Rebuilding a 20-Year-Old DMV System on AWS—On a 5-Month-Compressed Timeline

Tyler's Title, Lien, and Registration Search application had been running on-prem for over twenty years. It was the system of record that approved entities—banks, dealers, insurers, law enforcement—used to query protected vehicle information from the state DMV. It worked, but it looked and felt like software from another decade, and the relationship with the DMV had been deteriorating for years before I inherited it. A series of miscommunications, missed launches, and end-user frustrations under previous PMs had pushed our agency partners to begin raising questions about long-term fit. In our market, that meant a small set of established competitors who'd have been happy to take a $4M+ ARR account off our hands.

At the same time, Tyler corporate had set a strategic mandate to migrate on-prem applications to AWS. The cost case was clear at the executive level, but most product managers in our division were hesitant to volunteer their applications—the technical lift was substantial and the political risk of a botched migration was real.

When the DMV came to us asking for a major upgrade to the admin tooling, I saw a window most people would have closed. Instead of scoping the upgrade as a feature release on top of a 20-year-old codebase, I pitched a full re-architecture to AWS—bundled with the upgrade the DMV was already asking for. That meant aligning three audiences who didn't agree with each other: a skeptical director of technology who was hesitant to make this product our first full suite of web applications rebuilt in AWS, a corporate executive team that wanted cloud wins on the board, and a DMV partner who wanted features yesterday and wasn't interested in our internal architecture decisions.

The play wasn't "sell the migration." The play was making the migration the cheapest path to what every stakeholder already wanted.

I built the case separately for each audience. For the director of technology, I framed it as risk reduction—we were going to have to touch this system anyway, and doing it twice (once for the upgrade, once for the cloud migration) was the actual risky path. For our GM, I framed it as a large step in rebuilding the relationship with our agency partners at the DMV and alignment with our corporate executive team. For the DMV, I sold the upgrade and let the architecture decision be ours to make. After several rounds of pushback, all three signed off.

The original plan was twelve months: planning, development, testing, launch. Then the contracting process took five months longer than anticipated. We didn't get a deadline extension—the go-live date was tied to a contractual milestone the agency had committed to upstream. I had seven months to deliver what we'd scoped for twelve.

I rescoped the MVP three times. The first rescope happened immediately on contract signing—I sat down with the dev team and our DMV partners and ruthlessly cut anything that wasn't tied to the contractual deliverable. The second rescope came when the state mandated we switch to a new authentication and authorization service mid-build, which broke our planned data migration path. The third rescope came when our lead developer and I had to architect a workaround: legacy users would authenticate through the new service on first login, which would link their old and new accounts and let us migrate their data behind the scenes without forcing a hard cutover.

Through all three rescopes, I held two things constant: the contractual go-live date, and the user-facing functionality the DMV had told us mattered most. Everything else was negotiable, and most of it got negotiated.

I wrote the internal QA test cases and led our team through them. I wrote the external UAT plan and walked our DMV partners through it personally—partly because the testing rigor mattered, and partly because I wanted them in the system seeing it work before launch day. I co-authored the production release plan with our director of technology and we executed it on the contractual deadline.

Strategic operator. I spotted a window that combined a customer ask, a corporate mandate, and a competitive threat into a single play, sold it to three audiences who weren't aligned, and held the deadline through three forced rescopes. This is the kind of work a director of product puts on their most at-risk strategic account.