← Back to Blog
AI10 min read

How to Choose an AI Consultant: 7 Questions Every Business Should Ask

NDN Analytics TeamApril 12, 2026

The AI consulting market is crowded with generalists who claim expertise in everything. Choosing the wrong partner is expensive — bad AI projects commonly waste $100K–$500K before anyone admits failure.


Here are the seven questions that separate genuine AI expertise from sales-driven hype.


1. "Can you show me a deployed system, not a demo?"


Demos are engineered to impress. Production systems are engineered to work. Any credible AI consultant should be able to reference actual deployed systems with measurable outcomes — not just proof-of-concept models trained on public datasets.


**What to look for:** Live case studies with verifiable metrics. Client names you can call for a reference. Revenue or efficiency numbers you can validate.


**Red flag:** "We can't share client details" as a blanket answer. While NDA restrictions are real, reputable consultants have at least one reference they can share.


2. "What percentage of your AI projects reach production?"


This is the most important question and the one most consultants dread. Industry-wide, ~85% of enterprise AI projects fail to reach production. The best consultancies run at 60-80% production success rates — still not perfect, but dramatically better than average.


**What to look for:** Honest acknowledgment of failed projects with clear diagnosis of why. Consultants who claim 100% success are either lying or only doing trivial projects.


**Red flag:** Vague answers about "learning experiences" without specific data on deployment rates.


3. "Who owns the model and the data pipeline after the engagement?"


This is a business and legal question as much as a technical one. Some consultancies build on proprietary platforms that create vendor lock-in. Once they leave, you can't maintain or improve the system without them.


**What to look for:** Open-source model frameworks (scikit-learn, PyTorch, Hugging Face) and cloud-native pipelines (GCP Vertex AI, AWS SageMaker, Azure ML) that your team can take over.


**Red flag:** Proprietary AI "platforms" with licensing fees that scale with your usage. You're renting capability, not building it.


4. "What does your team actually look like?"


AI consulting requires a multi-disciplinary team: data engineers (to build pipelines), data scientists (to develop models), ML engineers (to deploy and monitor), and domain experts (who understand your industry). A team of generalist "AI strategists" can produce a roadmap but not a working system.


**What to look for:** Named team members with verifiable LinkedIn profiles and relevant technical backgrounds. Evidence of cloud certifications (GCP Professional ML Engineer, AWS ML Specialty).


**Red flag:** A team whose bios are full of strategy and MBA language with no technical specifics.


5. "How do you handle model monitoring and drift?"


AI models degrade over time as real-world data changes. A churn prediction model trained on pre-2024 data may perform poorly today. Professional AI teams build monitoring pipelines that detect when models need retraining.


**What to look for:** Specific monitoring tools mentioned (MLflow, Weights & Biases, Evidently AI). Clear policy on model retraining cadence. SLAs on model accuracy thresholds.


**Red flag:** "We'll set it up and hand it off" with no mention of ongoing monitoring. An AI system without monitoring is a liability.


6. "What's your approach to data privacy and security?"


This is non-negotiable in regulated industries. AI systems ingest sensitive data — customer records, financial transactions, health information. Security must be designed in, not bolted on.


**What to look for:** SOC 2 compliance or equivalent. Specific answers about encryption at rest and in transit, access controls, and data residency. HIPAA BAA willingness for healthcare projects.


**Red flag:** Security treated as an afterthought or a "we'll figure it out during implementation" approach.


7. "What does success look like at 6 months and 12 months?"


Vague success criteria lead to vague outcomes. A credible AI consultant will define specific, measurable outcomes upfront — not in the SOW boilerplate, but in conversation.


**What to look for:** Named KPIs specific to your business context. A baseline measurement methodology (you need to know where you started to prove you improved). Milestone-based payment structures tied to outcomes.


**Red flag:** Projects priced purely by time and materials with no outcome accountability.


How NDN Analytics Answers These Questions


We're not going to claim perfection — but we do have direct answers to every question above:


  • Deployed systems: NDN TraceChain in pharmaceutical supply chains, NDN Demand IQ in retail, NDN Care Predict in healthcare
  • Production rate: 70% of our projects reach production; 20% are redirected during assessment when we identify blocking constraints
  • Model ownership: 100% open-source, cloud-native — you own everything
  • Team: Named engineers with GCP and AWS certifications, available to reference
  • Monitoring: We include MLflow-based monitoring in all production deployments
  • Security: SOC 2 Type II process, HIPAA BAA available, GDPR-compliant architectures
  • Success metrics: Defined in the initial assessment, tracked quarterly

    • The best way to evaluate us is to start with the AI Readiness Assessment. It's a 2-hour working session — you'll know within an hour whether our approach matches your needs.


    Start with a free AI Readiness discovery call and we'll map a clear AI roadmap with you.

    Need Help Implementing AI/Blockchain Solutions?

    NDN Analytics specializes in enterprise AI and blockchain implementation. Our team can help you integrate cutting-edge technology into your existing workflows.