An AI automation agency designs and builds AI agents and automated workflows for your business. Good agencies start with your process, not with a tool. Use the questions below to compare partners.
10 questions to ask
1. Which of your projects are live today?
Ask for running systems, not demos. A live product shows the agency handles real data, real users and real failures. For reference, our live products include the AI HR Portal and NewsAgent.
2. How do you scope the project?
Expect a discovery session mapping your current workflow step by step. A fixed quote before anyone looks at your process is a warning sign.
3. Which systems will you connect to?
A good agency names your CRM, ERP, WhatsApp, email and databases, and explains how each connection works.
4. Where does a person stay in control?
Ask where the system asks for human approval. Look for confirmation steps before data changes or messages go out.
5. How do you handle wrong answers?
Ask about lookups through tools, validation rules and review steps. "The model is accurate" is not an answer.
6. Who owns the code and data?
Get written confirmation: you own the code, the prompts, the data and the accounts.
7. Where does my data go?
Ask which AI providers process your data, in which region, and whether self-hosting is an option.
8. How do you control AI costs?
Ask for an estimate of monthly AI usage costs and the limits in place: daily caps, caching, model choice per task.
9. What happens after launch?
Ask about monitoring, fixes, model updates and response times for support.
10. How do you measure success?
Agree on numbers before work starts: hours saved, response time, resolution rate or error rate.
Red flags
- Promises of "full automation" with no human review.
- No live projects, only demos and slide decks.
- Vague answers on data storage and ownership.
- Lock-in to a proprietary platform with no exit path.
- Pricing with no breakdown of build, AI usage and support.
Pricing models
| Model | How the model works | Best for |
|---|---|---|
| Fixed project price | One price for a defined scope | Clear, well-mapped workflows |
| Monthly retainer | Ongoing builds and support for a monthly fee | Several workflows over time |
| Build plus support | One-off build, then a smaller monthly fee | A single system needing upkeep |
Whatever the model, ask for three costs in writing: the build, AI and hosting usage, and support.
What a good proposal includes
- Your current workflow, mapped step by step.
- The future workflow, with every AI step and human check.
- Systems to connect, with access requirements.
- Guardrails and data handling.
- A timeline with milestones.
- Success metrics and how to track them.
- Costs split into build, usage and support.
How zdevs works
We start every project with a free discovery call and a mapped workflow. We build with human approval where approval matters, connect to the tools you already use and hand over full ownership. See our AI automation and AI agents services, or check pricing.