Resources · automationagencieshiring
Questions to ask before hiring an AI automation agency
The best questions to ask an AI automation agency are the ones whose honest answer could lose them your business. Ask what happens when the system breaks at 2am, who owns the build if you part ways, and to see a build they killed and why. A shop that answers those cleanly, even when the answer costs it the sale, is the one worth hiring.
Below is the full list we would hand a buyer, including the questions built to expose our own weak spots. For each one, the answer a good shop gives sits next to the answer that should end the call.
What questions actually matter before hiring an automation agency?#
Ten questions cover most of what matters before you sign. The pattern is simple: a good answer names a person, a process, or a limit, and a bad answer is a promise that nothing will ever go wrong.
| Question | The answer a good shop gives | The red-flag answer |
|---|---|---|
| What happens when this breaks at 2am? | A named person, a response time, and a monitor that catches it before you do. Break-fix sits in the retainer. | "It will not break," or no clear owner of the alert. |
| Who owns the build if we part ways? | You keep the rules doc, the logins, the data, and the audit trail, in writing. | The system lives on their account, or they keep the logic. |
| Show me a build you killed and why. | A real workflow they shut off, or one they told a client not to automate. | "Everything we build works." |
| Will you quote before seeing our tools? | No. They ask to see the actual workflow and tool list first. | A flat price off one sales call. |
| Where does a human approve things, and can the system skip that gate? | Named approval points, and the gate cannot be skipped on money or client work. | "It is fully autonomous," said as a selling point. |
| How do you test it against our real history first? | A shadow run against last month's actual work before it goes live. | Straight to live, or "we tune it as we go." |
| What is the smallest first version? | One workflow, live in a few weeks. | A months-long roadmap before anything runs. |
| What does month three cost, and what makes it spike? | Two numbers, build and run, and honesty that spikes come when a tool changes. | One number, no monthly line. |
| What happens when one of our tools changes its export or login? | Their monitor flags the drift, and re-wiring is part of upkeep. | "That will not affect it." |
| Can we talk to a client who took the build in-house? | Yes, someone running the system now without them. | Only references who are still paying. |
The money questions have a page of their own. What an AI automation build actually costs breaks the two numbers into honest bands and the five things that move them, and what month three actually costs covers the run number line by line. Every other row ties back to the same test: can this shop describe a moment where its work fell short, and what it does when that moment comes.
How does a good discovery call go from the other side?#
A good first call is mostly the vendor asking questions, not pitching. From our side of the table, three things happen before anyone talks price.
- We map the process first. We ask you to walk through the workflow step by step, including the steps a person does without thinking, because that is where the cost hides.
- We refuse to quote until we have seen the actual tools. A spreadsheet someone edits by hand and a tool with no way to connect are the two facts that change a price the most, and we will not guess at them.
- We insist on a shadow period. Before the system touches a live client, it runs against your last month of real work, and we compare what it would have done to what your team actually did.
You keep the process map even if you never hire us. That mirrors what you keep after a full build: the rules doc, the logins, and the audit trail. That is the difference between buying a system and renting one, and it is worth confirming in writing before you sign. You can see how we scope that first piece under the systems we build.
Which vendor answers should make you walk away?#
Three gates decide whether a deal is safe to sign. A vendor who fails any one of them will cost you more than the build ever saved.
The scope gate is about honesty on price and size. A shop that quotes a flat number off one call, or opens with a months-long roadmap, has not looked at your actual work yet. The proof gate is about whether they test before they trust: a shop that has never killed a build, or that wants to go straight to live, is asking you to be the test. The ownership gate is the one buyers forget. If the system lives on the vendor's account and the logic leaves when they do, the system is not yours to keep. Stop paying and it walks out the door with them.
When is the honest answer that you should not hire anyone?#
Sometimes the right answer is do not buy this yet, and a good shop will say so and lose the sale. If the workflow runs a few times a month, or the process changes every few weeks, the automation costs more to keep alive than the hours it saves.
Two other cases point away from an agency. If the workflow is internal, forgiving, and someone on your team enjoys tinkering, the honest call is often to build it yourself or hire later. And if a standard tool already fits your process, you may not need custom work at all, which is the first thing to settle in custom agents versus off-the-shelf automation.
The agency worth hiring is the one that runs those checks with you and tells you when the answer is no. A shop that does this loses the sale today and keeps the relationship. People come back with the right workflow later, and they come back to the shop that told them the truth the first time.