There is a predictable pattern to failed AI implementations. The organization identifies a problem — or feels pressure to "do something with AI." They evaluate vendors. They select a tool. They deploy it. Six months later, adoption is low, the ROI is unclear, and nobody is quite sure what went wrong.
What went wrong, almost always, is that nobody asked the right questions before the decision was made. Not the vendor questions — those get asked. The internal questions. The ones about people, processes, and what the organization is actually trying to achieve.
Here are the questions we ask every client before any technology enters the conversation.
Question 01
Are you solving a people problem, a process problem, or a technology problem?
This sounds obvious. It isn't. Most organizations reach for AI when the underlying issue is that a process is broken, a team is understaffed, or institutional knowledge is held by people who are leaving. AI will not fix a broken process — it will automate the broken process and make it faster. Getting this diagnosis right before selecting any tool is the most important thing you can do.
Question 02
Where does human judgment actually matter in this workflow?
Not where you think it matters — where it actually matters. The distinction is important. Many organizations protect human judgment in places where it is genuinely just habit, while exposing it to automation in places where it is genuinely critical. Mapping this before you deploy means you design the system correctly from the start rather than discovering the failure mode in production.
Question 03
What institutional knowledge lives only in people's heads — and what happens if those people leave?
This is the question that reveals the most about an organization's actual AI readiness. If the answer is "a lot," then the first priority is capturing that knowledge, not automating it. An AI system trained on documented processes cannot compensate for undocumented expertise. The capture has to come first.
Question 04
Who in your organization will be most affected — and have you spoken with them yet?
The people closest to the work you are trying to automate or augment almost always know things that leadership doesn't. They know where the edge cases are. They know which parts of the process look simple but aren't. And they will tell you, if you ask before the decision is made rather than after. Organizations that include frontline employees in AI planning have significantly higher adoption rates. This is not a coincidence.
Question 05
What does your data actually look like — and who controls it?
AI systems are only as good as the data they learn from or operate on. Most organizations significantly overestimate the quality and accessibility of their own data. Before evaluating any vendor, you need an honest picture of what you have, where it lives, who can access it, what its gaps are, and what the privacy and compliance implications are of using it. This audit routinely surfaces surprises that change the entire direction of the implementation.
Question 06
What would a failed implementation actually cost you — beyond the contract value?
Vendor proposals quantify the upside. They rarely quantify the downside. A failed implementation costs the direct contract value, the internal time spent on integration and training, the opportunity cost of the months spent on something that didn't work, and — most expensively — the erosion of employee trust in future technology initiatives. That last one is the hardest to recover from.
Question 07
If this works exactly as planned, what will you do with the capacity you free up?
This question separates organizations that are serious about AI from those that are responding to pressure. If the answer is "we'll figure that out later," the implementation will likely underdeliver — not because the technology fails, but because the organization hasn't decided what success actually looks like. If the answer is specific and concrete, that specificity should drive every design decision in the implementation.
The pattern these questions reveal
When we work through these questions with clients, a pattern almost always emerges: the organizations that ask them thoroughly before selecting a tool end up with narrower, more specific implementations that actually work. The organizations that skip them end up with broad deployments that generate enthusiasm at launch and quiet abandonment six months later.
The questions are not a delay tactic. They are the work. The technology selection is the easy part.