Skip to content

Insights

What should you ask a software development company before you hire one?

By Rare Agentics engineering teamLast updated First published 8 minute read

Ask who will own the code, data and accounts, how the price is fixed and what happens when scope changes, and exactly who will do the work, including how any AI agents are used and controlled. Then ask how they test and secure what they build, how often you will see working software, and what happens after launch. A good company answers each question plainly and puts the answer in writing; vague answers are the warning sign.

Why do the questions matter more than the portfolio?

A portfolio shows what a company built for someone else, under someone else's budget, with someone else's decisions. It tells you they can make software. It does not tell you what working with them will be like when the scope moves, a deadline slips or you want to take the system elsewhere.

The questions below are about those moments. They are grouped by the things that cause most of the trouble in software projects: ownership, price, who actually does the work, security, delivery and life after launch. You do not need to ask all of them in the first call. You do need answers to all of them before you sign.

One rule applies throughout: if an answer matters, it belongs in the contract or the statement of work, not only in a sales conversation.

Who will own the code, the data and the accounts?

This is the question to ask first, because it decides how much leverage you have for the rest of the relationship.

  • Whose name is on the code repositories? The answer you want is yours, from the first commit, with the company invited in as users.
  • Whose cloud accounts will it run in? Hosting, database, email and AI provider accounts should be registered to your business and billed to you.
  • Who owns the intellectual property? The contract should assign it to you, including anything written specifically for your project.
  • What happens if we part company? If the honest answer involves a handover project, an export or a licence fee, you do not own the system yet.
  • Is there any per-seat or platform fee? Some suppliers build on their own platform and charge you to keep using it. Know before you start.

Our article on owning your code and data with a software partner goes through each of these in detail, with a checklist.

How is the price set, and what happens when the scope changes?

Most disputes in software projects are about money, and most of those start with a scope change nobody priced.

  • Is the price fixed, or is it an estimate billed by the day? Both can work. A fixed price puts the estimating risk on the people doing the work. A day rate puts it on you.
  • What exactly does the price include? Design, testing, deployment, documentation, project management and the first weeks after launch are often priced separately.
  • What happens when we change our minds? Ask for the process in writing: a change is described, re-scoped and re-priced before it is built.
  • Can we start small? A good company can offer a small, paid first step that proves the working relationship before a large commitment.
  • What will it cost to run? Hosting, AI model usage and support are real costs. A supplier should raise them before you ask.

Our MVP cost guide sets out what drives the price, and our pricing page shows how we charge.

Who will actually do the work?

The people in the sales meeting are often not the people who write the code. Ask directly.

  • Who will lead the work, and how senior are they? Ask for names and roles, not a team size.
  • How much of the work is done by junior staff, contractors or offshore partners? None of these is wrong in itself. Not knowing is.
  • Who reviews the code before it is merged? The answer should be a named, experienced engineer, every time.
  • Who can release to production? It should be a named person, through a controlled process.
  • What happens if the lead engineer leaves or is ill? Ask who takes over and how quickly they can pick up the work.

How do you use AI agents, and how are they controlled?

Many software companies now use AI tools somewhere in the work. That can make delivery faster and cheaper. It can also mean code nobody has properly read. The questions are not whether they use AI, but how.

  • What do agents do in your process, and what do people do? You want a clear split: people plan, decide and check; agents do volume work within limits.
  • Can an agent reach our production systems? The right answer is no, enforced by access controls rather than instructions to the agent.
  • Does a person review every change an agent makes? If not, who is accountable for it?
  • What data from our systems will be sent to AI providers, and what do those providers keep? Ask for the providers by name and check their terms.
  • What is recorded? You want a record of which agent acted, the sources it used, what it did and who approved it.

Our guide to keeping AI agents away from production explains what good controls look like, so you can judge the answers.

How will you test and secure what you build?

Security is cheap to design in and expensive to add later. Ask before the first line of code.

  • How are access rules tested? Especially the rule that one customer can never see another customer's data. Ask whether there are automated tests for it.
  • Where will secret keys live? On the server and in a secrets store, never in front-end code or the repository.
  • Are there separate environments? At least staging and production, with backups, and a restore that has actually been tried.
  • Will you review the security of what you build before launch? And will you fix what you find within the price? If the app was built quickly with AI tools, see what a security review for AI-built apps checks.
  • Do we need a penetration test as well? Some customers and insurers require one. A good supplier will tell you when, and work alongside the firm you choose. Our comparison of a penetration test and a security review explains the difference.

How often will we see working software?

The longer you go without seeing the thing you are paying for, the more expensive a misunderstanding becomes.

  • How often will we see working software? Weekly is a good answer. "At the end of the phase" is a risky one.
  • Where can we try it? A staging environment you can use, not a demo someone drives for you.
  • How will decisions be recorded? You want one place where scope, decisions and approvals are written down.
  • What will you need from us, and when? Slow answers from your side delay the work too. A good supplier tells you up front.

What happens after launch?

Launch is the start of the system's life, not the end of the project.

  • Is there a warranty? Ask what it covers, for how long and what counts as a defect.
  • Who fixes things when they break, and how fast? If there is a support plan, ask for the response times in writing.
  • What documentation will we get? Enough for another team to take over without you having to call the original builders.
  • Can we take it in-house later? The answer should be yes, with no fee and no friction.

What are the warning signs?

QuestionA good answerA warning sign
Who owns the code?You do, in your repositories, from the first commit"We hand it over at the end" or "you pay us a licence to keep using it"
How is it priced?A written price, with a change processA loose estimate and "we will see how it goes"
What does the price include?A list, including testing and deploymentTesting or deployment quoted separately later
Who does the work?Named senior people, and who reviews every changeA team size, with no names
How do you use AI agents?A clear split, with access controls and a record of every action"We use AI to go faster", and nothing more
Can agents reach production?No, by construction"We tell them not to"
How do you test access rules?Automated tests for the rules that must never break"We test it manually before launch"
How often will we see it?Working software every weekA demo at the end of each phase
What happens after launch?A written warranty and a choice of support or handoverSupport only at a day rate, with no warranty
Can we leave?Yes, and nothing changes hands because it was always yoursA handover fee, an export or a notice period

No single warning sign is a reason to walk away. Several together usually are.

How do you compare the answers you get?

Put the answers side by side in writing. Ask each company to confirm theirs in the proposal, because a promise that will not go in writing was never really a promise. Compare like for like: a cheaper quote that leaves out testing, deployment or support is not cheaper.

It also helps to decide what kind of partner you want before you start comparing. An agency, a freelancer, an in-house hire and an agent-assisted senior team each answer these questions differently, and each suits a different business. Our comparison of agencies, freelancers and in-house teams sets out the trade-offs.

How Rare Agentics can help

Here are our own answers, so you can hold us to them. Your code lives in your repositories and your data in your cloud accounts from the first commit. Every piece of paid work has a fixed price in writing before it starts, and any change is re-scoped and re-priced in writing. Senior engineers lead every build, agents do the volume of the work, agents cannot reach production, and a person signs off every release. Any defect against the agreed acceptance criteria, found in the 30 days after release, is fixed at no cost.

If you want to judge how we work on something small first, a proof sprint delivers one bounded feature, workflow or working prototype in 1 to 2 weeks, from £3,000. For a full product, see our bespoke software builds, or for AI that acts within the bounds you set, our custom AI agents. Or start a conversation: within two working days you get a first view, and within a week a written plan, a fixed price and the risks.

Related reading

Related services

Start a conversation

Have a product or a feature in mind?

Describe the outcome in plain language. Scoping it is free, and you get a written plan and a fixed price before anything is built. Within two working days you get a first view, and within a week a written plan, a fixed price and the risks.

Or start with a proof sprint, from £3,000

In plain language: what you want to exist, who it is for, and anything that worries you. It does not need to be technical.

Read by a senior engineer. Used only to reply to you.

Tell us what you need built.