The firm you choose matters more than the technology it uses. A capable team on an ordinary stack will get you further than a weak team on the newest one, and the difference shows up a year after launch, when the system needs changing and you find out who really understood it.
Most guides to choosing a software company are lists of questions. This one is too, but with the answers we think are worth hearing, including the uncomfortable ones. It is written for a company buying its first custom system or its first AI agent, and it is the list we would want a buyer to put to us.
Start with the problem, not the shortlist
Before you speak to any firm, write down the problem in your own words. Not the solution, not the feature list: the problem. "Enquiries wait until the next morning for a reply." "Job details are typed into three systems." "One person approves every order and goes on holiday twice a year."
A good firm will spend its first conversation on that sentence. It will ask what the work looks like today, who does it, what it costs you when it goes wrong and what you have already tried. A firm that moves straight to screens, timelines and a proposal has decided what to build before it understands what is broken.
One of our clients opened with this: "Our operations are not efficient. People duplicate work, redo things someone else has already done, and write the same record in several places." That is a problem statement. It names no software, and it gave us everything we needed to start asking questions.
Ten questions to ask, and what a good answer sounds like
- Who will actually do the work? You want names and roles, not "our team". Ask whether the people in the sales conversation are the people who will build it.
- What do you need to know before you can recommend anything? A good answer is a list of questions about your business. A weak answer is a request for your requirements document.
- Would you ever tell us not to build this? The honest answer is yes. Many problems close with better use of tools you already own, and a firm that never says so is selling hours.
- Why this approach and not another? For an agent, that means where it runs: inside your Microsoft 365, in a workflow tool, or as custom code. We wrote about where an agent should run and why the choice comes after the diagnosis.
- What will it be allowed to see and change? The system should inherit the permissions you already have. Be wary of anything that needs broad new access to work.
- What happens when it is wrong? Every system makes mistakes. For an agent this is the most important question on the list. A good answer names the person it hands to, the record it keeps and how you find out. Our page on an inbound call agent shows what that looks like for one phone line.
- How will we know it works before it goes live? Look for testing on your real cases, a small first release and a period where people check its work.
- Who owns the code, the accounts and the data? Get this in writing before anything is built. The answer can differ for custom software, for a product you subscribe to and for an agent, and a good firm can state each one plainly.
- What happens after launch? Ask who maintains it, how changes are requested and what training for the people who will use it is included. A system nobody is trained on becomes a system nobody uses.
- What can you show us from past work? Ask the firm for customer success stories and testimonials. A good firm shares them when you ask.
Warning signs
Some answers should end the conversation.
- A fixed scope before a single question about your business. It means the proposal was written for someone else and edited for you.
- Vague ownership. If the firm cannot say in writing who owns the code, the accounts and the data, you will find out at the worst possible time.
- No one can explain the failure path. "It is very accurate" is not an answer to "what happens when it is wrong?"
- Promises about the model instead of the process. The model will change. The process, the permissions and the person it hands to are what make an agent safe to run.
- Everything is custom. A firm that builds every part from scratch, including sign-in and reporting, is spending your time on problems that were solved years ago.
- The team changes after you sign. Ask directly, and ask for the answer in the contract.
What to have ready before the first call
You will get far more from every conversation if you bring four things.
- The problem in one sentence, and the people who live with it.
- A rough picture of the systems involved: what you use for sales, scheduling, accounts and email.
- The person who will own the result after launch, by name.
- What success looks like in plain terms: fewer missed calls, no weekly spreadsheet, replies the same hour.
You do not need a requirements document. Writing one before you understand the problem often fixes the wrong solution in place.
How we answer these ourselves
We would rather be judged on this list than on a portfolio page, so here is how we work.
An engagement with us runs in four steps: an intro call, a diagnosis, the build, then training. The diagnosis comes first because the route, whether a product we already run, an agent built inside your own systems or training for your team, is a consequence of it. Sometimes the diagnosis ends with advice to change a setting or a process and build nothing.
On ownership, question 8, our answer has three parts.
- Custom software we build for you. You own the code, and you own the intelligence in it: the rules, the prompts and the data that make it yours.
- Our own products. Scout360 and Orbit Fleets are ours. We own the code and the licence rights, and you get the right to use them for as long as you subscribe.
- Agents. You own the agent. If it relies on our infrastructure to run, that part is a paid software-as-a-service.
In every case your data stays yours. If you stop using one of our products, we give you a full export of it.
On past work, question 10: ask us, and we will share customer success stories and testimonials.
The short version
Choose the firm that asks the most about your business and promises the least about technology. Get ownership in writing. Insist on a clear answer to what happens when it is wrong. And be glad if someone tells you not to build.
Talk to sales. Bring the problem in your own words and these ten questions.
