
Box's Model Picker: Designing Around Vendor Lock-In Objections
Box lets buyers pick their LLM (OpenAI, Anthropic, Gemini) right on the product page — a rare CRO mechanic that turns a procurement objection into a conversion asset.
Aiko Tanaka
Design Director · · Updated
The Detail That Stops the Scroll
On Box's AI product page, the eye doesn't land on a headline benefit or a fluffy hero shot — it lands on a row of logos. OpenAI, Anthropic, Google Gemini, sitting side by side under a single line of copy: "Box AI supports best-in-class LLMs, giving you the flexibility to select the best foundational model from providers for your specific needs."
That's an unusual thing to feature above the fold. Most AI product pages sell the outcome — faster search, smarter summaries, magic answers. Box is selling choice. And for the buyer this page is actually built for — an enterprise IT or security lead evaluating vendor risk — choice is the outcome.
Reading the Hierarchy
The logo row is doing real layout work, not just decoration. Logos are high-contrast, instantly recognizable marks — your visual system processes them faster than it processes a sentence. Placed together, they trigger Gestalt's law of similarity and proximity: three distinct brand marks grouped tightly read as "one category, one decision" before you've consciously parsed any of the surrounding copy. It's a shortcut straight to the mental model "this is a switchable component," without Box having to spell it out in a paragraph.
That matters for an F-pattern scan. Enterprise buyers skim top-left to right, then drop down the left margin looking for the next anchor. A logo row breaks that horizontal scan and forces a pause — it's visually denser than surrounding text, so it earns a second look almost by default. Box is using contrast in information type, not just color, to hold attention exactly where the objection-handling needs to happen.
Why This Is a CTA-Flow Move, Not a Feature Callout
The placement is the tell. This isn't buried in a spec sheet or a help-doc appendix — it's positioned in the same visual zone as the primary conversion path, which means it's designed to be read in the same breath as the CTA. That's a deliberate sequencing choice: surface the objection, resolve it, then ask for the click.
Procurement and security teams at large orgs have a predictable, almost scripted set of blockers for any AI vendor: What model are you running on? Can we swap it if compliance flags a provider? Are we locked in? Most SaaS pages either ignore this (bury it in a whitepaper) or answer it defensively three scroll-lengths down in an FAQ accordion. Box answers it before the visitor has to ask, in the same visual field as the button they're about to click. That's a hierarchy decision: rank the objection as high-priority information, not supplementary detail.
This tracks with a core landing-page principle — align your CTA copy with your page content so nothing creates friction or a reason to bounce. If the CTA says "start free trial" but the buyer's actual blocker (vendor lock-in) never gets addressed on the page, that mismatch is exactly what sends them back to the search results. Box closes that gap in the same field of view as the ask.
The Restraint Is the Craft
What's notable is what Box didn't do. There's no dropdown gimmick, no "build your own AI stack" interactive toy, no comparison matrix with checkmarks across twelve rows. Just three logos and one sentence. That restraint matters more than it looks like it should.
Landing page best practice consistently argues for limiting the number of actions and decisions on a page — every extra choice is a tax on conversion. Box could have turned "pick your model" into an actual interactive selector with real functional weight, which would have been a much bigger design and engineering lift, and also a much bigger risk: more clicks, more decision fatigue, more chances for a buyer to get distracted before reaching the CTA. Instead they compressed the entire value prop into a static, glanceable proof point. It reads as reassurance, not as a task. The visitor never has to do anything to get the benefit of "you're not locked in" — they just have to see it.
That's a hierarchy call: the flexibility message needs to register as fact, not as an interaction. Interactions invite hesitation. Facts, presented with logo-level visual weight, get absorbed and moved past.
Where the Motion and Spacing Choices Would Pay Off
If I were pushing this further, I'd look at whether the logo row gets any entrance motion — a soft stagger-in as the section enters viewport would reinforce "these are interchangeable pieces" without adding a single word of copy. Motion is underused here as a tool for communicating modularity; three logos fading in with a 60–80ms stagger would visually perform the flexibility claim instead of just stating it.
I'd also check the whitespace ratio around the row. Trust signals like this benefit from generous negative space — it's the visual equivalent of a confident pause. Cramming the logos too close to body copy above or below dilutes the "this is important, look here" signal the grouping is trying to send.
The Takeaway for Your Roadmap
If your product has a real, defensible answer to a top-of-funnel objection — vendor lock-in, data residency, integration flexibility, whatever it is — don't let it live in an FAQ or a sales deck. Give it the same visual rank as your value prop and place it in the same eye-path as your primary CTA. A three-logo row solved more procurement friction here than a paragraph of legal-safe copy ever could. Audit your own page this sprint: find the one objection your sales team hears every single call, and ask whether your hierarchy is currently telling visitors it's important — or hiding it below the fold where only the already-convinced will find it.
Read next
More from the ABWatcher blog
Nov 22, 2026
Snake River Farms Tests a $25-Off Popup on Its Ad Landing Page
We caught Snake River Farms split-testing an email-capture form and $25-off modal on its landing page. Here's the math on what it would take to call a winner.
Nov 21, 2026
Snake River Farms Tests Mid-Page Email Capture on Its Ad Landing Page
SRF is A/B testing a mid-page lead-capture block and a $25-off popup against a simpler hero-only promo banner on its paid landing page, detected 2024 week-over-week.
Nov 20, 2026
Attio Is Testing a New Tier, Higher Prices, and a New Pitch
Attio is running a live pricing page test with a new $1,000 "Ultra" tier, higher Plus/Pro prices, and a new headline — here's what we caught and what it means.
Nov 19, 2026
Found Quietly Tested a 25% Price Cut on Its Pricing Page
We caught Found running a live pricing A/B test with an annual billing toggle that drops Found Plus from $35 to $26.25/mo. Here's what changed and why it matters.
Nov 8, 2026
Astro Proves "Zero Lock-In" With a Live Code Demo, Not a Claim
Astro's homepage replaces a trust badge row with a working, clickable framework demo — here's the hierarchy and interaction design behind it, detected 60% rare on astro.build.
Nov 7, 2026
Basecamp's Big Numbers: Why Hyper-Precise Stats Beat Vague Social Proof
Basecamp's homepage swaps "trusted by thousands" for 1.7 billion to-dos and 5 petabytes of data. We break down why precision beats vagueness as social proof — and how to test it.
From the glossary
Vocabulary the article uses. See all terms →
- Shrinkage Estimator— A statistical adjustment that pulls noisy per-variant or per-segment estimates toward an overall average to reduce overfitting to randomness.
- Flicker Test— A diagnostic check to confirm whether users briefly see the original page before a variant renders in client-side experiments.
- Pre-Experiment Bucketing Audit— A review process confirming users are being randomly and evenly assigned to experiment variants before launch.
- Above-Fold Exclusion Rule— A test-design rule that deliberately prevents changes to above-the-fold elements so an experiment isolates below-the-fold effects.
- Flicker-Free Testing— A technical setup that prevents users from briefly seeing the original page before a test variant loads.
- Orthogonal Test Design— A test structure where multiple variables are changed independently so their individual effects can be measured without interference.
See more like this
ABWatcher catches A/B tests like this across 1,000+ brands.
Watch live experiments at 1,000+ high-converting brands, complete with hypothesis and takeaway. 7 days of Pro, free, no card, then Free forever for ten watched companies.