Product that ships, at the scale of the business.

Role: Design and technology leadership on the cross-functional tiger team (marketing, analytics, UX, and the Outgo business).

Context

DAT Outgo is DAT's factoring product. It gives carriers fast access to cash for the loads they have hauled instead of waiting weeks for a broker to pay an invoice. For most of 2025, Outgo lived mostly on its own site. The team made a deliberate bet: put Outgo front and center on the DAT.com homepage, DAT's highest-traffic page, and measure what happened.

The challenge

The homepage was running a static content carousel the team had little visibility into and no clean way to measure. The obvious risk of a dedicated Outgo module was cannibalizing DAT's core load board business, since the new module pushed the carrier signup button further down the page.

Approach

Treat the homepage as an experiment. A small tiger team swapped the carousel for a focused Outgo signup module and ran it for 30 days against a baseline. Partway through, they shipped a second iteration that added a three-months-free-factoring offer and restored a load board signup button above the fold to offset any downside to the core funnel.

What the test showed

The results were immediate. The new homepage generated 204 marketing-qualified leads in the first two weeks and 370 over the full four weeks, directly from the new content. On day one alone, 18 people filled out the homepage form, 13 started a bank application, and 2 were approved. Core load board signup completions held essentially flat, so the team judged the modest, partly-offset impact on the core funnel a clear trade for the value created. It proved a buying decision could move through homepage content alone.

Why it stuck

Rather than folding the experiment back into a rotating carousel, DAT kept Outgo as the homepage's primary focus. By late spring it was Outgo's single largest source of new leads, roughly 17% of the product's total monthly qualified-lead volume, and it held up even as other channels softened in a cooling carrier market. A formal 50/50 A/B test of the free-factoring offer now keeps sharpening the channel.

Role: Leadership on the original seven-person cross-functional team.

Context

Project Gizmo is DAT's internal attempt to let AI agents do a meaningful share of the software development work itself, turning a written requirement into working, reviewed code with minimal human effort. It started as a scrappy, time-boxed experiment in the summer of 2025.

The goal

Build an AI dark factory for development: a system that could produce production-ready, DAT-styled software by pulling context from the same places a human engineer would. That meant reading requirements out of JIRA, referencing designs in Figma, using DAT's own Cargo design system components, and opening real pull requests in GitHub.

Approach

The team built Gizmo around a small set of specialized AI agents coordinated through seven MCP servers, four of them built in-house, connecting the agents to JIRA, GitHub, Cargo, and Figma. Requirements were written in Gherkin, a structured plain-language format, so the agents could interpret them reliably. In practice, an engineer hands Gizmo a JIRA ticket with a Gherkin spec and a Figma link, and Gizmo produces an implementation plan, asks for human sign-off in a JIRA comment, writes the code, opens a pull request, and merges it once approved.

What the prototype showed

The team had a working prototype in 29 days. The most concrete early result: a newly onboarded team using Gizmo shipped a customer-facing feature in four days. DAT leadership thought enough of the results to bring the prototype in front of the parent company's board as an example of applied AI.

Why it stuck

Rather than staying one team's side project, Gizmo became a real, if still small, engineering team, listed in DAT's Carrier Engineering org with a stated goal of ubiquitous usage across DAT. Adoption spread organically, team by team. One team stood up a dedicated channel to work through using Gizmo on their own tickets; by the following summer, a separate team was setting up Gizmo's packaged tools on their own, including its Jira and Confluence MCP servers and a Figma-to-Angular component generator. The codebase grew into a shared repository publishing installable packages, the kind of durable infrastructure a one-off prototype never needs.

Role: Design leadership over the team behind this work, partnering with research, product, and engineering.

Context

DAT One is the digital evolution of DAT's truck-stop load boards, matching brokers' shipments with carriers' trucks. Brokers post shipments and search for trucks; carriers post trucks and search for shipments. Both sides vet each other for compliance.

The problem

Brokers did not trust the carriers they found on DAT One, and had no way to quickly tell whether a carrier met their company's qualification requirements. A post-COVID surge in freight fraud, from stolen and sold carrier identities to double brokering, raised the stakes, and DAT had historically let anyone sign up without identity verification. DAT's vetting tools also lived outside brokers' actual DAT One workflow, while competitors offered more robust vetting.

Approach

Research mapped how brokers actually vet carriers, a lengthy process sometimes handled by a dedicated compliance team, and the long list of requirements that matter, from FMCSA safety rating and authority age to insurance minimums, power units, crash counts, and CSA scores. Two hard constraints shaped the design: limited access to public FMCSA data, and no true filtering in search because of backend limitations. The team worked within them by annotating search results with a clear visual indicator: a green check when a carrier met the broker's qualification settings, and a red mark when they did not.

What the data showed

The hypothesis was that brokers with qualification settings would need fewer clicks per search. The experiment confirmed it: their click-to-search ratio came in 8.73% lower than the control group, with a 95% confidence interval between -10.38% and -7.09%. That was strong enough for the team to feel confident rolling qualification settings out to all brokers.

The reshape

Then real-world feedback exposed a flaw. Around 80% of carriers carry a "No Rating" status with the FMCSA, which does not mean they are bad actors, yet they were being labeled with a red X that brokers read as unqualified, making it harder for good carriers to find shipments. FMCSA insurance data is also inaccurate and rarely updated, so carriers with more than enough coverage were mislabeled. The team removed the "Satisfactory" and "No Rating" options and narrowed the setting to exclude only "Conditional" and "Unsatisfactory" carriers, temporarily pulled the auto-liability and cargo-insurance minimums until a more reliable data source was found, and replaced the red X with a neutral dash.

Next steps

The roadmap explores admin-controlled default settings, reusable qualification templates applied across every DAT One experience, and posting shipments to only carriers who meet a broker's pre-selected settings. Because hiding shipments outright would hurt the load-to-truck ratio and the perception of available freight, the team is exploring hiding the broker's contact information instead, and telling the carrier why, an education-first approach competitors are not taking.