Four tools that will tell you not to buy anything.
A readiness assessment, an ROI calculator, a stack builder and an EU AI Act Article 50 check. No sign-up, no email gate, and every result shareable by sending the link. Each one publishes its own logic, so you can check the answer instead of trusting it.
Four questions, four tools.
Should we start at all, will this specific thing pay back, and what should it be built on. Each answers one of them and none of them pretends to answer the others.
Article 50 applicability check
Which of the five paragraphs of EU AI Act Article 50 reach your system, which duty holder each falls on, and the date each applied from.
How it works
Eight questions mapped onto all five paragraphs, evaluated separately because they carry different duties and different dates. Five verdicts, three of which sell nothing, and an explicit “I would have to check” on every question that produces a cannot-tell rather than a guess. It runs entirely in the URL with no JavaScript, so nothing is stored and a result can be forwarded to counsel.
Who it is for
Anyone who has been told the December 2026 deadline applies to them, and anyone who needs to work out whether it does.
AI readiness assessment
Whether your data, governance, stakes and constraints support an AI project yet, and which engagement fits if they do.
How it works
Ten questions across four dimensions, then fourteen named rules evaluated in order. It returns one of nine fixed-scope engagements between $2,500 and $8,900, a scoped build from $18,000, or no engagement at all. Four of the rules return the last one, and the whole rule set is published on the page so the result can be argued with rather than trusted.
Who it is for
Anyone deciding whether to start, and anyone who has been told they need something and wants a second opinion with its reasoning attached.
Chatbot ROI calculator
Hours saved, cost saved and payback period against a real build cost, including the inputs where the honest answer is no.
How it works
Monthly ticket volume, average handling time, loaded cost per hour, deflection rate, build cost and running cost. It returns the payback period, the position at eighteen months, and the deflection rate you would need to break even, which is usually the more useful number. On low volume it says plainly that this is not worth building, and why.
Who it is for
Anyone building a business case for a support chatbot, and anyone who has been handed one by a vendor and wants to test its assumptions.
AI stack builder
Models, vector store, hosting and evaluation, with the trade-off behind each choice rather than a list of product names.
How it works
Four choices produce four layers, each carrying what it costs you and the condition that would reverse it. Two of the options recommend no component at all: no vector store when the material fits in the prompt, and no hosting decision until somebody establishes the residency requirement in writing.
Who it is for
Engineers and technical buyers sizing an architecture, and anyone who needs to explain to a non-technical stakeholder why a choice was made.
Four properties, and all four are checkable from the page.
Every firm says its calculator is honest. The only version of that claim worth anything is one you can verify without taking our word for it, so each of these is stated in a form you can test.
Nothing is collected
No sign-up, no email gate, no analytics event carrying what you entered, and no database anywhere behind any of the four. Three of them run entirely in your browser with no server call at all, which is why they load instantly and keep working offline once opened. The Article 50 check is a plain form, so its answers travel in the web address to be evaluated and are discarded after the page renders.
Results are shareable by link
Your inputs are encoded in the query string of your own address bar, so sending the link sends the result. A colleague opening it sees exactly what you saw, with the state restored. That is the whole sharing mechanism and it stores nothing anywhere.
Every one can tell you to do nothing
The Article 50 check has five verdicts and three of them sell nothing. The readiness assessment returns no engagement on four of its fourteen rules. The ROI calculator returns a clear no on low volume and thin margins. The stack builder recommends no vector store when the material fits in the prompt. A tool that always recommends buying is recognized as a sales form within about thirty seconds and correctly ignored.
The logic is published
The readiness rule set is printed on its page in evaluation order. The ROI arithmetic is written out line by line. The stack trade-offs are in full prose below the tool. The Article 50 check names the paragraph and the duty holder behind every verdict it gives. You can check any of them by hand, which is the only thing that makes a result worth acting on.
They are a first pass, not a diagnosis.
None of the three has seen your corpus, your query mix, your ticket history or your error tolerance, and those decide the details that cost money. A tool that gave you a number for something it cannot see would be doing the exact thing this firm exists to argue against. Treat the outputs as a way to have a better first conversation, whether that conversation is with us or with the supplier trying to sell you something.
The prices these tools name are real and published in full on the services page, with what is included and what is excluded. If procurement is the gate, the security posture page answers the questionnaire in advance and names who issues each attestation and what MetaMinds delivers in front of it.
About the tools.
These answers are the same strings this page puts in its structured data, so a person and an answer engine read identical text. Open one to read it.
Are these AI tools free?
Yes, all four, with no sign-up and no email address required. Three of them run entirely in your browser with no server call at all. The Article 50 check works differently and it is worth being precise: it is an ordinary form, so your answers do reach our server in the web address in order to be evaluated, and they are used to render the page and then discarded. There is no account, no database and no analytics event behind any of them, and nothing you enter is stored.
What are the four tools?
An AI readiness assessment, which asks ten questions on data, governance, appetite and constraints and returns a named next step or no engagement at all. A chatbot ROI calculator, which returns hours saved, cost saved and payback period against a stated build cost. An AI stack builder, which returns a four-layer architecture with the trade-off behind each choice. And an EU AI Act Article 50 transparency check, which names the paragraph and the duty holder behind every verdict it gives.
Can I share a result with a colleague?
Yes. Every input is encoded in the query string of the page URL, so copying the link from the address bar and sending it reproduces the exact result for whoever opens it. Each tool also has a copy button. Nothing is stored on any server to make this work.
When do these tools recommend nothing at all?
Each one publishes the conditions under which it recommends nothing at all. The readiness assessment returns no engagement on four of its fourteen rules, including when nothing is deployed and nothing breaks when the answer is wrong. The ROI calculator returns a clear negative on low ticket volume. The stack builder recommends no vector store when the content already fits in the context window.
What does MetaMinds charge if I do want help?
Fixed-scope engagements run from $2,500 for a three-day vendor AI assessment to $8,900 for an ISO/IEC 42001 gap assessment, with an AI security assessment at $6,500 and a hallucination and grounding audit at $5,900. Scoped builds start at $18,000. Every price is published in full on the services page.
Do these tools work without JavaScript?
The Article 50 applicability check works completely without JavaScript, because it is a plain form submitted by GET and its result is rendered on the server from the URL. The other three need JavaScript for the interactive part, but no page is a blank shell without it: the ROI calculator pre-renders a complete worked result for its default figures and prints the full arithmetic, the readiness assessment publishes its entire rule set in evaluation order, and the stack builder publishes every trade-off in prose. In each case you can reach the same answer by reading.
Which of these tools should I use?
Start with the Article 50 applicability check if somebody has told you an EU AI Act deadline applies to you and you need to know which paragraph they meant. Use the AI readiness assessment if you are deciding whether to start at all, or want a second opinion with its reasoning attached. Use the chatbot ROI calculator if you are building or testing a business case for support automation. Use the stack builder if the decision is already made and the question is what to build it on.
Do I have to give an email address to see a result?
No result on any of the four tools is gated behind an email address: the answer is on screen before anything asks for one. Three of them, the readiness assessment, the ROI calculator and the stack builder, then offer an optional one-field box that will email the result to you if you want it sent, and skipping it shows the identical result. Nothing else is collected with that address, and the message it sends carries a fixed body and a link this server builds rather than any text typed into the form.
Can I use a result from these tools in a document for my own team?
Yes, and that is what the shareable URL is for. The result is a real web address that reproduces the same output for anyone who opens it, so it can be pasted into a ticket, a board paper or an email to counsel without MetaMinds sitting in the middle of that conversation. Each tool also publishes the logic behind its verdict on the same page, which is what makes the output defensible when somebody asks where the number came from.
Send us a result, including one that said no.
A tool output is a better opening than a requirements document, because it comes with the reasoning already attached. Thirty minutes with the engineer who would do the work.
Typical reply within one business day · Fixed-scope engagements $2,500 to $8,900