Stormy Review: strong workflow, thin self-serve help (75/100)

Stormy is a web product for rental and field-service operations, reviewed across 12 crawled pages.

Fact	Value
Domain	stormy.ai
Category	AI operations automation for rental property and field-service businesses
Pricing	Site metadata: unknown; observed surface: Usage $0 today, Production from $300/month prepaid, Enterprise as a conversation
Crawl	12 pages on 2026-10-21
Platform	Web

Stormy earns its 75/100 through a concrete operating story, clear entry points, and a broadly strong public surface. The homepage turns a late-night repair into one coordinated job across phone, Slack, text, and email, while the consultation flow makes the next step concrete.

Usability is the clearest strength at 88.2/100, followed by technical SEO at 86.3/100, risk and stability at 84.8/100, and accessibility at 83/100. The sampled pages also show a coherent design system, no reported horizontal overflow at 390px, and a solid semantic baseline.

The score is held back by uneven depth. Docs and self-serve help is the largest weakness at 35.5/100: the crawl found a human support route but no observed self-serve support content. Decision-support surfaces score 68/100, and performance scores 68.5/100 despite homepage field signals of 1.27 s LCP, 0.01 CLS, 514 ms TTFB, and 78 ms desktop INP. Message breadth and support discoverability are the most direct opportunities to improve the buyer journey.

Stormy makes its first promise clearest to rental and property-operations teams, but its broader category is not yet arranged as a deliberate expansion path. The H1 says, ‘Hire an AI employee to run your rentals,’ while the meta description says ‘AI for businesses that run on jobs’ and the SoftwareApplication audience spans equipment service, trades, property management, freight, logistics and yards. The 4B leak workflow gives the promise operational shape, and Mike Brennan’s named role, the $1.7M raise and Y Combinator backing add credible context. The supplied evidence contains no quantified customer outcome or third-party review count. Lead with rentals as the first wedge, then introduce adjacent segments and add one attributable before-and-after result.

Stormy explains the operator’s problem better than it defines the full buyer decision. The 11:48 PM leak sequence shows a plumber booked for 9–11, a tenant asked for access and an owner approving the cost, so the audience can see the work rather than decode an AI claim. The site also makes activation concrete: $0 today, 30¢ per call minute, 20¢ per text and production from $300 a month prepaid, with no plan or minimum. The remaining confidence gap is what happens when automation gets something wrong. ‘We watch its work and fix what goes wrong’ is reassuring, but the supplied evidence does not establish an SLA, escalation path or recovery promise. Name the primary rental/property-management buyer and add a concise failure-handling block.

Stormy’s main commercial paths are short and easy to recognize, while help fails at the point of discovery. A first-time visitor can understand the offer within 10 seconds from the rental-focused headline and supporting copy. The homepage exposes Pricing and repeated Hire Stormy or Get started for free actions, and the tested routes reach How it works, pricing or /start in one homepage-to-destination step. The support page itself is direct, but the link graph records /support as depth -1 with in-degree 0, so a visitor cannot find it from the primary journey. Add Support or Help to the global footer or navigation; the page already provides a clear no-account-required contact route. The conclusion is scoped to 12 of a 40-page crawl budget.

Stormy’s sampled pages have a strong semantic baseline, but the audit still identifies concrete accessibility work. The homepage exposes header, navigation, main and footer landmarks; the sampled DOM reports zero positive tabindex values and zero unlabeled inputs. Lighthouse reports color-contrast failures on both homepage runs and on the funding article, but the supplied records do not include selectors, ratios or failing nodes, so the affected pairs need a node-level rerun before targeted CSS changes. The privacy page reports only a section landmark and hasSkipLink:false. Add a main landmark and a first-focusable skip link there, then verify the homepage prohibited ARIA attribute and the article’s non-descriptive link after locating their nodes. This automated review covers 12 of 40 crawl-budget slots.

Stormy’s visual execution is coherent across the sampled marketing templates, and the responsive foundation is mostly sound. The homepage establishes a clear primary Hire Stormy action beside a quieter Book a demo action, while pricing and blog preserve the same card-and-rule language. Sampled layout facts report scrollWidth390: 390, with no reported horizontal overflow at the 390px viewport. Three measured issues keep the execution from being finished: the homepage has an empty 30×44px button, pricing reports 12px paragraph text on mobile, and PSI reports a homepage color-contrast failure in both mobile and desktop runs. CSS extraction lists 40 colors, 39 font sizes and 21 radii, but those counts include unused or library rules. Verify computed styles before consolidating tokens.

Stormy’s homepage field signal is healthy, but the lab results expose a route-level performance problem that deserves immediate work. Mobile field data reports LCP 1.27 s, CLS 0.01 and TTFB 514 ms; desktop field INP is 78 ms. In mobile lab conditions, the homepage reached 3.15 s LCP, 795 ms TBT, 3,486 KiB transferred and 16.4 s Time to Interactive. /hire is the clearest outlier, with 15.8 s LCP, 1,419 ms TBT and 0.639 CLS across 10 layout shifts. Profile /hire first to identify the LCP request, reserve media dimensions, remove above-the-fold blockers and defer below-fold work. Then reduce the homepage’s estimated 714 KiB unused JavaScript and 4.6 s of main-thread work, checking field data after each release.

Stormy’s strongest writing makes the product observable through a scene, a time and a consequence. ‘A leak at 4B, handled overnight. One job, four places.’ gives the homepage a compact operating story, while the consultation copy names the audience, the conversation and the output. The funding article sustains that point of view with a front-desk-to-back-office thesis. Supporting pages are less finished: the privacy page averages 51.5 words per sentence across 22 sentences, the blog index contains 107 rendered words and one visible article, and the funding article title is 79 characters. The crawl also reports 68 homepage images without alt text, plus 2 on pricing and 2 on the consultation page. Keep the concrete scenes, then apply the same sentence discipline and image-label QA to the supporting surface.

Stormy’s pricing page presents three readable choices, but it leaves the central plan decision to the buyer. Usage shows $0 today with 30¢ per call minute and 20¢ per text; Production starts at $300/month prepaid and shows $5, $4.50 and $4 interaction rates as prepaid amount rises; Enterprise says Let’s talk for many locations and higher volume. Desktop cards separate the options, and the mobile capture stacks them without losing the price or feature copy. What is missing is a best-fit recommendation, a named tradeoff and a worked cost example. Add one fit line for testing, a live operation and many locations, then show two or three representative bills and explain the trigger above the $900 tier. The buyer should not need a sales conversation to model the choice.

The supplied crawl does not establish an active traffic-suppression event: sampled public pages are live, indexable, HTTPS-consolidated and raw-rendered. The clearest resilience issue is a retired URL still present in the sitemap: /blog/10-mobile-app-paywall-design-principles returns 410 and noindex while remaining listed. Remove it from the sitemap and check internal links. The /pricing/ variant also returns 200 without a redirect while /pricing is the inventory URL, creating a small canonical-policy risk. Wayback supplies 24 sampled snapshots from August 2024 through October 2026, but the aggregate history cannot prove migration healing or quantify change. Google Search Console is not connected, so no traffic loss, recovery timeline or seasonality claim is supported. Connect them before attributing demand changes to technical work.

Stormy’s sampled content clears a competent editorial bar, with the strongest work appearing on the homepage, consultation flow and funding article. QA still finds four concrete fixes: an internal article URL returns 410, the funding title is 79 characters, privacy sentences average 51.5 words, and image-alt gaps affect 68 homepage images plus 2 on pricing and 2 on the consultation page. The funding article’s BlogPosting data matches its visible title, author Robert Lukoshka, description and September 29, 2026 publication dates, so the evidence supports internal consistency on that sample. Repair the dead URL first, then fix alt text, shorten the title and edit privacy sentence-by-sentence. This is a 12-page sample from a 40-page crawl budget, not a whole-domain content judgment.

Stormy provides a clear human escalation route, but the captured self-serve layer is too thin for an AI operations product. The discovered support page is 46 words, lists installation, Slack setup, account access, billing, security and product questions, and gives a no-account-required email route. Publish a help landing page and canonical guides for setup, calls, texts, dispatch, back office, integrations, billing and troubleshooting, while retaining email for escalation.

Stormy’s sampled technical foundation is strong: robots.txt allows public paths and declares the sitemap, HTTP redirects to HTTPS in one hop, the www host consolidates to stormy.ai, and raw and rendered text counts match across all 12 sampled URLs. The cleanup list is specific. The verified sitemap contains 1,717 URLs while one listed blog URL returns 410 and noindex; /pricing/ returns 200 without redirect; and the internal-link graph reports three orphans: /hire, /support and the retired blog URL. The homepage also exposes Organization, SoftwareApplication and WebSite JSON-LD, with /llms.txt present. Mobile lab LCP remains a technical opportunity at 3,154 ms on the homepage and 5,637 ms on the sampled article, despite faster supplied field LCP. Remove the dead URL, normalize pricing and add links to live orphaned pages.

Stormy is a strong fit for rental and property-operations managers, plus adjacent field-service operators who need calls, dispatch, and back-office work brought into one job flow. Its best work is specific: the site explains the operational outcome quickly, makes starting and pricing easy to reach, and keeps the public pages coherent.

Three fixable weaknesses keep it below the top tier:

• Narrow the message around the primary rental audience before broadening into the wider job-running story.
• Make the support route easier to discover and add practical self-serve help for buyers evaluating an AI operations product.
• Run the serious mobile performance pass called for by the performance summary.

The product story is credible and concrete. The next improvement is depth around evaluation and support, not a new promise.

This 13-dimension review combines a direct crawl of 12 pages dated 2026-10-21 with product facts from the homepage's JSON-LD, sampled semantic and DOM checks, homepage CrUX/PSI field data, and the available Wayback aggregate. The public score table and dimension summaries are the source for this frame; no owner-only findings or roadmap material are used here.

Dimension 13, Review-content integrity, is excluded as not applicable, so no section is generated for it. Google Search Console enrichment is pending: it was not connected and access was unavailable for this crawl.

For the scoring definitions and evidence policy, read How SiteList scores.