Before You Hire an AI Consultant, Answer These Five Questions

Before You Hire an AI Consultant, Answer These Five Questions

Hiring an AI consultant has become the new "we need a website" — a reflex, not a decision. Across my work with SMEs in Vietnam, Singapore, and the UAE, the most expensive mistake I see founders make is not choosing the wrong consultant. It is walking into the engagement without knowing what they actually need. This AI consultant SME checklist exists because that gap costs real money and real months.

The Real Problem Is Not the AI

Before we get to the five questions, one frame to carry: AI consultants are problem solvers, not problem finders. The best ones will push back and help you sharpen your brief. Most will not. They will take your budget, run a discovery sprint, and return a roadmap that sounds credible but is built on a brief you gave them before you had clarity.

The diagnostic below is designed to give you that clarity first.

---

Question 1: What specific decision or action do I want to make faster or better?

Not "I want to use AI in my operations." Not "I want to automate." Those are categories, not problems.

A furniture manufacturer in Ho Chi Minh City came to me asking about AI for customer experience. After thirty minutes of conversation, the real ask emerged: their sales team was losing deals because quote turnaround took four days. The solution was a configuration and pricing tool that took quotes from four days to four hours. It had an AI component, but the problem was a process bottleneck, not a customer experience gap.

If you cannot write the problem in one sentence, you are not ready for a consultant. You are ready for a working session with your own team.

---

Question 2: What data do I already own, and how clean is it?

AI runs on data. Not aspirational data. Not data you plan to collect. Data you have right now, in a structured, accessible form.

Ask yourself three things. First, where does this data live — CRM, spreadsheets, a POS system, or someone's head? Second, how much history do you have? A meaningful pattern usually needs at minimum twelve to eighteen months of consistent records. Third, who in your organization understands this data well enough to QA an AI output?

In Singapore, I watched a retail group spend three months and a significant retainer on a demand forecasting model, only to discover mid-engagement that their inventory data across three locations used inconsistent SKU codes. The model could not train. The engagement stalled. The data problem was discoverable in week one — if anyone had asked.

---

Question 3: Which single process, if improved, would create the most visible business impact?

Scope kills AI projects. The instinct is to think big: "Let's transform the whole customer journey." The execution reality is that transformation happens one process at a time.

Pick one. Map it from input to output. Identify where the time goes, where the errors happen, where a human is making a judgment call that a well-trained model might handle more consistently. That map is your starting brief.

Consultants who see a clear process map trust you more, scope more accurately, and deliver faster. Consultants who see a vague ambition either over-promise or pad discovery phases. Both outcomes drain your runway.

---

Question 4: Do I have an internal owner, or am I outsourcing the thinking entirely?

This question separates sustainable AI adoption from expensive dependency.

An internal owner does not need to be a data scientist. They need to understand the business problem deeply, be able to communicate with a technical team, and have the authority to make decisions about process changes. In SME terms, this is often an operations lead, a head of growth, or a sharp founder who is willing to stay close to the engagement.

I have seen strong AI implementations in mid-size companies in Dubai and Hanoi that had no in-house engineers. What they had was one sharp operator who sat in every session, questioned every assumption, and owned the rollout. And I have seen well-resourced implementations fail because the founder handed the brief to a consultant and came back three months later expecting a finished product.

Outsource the execution. Do not outsource the judgment.

---

Question 5: What does success look like in ninety days, and how will I measure it?

If you cannot define a ninety-day success metric before signing a contract, your consultant has no accountability and you have no signal.

The metric does not need to be revenue. It can be: quote turnaround time drops from four days to one. Customer support ticket resolution rate improves by twenty percent. Manual data entry hours per week drop from forty to ten. These are specific, measurable, and achievable within a short engagement window.

A defined metric also protects you from scope expansion. When new ideas emerge mid-engagement, and they always do, you have a shared reference point. Is this new idea going to help us hit the ninety-day metric? If yes, consider it. If no, log it for the next phase.

---

Why This Matters More for SMEs Than for Enterprises

Large companies can absorb a failed AI pilot. They have dedicated innovation budgets, benched technical talent, and enough organizational mass to run parallel experiments.

You do not. For an SME, a poorly scoped AI engagement is not a learning expense. It is an opportunity cost that delays a real solution by six to twelve months, strains a key team relationship, and sometimes creates AI skepticism that is hard to reverse.

The AI consultant SME checklist above is not about slowing down. It is about entering the right conversation faster, with more leverage, and with a clearer standard for what done looks like.

---

What to Do Next

One: Run the five questions as a sixty-minute internal session before any vendor call. Bring your operations lead or whoever owns the process you want to improve. Write one-sentence answers to each question. Where you cannot answer, that is your real starting point.

Two: Map the one process you want to improve, end to end, on a whiteboard or a shared doc. Input, steps, decision points, output, time per step. A consultant who receives this document will give you a sharper proposal and a more realistic timeline.

Three: Define your ninety-day metric before you issue a brief. Write it in the language of your business, not in AI terminology. "Reduce X from Y to Z by this date" is a complete metric. Bring that number into the first vendor conversation and watch how differently the room responds.

The right AI consultant will welcome all five answers. The wrong one will tell you not to worry about it — they will figure it out in discovery. That response is your signal to keep looking.