Most AI projects that fail don't fail because the technology was wrong. They fail because the wrong questions were asked at the start — or no questions were asked at all. Someone heard about AI, picked a vendor, ran a pilot, and six months later there's a half-built model nobody uses.
Before you commit budget, headcount, or your team's attention to an AI initiative, work through these five questions. They're the same ones our consultants ask in the first 30 minutes of every engagement — and they've saved more than one client from an expensive mistake.
The failure rate for AI projects is well-documented. Gartner research has consistently found that a significant proportion of AI pilots never make it to production — not because the technology doesn't work, but because the business case was unclear from the start. These five questions exist to change that.
Question 1: What problem are we actually trying to solve?
What's the specific business problem — not the tool we want to use?
This sounds obvious, but it's where most AI investments go wrong. "We want to use AI" is not a problem. "Our sales team spends 4 hours a day on manual proposal generation, and the bottleneck is costing us 20 deals a quarter" — that is a problem.
If you can't state the problem in concrete operational terms — with numbers — you're not ready to invest yet. Go back and quantify it first. If the problem is proposal generation, for instance, that's a well-defined process AI can compress significantly — tools like Drafted exist precisely because the problem was specific enough to solve.
Question 2: What's the actual value at stake?
If we solve this perfectly, what's the dollar impact in the next 12 months?
Multiply your estimated savings or revenue gain by a probability of success (be honest — pilots fail 40-60% of the time even when well-run). If that number isn't at least 3–5× the projected cost of the project, the math doesn't work — and no amount of AI sophistication will fix it.
Common values worth measuring:
- Hours saved × loaded labor cost
- Errors prevented × cost per error
- Throughput increase × revenue per unit
- Customer churn reduction × LTV
- Time-to-decision improvement × deal velocity
Question 3: Do we have the data?
Do we have enough high-quality, accessible data — or will we need to build it?
This is where projects quietly die. The AI is the easy part in 2025. The hard part is the data: is it clean, is it accessible, is there enough of it, and is your team allowed to use it?
Before greenlighting an AI project, walk through:
- Where does the data live — and who controls it?
- How clean is it, honestly? When was it last audited?
- Are there compliance or privacy constraints on its use?
- If we don't have it, how long would it take to collect?
If the answer to "do we have the data" is "kind of," budget 30-50% more time and cost than the vendor quotes you. That's not pessimism — it's the average.
Question 4: Will our people actually use it?
Who needs to change how they work for this to succeed — and have we talked to them?
The most technically impressive AI system in the world is worth zero dollars if the people it's meant to help refuse to use it. Adoption is the #1 reason AI projects underperform — not model accuracy.
Before investing, identify every role that has to change behavior for the project to deliver value. Then talk to those people. Find out:
- Do they see the current process as broken?
- What did past technology rollouts feel like for them?
- What would have to be true for them to trust the new system?
- Who are the natural champions — and who are the likely skeptics?
If you discover the answer is "they're going to hate this," fix that before you write any code. One pattern that consistently improves adoption: start with AI that fits into how people already communicate rather than forcing a new workflow. Speakeasy is a good example of this approach for client-facing teams.
Question 5: How will we know if it worked?
What does success look like at 30, 90, and 180 days — in measurable terms?
"It seems to be working" is not a success metric. Before kickoff, write down:
- 30-day goal: A working pilot in one team or function, with baseline metrics captured
- 90-day goal: Measurable improvement in your target metric, with clear comparison to baseline
- 180-day goal: Sustained improvement + path to scale across the business
If you can't articulate these milestones, you don't have a project — you have a science experiment. Both can be valuable. Just be honest about which one you're funding.
If you can answer all five questions clearly and the math still works, you're ready to invest. If you can't, the answer isn't "no AI" — it's "not yet."
How we work through this with clients
In a typical first engagement, we spend a week working through these exact questions with stakeholders. Sometimes the answer is "yes, here's the path forward." Sometimes the answer is "AI isn't the right fix yet — here's what to do first." Either outcome saves a lot more money than skipping the questions.
If you're staring at an AI initiative and not sure whether to greenlight it, we can help you work through the framework in 30 minutes. Book a free strategy call and we'll talk through your situation specifically.
Want help applying this framework?
Free 30-minute strategy call. We'll work through these 5 questions for your specific situation and give you a clear go/no-go recommendation.
Book a Free Call