How to Choose the Right Partner for the Job: Key Questions
The questions worth asking before you hire an AI or automation partner, from Solar Quest Ai — plus a quick comparison and an honest FAQ.
Choosing the right partner for an AI or automation project means asking about failure rates, data handling, ownership, and post-launch support before you ask about price. At Solar Quest Ai, founded in 2022 on Marszałkowska in Warsaw, we've seen roughly 70% of failed builds trace back to unclear scope, not bad code [1].
What should you ask before hiring an AI partner?
Most people ask first about price. That's the wrong first question. Ask about the failure rate. Ask what happens when the model drifts, when the API changes, when your data volume triples in a quarter. A good partner will answer with specifics, not slogans.
Here's the short list we hand to prospects who are also talking to two or three other shops:
1. What percentage of your builds are still running two years after launch, and can you name three?
2. Who owns the code, the prompts, the fine-tuning data, and the model weights when the contract ends?
3. What is your process when a third-party model deprecates an endpoint we depend on?
4. How do you handle personally identifiable information under GDPR, and where is it stored [2]?
5. What does month-13 look like, meaning after the initial warranty period ends?
6. Can I speak to a client who fired you, or a project that went sideways?
The last one matters most. Anyone can produce a glowing reference. The partner who'll send you to a client that walked away is the partner who has nothing to hide. For more on how we think about long-term builds, see our notes from behind the work.
How do you compare AI vendors fairly?
Comparing vendors is hard because everyone's deck looks the same. Strip the decks away. Look at four things: who owns the output, how they price, who they're actually built for, and when you shouldn't hire them.
| Partner type | Typical cost (PLN) | Best for | Avoid when |
|---|---|---|---|
| Solo freelancer | 8,000–25,000 | One-off scripts, prototypes | You need 24/7 uptime |
| Boutique studio | 30,000–120,000 | Custom workflows, integrations | Budget is under 20k |
| Enterprise consultancy | 250,000+ | Regulated industries at scale | You're pre-revenue |
We sit in the boutique row. A recent project, a February 2024 build for a 41-year-old logistics operator named Marek, ran 47,000 PLN and replaced roughly 22 hours of weekly manual sorting. He'd been quoted 180,000 by a larger firm for the same scope. Bigger isn't better. Bigger is bigger.
Why does post-launch support matter more than the build itself?
Because the build is a snapshot. Your business isn't. A tool that fit perfectly in March will chafe by October if nobody's maintaining it, and the maintenance conversation is the one most vendors avoid until the contract is already signed and you've paid the deposit.
Ask what happens in month 13. Ask about retainers. Ask whether bug fixes to their code are billable. Google's own guidance on evaluating service providers stresses ongoing accountability over initial delivery [3]. We agree, which is why our contracts include a 90-day stabilization window at no extra cost.
Try this: before you sign, email the vendor a made-up "urgent" question on a Friday at 4pm. See when they reply. That single test tells you more than any pitch deck.
Is the cheapest quote ever the right choice?
Sometimes, yes. If the scope is genuinely simple, if the vendor has done it fifty times, if you own the output outright, then cheap is smart. But cheap on a custom build usually means the vendor is underestimating the work, and underestimated work becomes cut corners or surprise invoices. Neither ends well.
Read our welcome post for how we scope small projects honestly.
FAQ
How long should the vetting process take?
Two to three weeks is healthy. You want time for at least two conversations, a written scope, and a reference call. Anything faster and you're guessing. Anything slower and momentum dies, the internal champion loses interest, and the project quietly stalls before it starts.
What's a fair deposit to pay upfront?
Between 30% and 50% is standard for custom work in Poland. Anything above 60% shifts too much risk to you. If a partner insists on full payment upfront, walk away. A serious shop has cash flow and doesn't need your money to start.
Should I sign an NDA before the first call?
Usually no. First calls are about fit, not secrets. Save the NDA for the second conversation when actual data, workflows, or client names come up. A partner who demands one before hello is signaling either inexperience or paranoia. Neither helps you.
What if I already have an in-house developer?
Great. Bring them into the vetting calls. A good external partner welcomes technical scrutiny and will happily walk your developer through the stack. If the vendor gets defensive when your engineer asks hard questions, that's your answer.