Where AI lives in CloudERM (and where it doesn't)

Updated 2026-10-28: an earlier version of this post described hour-meter digits recognized in the Inspect app, an in-app side-by-side damage view, and rate recommendations, natural-language search, and demand prediction as things CloudERM already does. That was wrong. What shipped is Inspection Analysis and, since 2026-10-19, the CloudERM connector for Claude, both described below; the other three are on the roadmap, and this version says so.

The current wave of "AI in rental software" mostly means one thing: outbound automation. Tools that scrape construction permits in your service area, generate cold emails to the procurement person, and stuff your sales pipeline with leads you didn't ask for.

We went the other direction. Every AI surface in CloudERM is aimed at the operator running the platform, not at your customers' inboxes. Here's what that looks like in practice.

The first thing that ships is Inspection Analysis, an add-on that reviews each completed inspection within minutes of the crew finishing it. Every photo is checked against the question it answers and for quality; multi-photo sets are checked for coverage; the asset number is looked for in the photos; and the hour-meter photo is read and compared with the reading the crew entered. Hour meters are the most-edited field in an inspection and the most-misread, so a reading that disagrees with its own photo is worth catching before it reaches an invoice.

On an off-rent inspection, the return photos are compared with the dispatch photos and new damage is outlined on the before-and-after pair, with a confidence level. A comparison the model can't call is marked inconclusive, not clean. The model surfaces candidates — it does not bill the customer, change a pass or fail, or edit a record. We thought hard about whether to automate the charge; we landed on "no, the model is an assistant, the human is the decider."

The results land on the inspection's Analysis tab in the web app: an A–F grade and a 0–100 score for how well the inspection was documented, with the machine's condition reported separately so new damage never counts against the crew who found it. The Technician Report averages the findings per inspector, which turns "our inspections are inconsistent" into a coaching conversation with names and numbers.

Also shipping: the CloudERM connector for Claude. Add CloudERM as a connector in Claude, sign in with your CloudERM login, and ask the questions that usually take three searches and an export — which rentals are overdue, what a customer owes, whether a unit is out and when it's due back, how fleet utilization is trending. Claude signs in as you and sees only what your CloudERM role allows. It is read-only: it can look things up, but it can't create, change, or delete anything, and every lookup is logged.

On the roadmap: per-asset rate recommendations. Setting rental rates today is a judgment call held in someone's head — age, brand, comparable assets, recent utilization, the regional going rate. That number then lives unchanged for years because nobody has time to revisit it. A recommendation that factors in utilization, asset age, maintenance cost, and comparable rates — and shows the operator the inputs behind it — is the next obvious place for AI to help. The manager will still pick the number.

Also on the roadmap: natural-language search inside the platform itself, so an operator can ask for the slice the canned report doesn't cover without filing a ticket, and fleet demand prediction, so Monday morning starts with a view of which categories will run tight by Friday and why. Both will follow the same rule: inputs visible alongside the output, and the team decides what to do with it.

The pattern across all of it: AI on the operator's side of the wheel, assists rather than replaces, evidence visible alongside the finding. Suggestions you accept, modify, or override — never opaque decisions handed down by a model.

Why not also do the AI-BDR thing? Two reasons. First, our prospects — mid-market rental dealers — are not won by cold-outbound automation. They're won by demos that match their fleet shape and references from operators with similar workflows. The AI-BDR pipeline is a cost we'd pass on to customers without a corresponding win in product utility. Second, it makes the rental software stranger for the operator. If you're trying to figure out why a quote was generated for a prospect you've never heard of, the answer "the AI did it" is the worst possible explanation. We're not in the business of explaining model outputs to confused yard managers.

The pages that go deeper: a feature-by-feature breakdown on the [/ai page](/ai), and the wider competitor framing — including which cloud peers lead with AI-for-outbound — on the [/vs page](/vs).