OXVELO / Research
Oxvelo Research

Research.

Position papers, research reports, open studies and tools on AI-native professional services — what actually changes when a regulated industry's operating model is rebuilt rather than retooled. Every figure is attributed, every source is retrievable, and every correction we make is published.

Latest
Position paper · argument from published sources
15 August 2026 Engineering & Architecture ~32 min read Gerald Salam

The AI-augmented engineering firm.

Engineering is about to be disrupted. Not by the firms already in it.

Every serious attempt to bring AI to engineering has tried to build a better drafting tool. The tool was never the bottleneck — the enterprise around it was. This paper argues that the firm, not the engineering task, is the unit of transformation; that the engineer's job shifts from producing designs to encoding judgment and carrying accountability; and that the fastest lawful path into a licensed profession runs through someone else's licence before it runs through your own.

It states the seven strongest objections at full strength — licensure, ownership law, model capability, review capacity, the productivity evidence, the historical record, and procurement — concedes three of them, and answers the rest with the evidence.

Read the report →
73
Fact-checked sources, each with a retrievable link
14
Interactive charts and diagrams, all with data tables
90
Glossary terms defined inline, with deep dives
5
Appendices, including a published verification log
Tools & open research
Interactive tool
August 2026Firm economicsInteractive

The leverage calculator.

Enter your own utilization, overhead, operating margin and fee mix, and see what an AI-driven efficiency gain actually does to your operating profit. At the industry-median 40% fixed-fee mix, a threefold efficiency gain loses money — you need roughly 69% of revenue on fixed-fee before it pays. The model is shown in full, including what it ignores.

Open the calculator →
69%
Fixed-fee share needed before a 3× efficiency gain breaks even, at industry medians
3
Fee structures compared live against your own numbers
Open study · pre-registered, in field
August 2026Time-use studyOpen for participation

Where does an engineer's week actually go?

No published study has ever measured how design professionals spend their working hours — the industry argues about AI's productivity impact with no baseline to measure against. We are fielding the first one, and publishing the full instrument, sampling plan, analysis plan and declared conflict of interest before collecting a single response. Including what result would prove us wrong.

Read the instrument & take part →
0
Published surveys that ask clients whether they want their designers using AI
Open
Anonymised dataset released with the findings
Also published
Research report · original data, 169 sites scanned
6 July 2026 Accessibility & Compliance Gerald Salam

Oxvelo Compliance Report — July 2026

A benchmark of web accessibility compliance across three populations: a random sample of 123 top-500 websites globally, 22 leading US retail and hospitality brands under ADA Title III, and 24 Canadian federally regulated organisations under the Accessible Canada Act. Measures real-world WCAG conformance and compares reactive, litigation-driven regulation against proactive standards-based regimes.

Read the report →
169
Sites scanned across three jurisdictions
1 in 3
Top-500 sites scoring below Gold tier
93
Average score under the Accessible Canada Act, against 84 under ADA Title III

Also available: the scoring methodology behind the compliance grades.

What the labels mean

Research report — original data we collected, with a published methodology. Position paper — a thesis argued from published sources and original analysis, with no new data of its own. Open study — a research instrument published before collection, so the design can be judged before the findings exist. Interactive tool — a model you can run on your own numbers, with its assumptions shown.

We label these explicitly because the words are not interchangeable. A document that argues from other people's evidence should not borrow the authority of one that produced its own.

How we publish

Every claim in an Oxvelo report is checked against its primary source before publication, and the corrections that check produces are published alongside the report rather than quietly absorbed. Where a figure is our judgment rather than measured data, it is labelled as such at the point of use. Where a number could not be verified, it is removed and the argument is rebuilt on evidence that could.

We also grade our own evidence. Primary legal text and peer-reviewed research are treated differently from industry surveys, which are treated differently again from company-reported claims — and each report sets out that scale so a reader can weigh the argument for themselves.