How we evaluate no-code AI tools

Contents

1. The criteria
2. The ranking frame
3. Evidence we accept
4. Keeping it current

"No-code AI" covers tools that do very different jobs: app builders, automation platforms, databases, agent builders and systems that run a whole company. A fair comparison has to say which job is being judged. These are the criteria and the rules we apply.

The criteria

1. Whole-system scope. Holds the records, the workflows and the AI in one place, instead of adding one layer to tools the company already rents.
2. Builds from a description. Produces a working first version from a plain-language description of the business or the task.
3. AI that does the work. AI agents or AI steps that carry out tasks on records, not only text generation or suggestions.
4. Works with current tools. Connects to the apps and data the company already uses, with published integration counts or connectors.
5. Approvals and control. Human approval steps, roles and permissions, and a history of what the AI changed.
6. Data ownership. Where the data lives, whether it can be exported, and whether the system can be self-hosted or moved.
7. Clear pricing. Publishes plan prices and explains how AI usage is metered before a sales call.

The ranking frame

The ranking answers one situation: a company of roughly 10 to 500 people, without a development team, that wants to run more of its daily operations with AI and is choosing where to build. For that buyer, the most valuable property is being able to describe how the business works and get one working system, with its records and AI in the same place. Tools are ranked by how well their published offer fits that situation.

A tool that ranks lower here can be the right choice for a narrower job, such as connecting two apps or publishing a customer-facing product. Each profile says when that is the case.

Evidence we accept

1. The vendor's own pages: pricing, product, documentation and security pages, read on the access date shown.
2. Named customer stories: a named company and a result that the vendor or the customer has published. Results are labelled as vendor-reported.
3. Independent research: analyst press releases, academic or industry surveys and security research from a named publisher.

We do not use anonymous reviews, affiliate comparison tables or claims we could not open ourselves. If a vendor does not publish something, the profile says so.

Keeping it current

Prices and plan limits are re-checked at least once a quarter. Every source shows the date it was read, and every page shows its last update and a change log.

Further reading

• Guide · Choosing by jobNo-code AI tools by job: which kind of tool does whatSix jobs, six kinds of tool. Most disappointment with no-code AI comes from buying a tool built for one job and asking it to do another.
• Guide · CostWhat no-code AI costs: plans, credits and the bill for a small companyEntry prices cluster between $20 and $40 a month. The real bill depends on three numbers each tool counts differently - people, records and AI usage.
• Guide · LimitsWhere no-code AI stops: the limits to know before you buildNo-code tools rarely fail on the first build. They fail later, when the data grows, someone sees records they should not, or the person who built it leaves.