AI Email Add-in for Freight Forwarders

---

Background

GoFreight is a freight forwarding management SaaS — it helps freight forwarders run their business, from shipment tracking and document management to invoicing and accounting. The people using it day-to-day are operations staff at freight forwarding companies, and a core part of their job is handling incoming emails and documents: arrival notices, invoices, bills of lading, each containing shipment data that needs to be entered into GoFreight manually. The goal of this project was to use OCR and AI to recognize that data automatically and fill it into the system, so OP staff wouldn’t have to.

The question wasn’t whether to build an AI feature — it was what form it should take, and whether teams would actually adopt it. As lead designer, my job was to help the team figure out what to build before committing to building it.

---

Choosing the right concept to test

I ran a wireframe workshop with the PM and engineers. Before presenting options, we aligned on a persona and UX principles — including “User in Control” and “Build Trust”, both specific to AI adoption — so that design decisions later had a shared reference point rather than personal preference. Three options on the table:

The initial consensus landed on A — full integration felt like the right long-term answer. Before closing out the session, we added one more filter: which option could we put in front of users fastest, and which would leave the least wasted work if we needed to change direction?

Email Add-in won. It was the most isolated concept, which meant we could build a believable prototype without touching production systems. If we pivoted, nothing learned would be thrown away.

---

Making it real enough to test

Most concept validation at this stage would be Figma prototypes. I built with Cursor instead — a fully interactive, browser-based experience that looked and behaved like an actual Outlook plugin.

The difference mattered. A Figma mock of an Outlook add-in looks like a Figma mock. Something that runs in a browser, mimics the Outlook sidebar, and responds to real interactions — that’s something a freight forwarder can actually imagine using. It was enough to lower the “is this real?” question and recruit companies before any production code existed.

Two core workflows in the prototype:

• Smart Data Update: AI parses the email to extract shipment fields (HBL/MBL, ETD/ETA, container numbers). The user reviews and confirms before anything syncs to GoFreight.
• Document Archival: One click sends the email thread and attachments to the correct shipment’s document center, labeled automatically.

---

What beta told us

7 freight forwarding companies, 12 users. We ran the beta through frequent user conversations rather than structured metrics — the goal at this stage was directional, not performance benchmarks.

The concept held. Companies were interested enough to participate, and users understood the value quickly. But one assumption didn’t survive contact with reality: we’d built for Outlook because we assumed that’s what freight forwarders use. Several beta partners were on Gmail.

If we continued down this path, every feature would need to be built twice — separate plugins for every major email client, maintained in parallel indefinitely.

---

The pivot

We kept the job-to-be-done and dropped the delivery mechanism.

The core of the feature — parse a document, extract shipment data, sync to the system — didn’t require email at all. We moved it inside GoFreight: users upload an email or document directly through the platform, and the same extraction logic runs. No email client dependency, no plugin maintenance overhead.

Because the prototype had been kept isolated from production systems, the pivot required no rollback.

The feature is now in active development as a native platform capability.

---

What I took from this

The approach became GoFreight’s internal reference for how to run early-stage concept validation.