Strixon

Strixon operates in the global supply chain and logistics technology sector, and its website is the kind that gets one decisive visit: a founder or an M&A advisor meets the firm, then looks it up afterwards to decide whether the conversation was real. For a firm whose entire pipeline is a handful of high-value relationships a year, that post-conversation lookup is the site's whole job.

The engagement was a redesign of strixon.io, run through an AI-based workflow: research, strategy, design and build handled as one continuous activity by one person, with Claude Code as the working layer throughout and the discovery material kept as a structured, sourced knowledge base rather than a deck.

The existing site was destroying credibility rather than building it, and doing so invisibly. Its highest-volume use is the lookup after a meeting, and what a checker found was Lorem Ipsum case studies, testimonials attributed to the firm's own marketing director, unlabelled performance metrics, a default "Hello world" blog post published as thought leadership, and dead privacy policy and terms links. None of that produces an analytics event or a complaint. It just quietly ends conversations, which is the largest commercial cost a site like this can have and the hardest to see.

Underneath the surface problems sat an identity conflict the redesign could not proceed without resolving. The homepage described an investment firm deploying up to 50M dollars per strategy in control positions. The one case study described a four-month AI consulting engagement bylined to a differently named entity, ending in a request-a-proposal button. Those are two different companies with two different audiences, and a designer choosing either implicitly makes a business decision for the client.

The deliverable that mattered most was not a design, it was a decision forced into the open. The discovery phase documented the identity conflict as a blocking question with two readings, investment firm or applied-AI services firm, and put it to the client rather than designing around it. The client resolved it to the services reading, which retired the investment positioning entirely and inverted the assumption most of the early strategy had been written under. The knowledge base records exactly which conclusions survived that inversion and which did not, which is what let the work continue instead of restarting.

The redesign itself was rebuilt on proof discipline. Everything on the new homepage traces to the verified evidence audit, and everything that could not be verified was quarantined: illustrative content lives in one clearly named file with a build-time content check standing between it and production, so placeholder material cannot silently ship as fact again. The readiness assessment gated high-fidelity design on the blocking decisions and let low-fidelity structure proceed, which kept the project moving without letting it outrun its facts.

The most valuable thing found in discovery was invisible by construction. A site that quietly fails the post-meeting check produces no event, no complaint, and no data; the deal just cools. Finding it required reading the site the way its one important visitor does, rather than the way an audit checklist does. I now treat "who actually looks at this, and when" as the first research question on any credibility-driven site, because the answer reshapes everything downstream of it.

An AI workflow raises the ceiling on solo scope, and the discipline that keeps that safe is separating evidence from decisions. The volume of research a single person can produce with an assistant is no longer the constraint; the constraint is knowing which sentences are sourced and which are synthesis. The two-folder rule, research is evidence, product-knowledge is decisions, is a small structure that did more for the engagement's integrity than any individual document.