Product teams enter most years already locked into six-to-twelve-month roadmaps: research synthesized, dependencies aligned, headcount secured. The remaining work is execution — Gantt charts, kickoffs, launch plans. Figma's Avantika Gomes argues that this is exactly the moment to pause, because the ground under those plans has shifted.
Why a fixed plan fits badly right now
AI tooling is moving quickly. Agents and assistants are reshaping daily workflows, and tools like Claude and Cursor promise to make app building as simple as writing a few prompts. Users, in turn, expect more from the software they use every day. Meeting that expectation requires a faster cadence than conventional product methodologies support, and a more experimental posture to go with it.
A roadmap offers clear direction, which is seductive amid that change. Followed prescriptively, though, it encourages teams to cling to a direction long after it has stopped being the right one.
The Dev Mode pivot
Figma itself is no longer only a design tool; it is a multi-product platform spanning FigJam, Dev Mode and, most recently, Figma Slides. Gomes spent two years building Dev Mode from scratch and describes new product work as messy, chaotic and rarely linear. The industry myth she pushes back on is that launches depend on an unwavering commitment to a single vision. Vision matters, but shipping something new also takes planning, operational juggling and enough flexibility to follow user feedback — sometimes to the point of a full pivot.
Dev Mode was first conceived as a way to automate design-to-code. Early versions looked promising, but user feedback complicated the picture: developers working with design systems already have reusable code for their components, so they spend more time assembling existing pieces than generating new ones, and the auto-generated output wasn't always useful. Figma changed direction. Rather than doubling down on codegen, it built Code Connect, which lets developers customize the snippets shown in Dev Mode so they see code from their own design system instead of generated CSS. That approach is still being iterated on today.
In hindsight the adjustment looks obvious. In practice it carried real costs: launches slipped and team members who had spent months on the original vision were frustrated. Gomes's position is that this is acceptable.
Great products aren't derailed by twists and turns; they are defined by them.
Loom, she notes, began as a network of "expert" friends offering product feedback to companies and pivoted several times before its current video recording platform broke through. Slack originally started as a multiplayer online gaming platform called Glitch. Neither company was afraid to redirect its product when learnings challenged existing assumptions.
Building the reflex into the process
Figma's stated practice is not to let sunk costs block a change of mind. Teams question whether what they are building matters and whether it is actually good, and treat course correction as always available. Projects get deep into development, get tried internally, and are sometimes restarted because something doesn't click.
Dev Mode's public launch at Config in 2023 is the illustration. Figma had a roadmap it was excited to execute, but once the beta went live and feedback arrived, responding to users took priority over the plan: more than 200 fixes and new features shipped in a month.
Changing course is still frustrating, Gomes concedes — but it can also be freeing. The freedom to change your mind and gather more information, whether through user research, internal dogfooding or beta feedback, moves a team closer to something that does feel right. The point is not attachment to a particular outcome but arriving at the best one for users.
Her suggestions for the year ahead are open-ended: prototype an idea rapidly before product and design specs are finished; make the hard call to return to square one on a product that isn't landing; re-examine longstanding assumptions that may no longer hold. Leaving a familiar framework behind for a more fluid approach is the way to judge, case by case, when to follow the roadmap and when to go off-road.



