Know what you are buying
AI automation projects combine three things: process design, software integration and model behaviour. A strong partner is good at all three, and can explain trade-offs in plain language.
Questions to ask every vendor
Ask for specific answers, not reassurance:
- How will you measure accuracy before and after launch?
- What happens when the automation is unsure?
- Where does our data go, and who can access it?
- Who owns the code, prompts and configuration?
- What does support look like after go-live?
Red flags
Be wary of fixed promises about percentages of work removed before anyone has looked at your process, platforms that lock your data in, and proposals without a plan for monitoring.
A good first 90 days
Weeks one and two map the process and agree on success measures. Weeks three to six build and test one workflow with real cases. The rest of the quarter runs it with human approval, tunes it, and decides what to automate next based on results.
Want help applying this to your team? Book a free AI workflow audit or talk to us about your project.
