Agile transformation
From ad-hoc Kanban to two-week Scrum, starting with a team jam
Protecting the team's time first, making the case for the right hires, then a team jam whose retro became a team agreement and a move to two-week Scrum.
Written by
Bianca Lomeli
Technical Program Manager
Good engineers want to help. That's exactly why ad-hoc delivery is so hard to fix.
When I took over the ParsePort engineering team in October 2023, after Workiva's acquisition, the team was already shipping. But delivery was ad hoc. Work ran on a loose Kanban board, testing was thin, and there were few regular rituals. When someone needed something, they messaged an engineer on Slack, and the engineer, wanting to help, picked it up.
Nobody was doing anything wrong. The process just wasn't built for the stakes. Our customers were listed companies filing regulated reports, and a release that broke something in filing season wasn't a bug. It was a risk to someone's legal filing.
Here's how we turned it into a structured process, and what I learned along the way.
Protect the team's time first
Before changing how the team planned, I had to protect the time it had.
I set up a bi-weekly product review where Professional Services met the Product Manager to talk through the issues they were seeing, so problems surfaced early instead of arriving as urgent messages. I also made it a rule that work went into Jira, and I assigned it.
Over time that faded. Old habits are strong, especially when the people asking are your colleagues and the engineers genuinely want to help. So I put formal intake gates in place, graded by urgency, along with a support buffer for the work that really couldn't wait.
Together, those protected more than 25 engineering hours a week.
Make quality someone's job
Before we had automated QA, functional testing sometimes missed side effects in unrelated parts of the product. The worst case was a production release during filing season that changed how block text rendered. It could have affected customers' reports, so I made the call to roll back.
My rule of thumb: fix forward where you can, roll back when customers are at risk. But the better answer is not needing to choose. We strengthened our automated testing, and I made the case for a dedicated QA engineer. Release failures dropped significantly.
That QA engineer was one of four hires I made the case for, along with UX and two engineers, as the team grew from 5 to 10 people.
Let the team design the change
By mid-2025 the foundations were there, but scope creep was still pulling us off course. I didn't want to hand the team a new process. I wanted them to build it.
The team was spread across Europe, so in June 2025 I planned and facilitated a team jam to bring everyone together. We started with a retro on what was causing us pain. That retro became a team agreement, and the team agreement became a move from ad-hoc Kanban to two-week Scrum.
Because the process came from the team, the team owned it. Sprints also made something visible that no rule had: once the team was planning against its own capacity, everyone could see how a quick request in Slack pushed out the work we'd committed to.
Be honest about the learning curve
We ran 10 consecutive two-week sprints in five months. We didn't finish everything we planned. Early on, the team overestimated, so work sometimes rolled over, and priorities shifted as the business changed. That's normal. Part of my job was guiding the team to plan more realistically, sprint by sprint, through sprint planning, retrospectives and backlog refinement.
I also introduced a ticket framework with acceptance criteria, so engineering and product were working from the same picture of done. Tickets moved faster, with less back-and-forth.
What I'd tell anyone leading an agile transformation
Protect the team's time before you redesign its process.
Make quality a role, not a hope.
Let the team find the pain points. They'll own the fix.
Expect the first sprints to be messy, and guide the team through it.
Make capacity visible. It does more than any rule about where requests go.
A process isn't the goal. A team that can say what it will deliver, and keep improving at it, is.