, Dan Belcher will discuss what changes when agents take over building and testing: how roles will shift, which processes need overhauling, and which parts of the stack need rebuilding. What happens when an agent decides whether a Jira ticket needs test coverage, builds it, runs it, and ships the result before a developer opens their laptop? Quality has never been more important, so let's reimagine the role it plays in an agentic era.

What happens when an agent decides whether a Jira ticket needs test coverage, builds it, runs it, and ships the result before a developer opens their laptop?
This session walks through how Stratus built a fleet of orchestrated bots — powered by mabl's MCP and n8n — that read tickets, inject preconditions, author and execute tests, and push straight to CI/CD, pulling in a human only for the edge cases that genuinely need one. It's not zero-touch — it's an autonomous loop, and that distinction matters: the humans still in the loop aren't testing tickets one by one anymore, they're coaching quality across the whole system.
The team behind it saw a 10x jump in quality automation productivity within 30 days. Leave with the architecture, the routing logic, and a playbook for earning your engineering org's trust in autonomous testing.

Testing has always meant specifying exact inputs and expected outputs, but that breaks down fast against non-deterministic AI. Brian Miller learned this testing three of SealingTech's AI-driven products. His shift wasn't a new technique, it was a new mindset: describe the outcome you want to achieve and the element you're checking for, and let mabl's agents find both. He'll show how that same descriptive approach validates AI outputs with no fixed answer and locates UI elements with no IDs or labels, once thought unautomatable.

The hardest questions about agentic testing don't have answers yet, which is exactly why this room sets its own agenda. There are no slides here. You bring the problems you're stuck on: how to trust an agent-authored test, where human review still belongs, what breaks when agents write faster than teams can verify. We'll cluster the topics the room cares about most, split into small groups, and dig in. You'll leave with sharper questions, a few working answers, and the names of people wrestling with the same thing you are. 

New to mabl, or mentoring someone who is? The mistakes are the same ones Bill Montbleau has flagged for years, and they aren't just a novice problem: a new user recording a 500-step test no one can maintain, and a developer writing custom JavaScript to generate a random string mabl can already do out of the box, come from the same root cause. As more teams adopt agentic test authoring in mabl, those old habits and skills matter in a new way: knowing what a test should actually test doesn't go away just because an agent is writing it. Bill walks through the fundamentals every new mabl user needs, new or experienced, agent-assisted or not.

The hardest part of testing data center networking isn't the UI: it's validating data that lives on private networks a testing platform can't reach. Ojasvi Agarwal, a systems and software engineer at HPE, shows how her team wired a custom backend into mabl so its product can prepare for weekly releases.

She'll walk through the backend integration itself, how her team authors tests, and how mabl helped them move beyond brittle tests where a single UI change could trigger hours of DOM hunting and put release timelines at risk. You'll see how they use mabl and MCP to retrain tests quickly enough to keep pace with frequent UI changes, triage failures efficiently, and validate their own datacenter assurance workflows. Beyond mabl, she'll also share how the team validates the output of their own agentic chatbot against golden answers to ensure trustworthy results.

You'll leave knowing how to extend mabl to environments it can't access directly and build a resilient test suite that keeps you release-ready, even as your product evolves week after week.

Buying a testing platform is the easy part. Getting real value from it is where most teams stall. In this fireside chat, a Senior QA Analyst from Avant-Garde Health, a healthcare technology company, shares the honest story of their mabl rollout: the early friction, the turning point, and how partnering with Procap, a mabl systems integrator, helped them put Test Authoring and Test Recovery to work — cutting manual regression effort and building coverage that's held up over time. Attendees leave with practical, repeatable lessons for getting to value faster with mabl, whether they're just starting out or scaling up.

Two days, the mabl MCP server, and no assigned problem. Point mabl's agents at whatever you've been meaning to try: a test suite that authors itself from your Jira backlog, a failure-triage bot that lives in your team's Slack, a coverage check that runs on every pull request, or something nobody's thought of yet. You'll have the Test Authoring, Test Execution, and Failure Analysis skills to build on, mabl engineers on the floor to unblock you, and demos on day two. Bring a laptop with VS Code and your coding agent of choice. You'll leave with a working prototype and learnings from other builders. Projects will be judged on user value, strategic value, technical achievement, and presentation.

Teams are completing tasks 55% faster than they were a year ago. Verifying code changes, keeping tests maintained, producing evidence someone can trust, getting ready to release — the jobs haven't changed, but who does them has. In this keynote, experience an entirely new way of working with mabl by following one code change through the entire pipeline. When agents ship the code, someone still has to stand behind it. Here's how.

Standing up an agentic testing platform is the easy part. The real work is getting an organization to trust agents, change how teams work, and sustain adoption past the pilot. In this fireside chat, Marcio Duarte of mabl partner Abstracta draws on years of guiding engineering teams through large-scale rollouts to unpack what actually stalls them. Expect the resistance patterns that surface in nearly every implementation, and patterns from successful rollouts. You'll leave with the people signals that predict success and a change approach you can run with your own team.

Manual testing and flaky Espresso tests were slowing our releases down, so we adopted mabl Mobile to automate native app testing and cut manual effort by 75% while eliminating flakiness. Integrating mabl into our CI/CD pipeline lets us verify build quality automatically for every individual build, then run our full test suite before anything reaches final QA on devices. The result: faster releases, fewer surprises, and a testing process the whole team trusts.

When agents write the first draft and set up the plan, experienced engineers lose the steps that used to make them feel useful — and morale, not tooling, becomes the real adoption risk. In this fireside chat, a panel of leaders talk through the cultural side of agentic development: how to reframe upskilling so it doesn't read as a threat, how to keep critical thinking alive when the AI skips ahead, and what it takes to hold a team's confidence steady through constant change. You'll leave with a practical way to lead your team through the transformation.

Your coding agent knows how to write code. It doesn't know how your team tests. Dani Paul, mabl's Lead Agentic Experience Engineer, has been building the public skill library at github.com/mablhq/skills, teaching Claude Code, Copilot, and Codex to set up a project, design coverage, author tests, and debug failures against mabl. She'll show what each skill does, where the library goes next, and how to turn a process your team already repeats into a skill of your own. You'll leave with the plugin installed and your first skill drafted.

Every engineering team is working out how much to hand to agents, and most of the honest conversation happens behind closed doors. This panel makes it public. mabl's own engineers build an agentic testing platform using coding agents every day, and they'll talk about what actually changed: where they let agents run, where they pump the brakes, how code review shifts when an agent wrote the diff, and what broke before it worked. You'll leave with concrete patterns for putting agents to work on your team, and a candid read on what's still hard.

Every engineering team has a few developers getting 2-5x from coding agents. Very few have made that the default for everyone. Geoff Cooney retraces how 25 engineers at mabl went from ad hoc agent use to agentic development as the standard workflow across 75+ repos: the friction that forced each architectural decision, the shared context and skill layers underneath it, and the governance that took PR volume from 291 to 732 a month on flat headcount. You'll leave with the playbook to run at your own org, and learn from the challenges we faced. 

If your team learned mabl two years ago, they learned a workflow that isn't the fastest path anymore. Dennis Nelson walks through how onboarding works now: what a new hire should learn in week one, what experienced users need to unlearn, and where the Test Authoring and Failure Analysis skills replace steps people still do by hand. He'll share the sequence mabl University uses and the checkpoints that show a team is actually ramped, not just trained. You'll leave with a ramp plan you can run for your next hire, and a list of habits worth retiring.