Seven Questions to Ask a Software Development Company Before Signing a Contract
The AI vendor market has largely emerged over the past twelve months, so you’re choosing from companies with no track record. Rankings and portfolios don’t make up for that. Here are seven questions that do—along with what the answer sounds like when a vendor is hiding something.

There are now eighteen times more Google searches for “AI software house” than there were a year ago, and most providers in this market have been in business for less than twelve months. So you’re choosing from companies with no track record, and their portfolios and rankings don’t make up for that.
Below are seven questions that do make up for it. Next to each one, I’ve included what the answer sounds like when the vendor would rather you didn’t ask any further questions.
1. Who owns the code once we end our collaboration?
Three honest answers: the code is yours and you get the repository; the code belongs to the contractor and you have a license; or the system runs on someone else’s platform and there’s nothing to hand over.
None of these is wrong in and of itself. What’s wrong is the fourth option: “We’ll figure that out when the project ends.” It needs to be decided at the beginning, because in two years, that could mean the difference between a migration costing a few thousand and building everything from scratch.
2. Who pays for model usage and how does it scale?
The cost of language models is based on the number of cases, not the number of employees. When traffic doubles, this cost doubles, but the subscription fee usually does not.
A good answer includes a figure and a condition: this much at this volume, that much at twice the volume. The answer “these are small amounts” means that no one has actually calculated it.
3. What happens if the system stops working on Saturday
The question seems trivial, but it’s decisive for the partnership.
You want specifics: who answers the phone, during what hours, what’s included in the subscription fee, and what’s charged separately. The answer “we’re available” isn’t an answer. Ask specifically about Saturday at 6 p.m., because that’s usually when things break down.
4. Am I speaking with the person who will actually be doing the work?
Sometimes one person attends the first meeting, but someone else manages the project.
That’s not a problem in itself, but you need to know in advance. Ask directly who your point of contact will be after you sign the contract and whether it will be the same person who’s making the promises now. The discrepancy between who’s selling and who’s delivering is the most common source of disappointment.
5. What is the basis for this quote?
A quote without line items isn’t a quote—it’s just a number.
You should see a breakdown: integrations, data cleansing, development, testing, and maintenance. If any item is listed as zero, ask why. Most often, data cleansing is missing—and this can be the largest part of the project—and it gets added later.
6. What exactly does the warranty cover?
In this industry, the word “warranty” can mean anything.
Agree on three parameters: what is covered, within what timeframe, and whether the repair is included in the price. A warranty that covers only bugs reported within the first month is a warranty for a demo, not for the system.
7. What’s the plan if it doesn’t work?
This is a question almost no one asks, yet it reveals the most about the vendor.
A good provider will respond calmly: “The first phase covers one process; after that, you’ll have the data; the contract ends on such-and-such a date, and the data remains with you.” A bad provider will interpret this as a lack of trust and start trying to convince you that it will work out.
The reaction to this question is more valuable than the answer itself.
What You Can’t Tell from a Portfolio
A portfolio shows what a company has built. It doesn’t show how many of those systems are still in use today.
Ask about an implementation from two years ago and whether the client is still using it. That one question reveals more than an entire gallery of screenshots.
What happens when you reach out to us
We answer all seven questions before you sign anything—in writing, if you’d like.
The owner will reach out to you, and that same person will lead the project later on. The conversation lasts 45 minutes and concludes with a price range broken down by item. The code and data remain with you.
The first scope covers a single process, not the entire ecosystem. The team subscription starts at 3,875 PLN per month, and you can calculate the cost using the calculator before the call. The guarantee covers repairs at no extra charge and doesn’t expire after one month.
Questions & Answers(FAQ)
Ask these seven questions before signing: Who owns the code once the project is complete? Who pays for the use of the models and how does that cost increase over time? What happens if there’s a failure over the weekend? Who will be the point of contact after signing the contract? How is the pricing broken down? What does the warranty cover? And what’s the exit strategy? The response to the last question says the most.
Three things: ownership of the code and data upon termination of the collaboration, the scope and response time under the warranty, and the terms for terminating the contract. The phrase “details to be determined” in connection with any of these is a red flag.
Scope of responsibility after implementation. A traditional project ends with acceptance, whereas a system based on language models requires ongoing fine-tuning because the company’s offerings and procedures change. Ask whether maintenance includes updating the system’s knowledge base or only troubleshooting.
No. A portfolio shows what a company has built, but it doesn't show how many of those systems are still in use today. Ask about an implementation from two years ago and whether the client is still using it.
The quote is free, but it should be broken down into the following categories: integrations, data cleanup, development, testing, and maintenance. If any of these categories is listed as zero, ask for clarification. Data cleanup is most often omitted, and it can be the largest part of the project.
Got a similar process on your side?
If something in this article sounds like your day-to-day - let's talk. We'll tell you plainly what can be improved, and what's not worth touching.
Services related to this topic
Custom software with AI
Off-the-shelf tools force you to compromise. We build a system around your process, with AI where it genuinely gives an edge.
AI agents with integrations
Not another chatbot. An agent wired into your email, CRM and knowledge base that takes over repetitive work.
AI automation for business
Repetitive work a machine does faster and error-free. AI only where it pays off.