How to Evaluate an AI Consultant in 2026

TL;DR: Most mid-market businesses choosing an AI consultant look for the flashiest demos and the most impressive portfolio. Wrong criteria. The right criteria are: does the consultant understand your specific operational gaps, do they have a track record of implementations that stick, are they honest about what takes time and what takes money, and do they design implementations that produce sustainable outcomes rather than impressive launches. This post breaks down what to actually evaluate when you are choosing an AI consultant in 2026.

The Evaluation Criteria That Mislead Most Businesses

When a mid-market business starts looking for an AI consultant, there is a standard set of criteria that almost every business uses to evaluate options. Those criteria are almost entirely wrong.

The first criterion is usually the portfolio. Show me the impressive implementations you have done. Show me companies like ours that you have worked with. The implicit assumption is that if a consultant has done impressive work for similar companies, they will do impressive work for you.

The second is usually the breadth of capability. Can you handle everything. Can you do data infrastructure and workflow design and AI systems and change management all internally. The implicit assumption is that a full-service consultant is more valuable than one that specializes.

The third is usually the confidence in the pitch. The consultant that walks in and quickly proposes a solution is more competent than the one that spends weeks asking clarifying questions. The implicit assumption is that faster is better and that a consultant who knows the answer already is more capable than one that needs to learn your specific situation.

All three of these assumptions produce bad consultant choices. And there are good reasons for each assumption to be wrong.

Portfolio and past work tell you what a consultant has done. They do not tell you whether those implementations stick, whether they produced sustainable outcome improvements, or whether the consultant has a track record with mid-market businesses specifically. Impressive portfolios often hide the implementations that failed quietly, because failed projects do not make good portfolio examples.

Breadth of capability sounds valuable until you realize that full-service consultants have built their model on selling more services, not on designing the most focused implementation. Breadth often means less depth. A consultant that specializes in the specific area you need and partners with specialists in other areas is often more valuable than a generalist.

Speed and confidence in the pitch sound like competence. They are usually a sign that the consultant is proposing a solution based on how they normally work, not on what your specific situation requires. The consultant that asks the most questions is usually the one that understands that every business is different and that getting the diagnosis right takes more time than the sales process wants to spend.

What to Actually Evaluate

If the standard criteria are wrong, what should you actually look for when evaluating an AI consultant. Six things, specifically.

1. Operational Diagnosis Before Solution Prescription

The best indicator of a good consultant is whether they insist on understanding your specific operational situation before proposing a solution.

What this means: The consultant should want to do a readiness assessment or operational audit before scoping implementation. Not a cursory one. A real one. They should spend time with the people who run your most important workflows. They should ask questions that make you think, not questions that are just gathering information they have heard before.

If a consultant proposes a solution in the first few meetings without doing genuine diagnostic work, that is a red flag. It means they are working from a template, not from your reality.

2. Honest About Sequencing and Timing

The second best indicator is whether the consultant is honest about what takes time and what takes money.

What this means: Real AI infrastructure takes time to build. The consultant that promises to have you AI-native in three months is selling you something that is not actually AI-native. The consultant that is honest about a 12 to 18 month roadmap, but is clear about what gets done first and what the early wins are, is being realistic.

The consultant that insists that data quality does not matter because modern AI is robust is not being honest. The consultant that explains why data quality is a prerequisite and what fixing it requires is being honest.

3. A Track Record Specific to Mid-Market

Your situation is different from enterprise, and it is different from startups. A consultant that has primarily worked with enterprise clients may understand sophistication but not constraints. A consultant that has primarily worked with startups may understand growth but not operating at scale.

What this means: Look specifically at the consultant's track record with mid-market businesses. How many. What industries. What are the outcomes. The consultant should be able to name projects, results, and lessons learned specifically from mid-market experience.

4. Implementations That Stick

The most important question is whether past implementations have actually stuck. Not whether they launched successfully, but whether the organization is still using the AI infrastructure two years later, whether it is still producing value, and whether it is still being refined.

What this means: Ask specifically about retention and sustainability. How many clients are still using what was built. How often do the systems get adjusted based on new learning. Are there patterns of initial enthusiasm followed by quiet abandonment.

The consultant that is honest about this is valuable. They should be able to explain which implementations stuck and why, and which ones did not and why. The gap between those two categories tells you what the consultant has learned about what works.

5. Clear About Their Own Limitations

No consultant is genuinely full-service. A good consultant is honest about what they do well and what they partner on.

What this means: Listen for where the consultant says "we bring in a partner for this" or "this is not our core competency." A consultant that claims to do everything is not being honest with you. A consultant that is clear about what is in their sweet spot and what is not is being realistic about how to structure implementation.

6. Focused on Outcomes, Not Activities

The final criterion is whether the consultant measures success by what gets done or by what the business actually achieves.

What this means: The consultant should be talking about outcomes. Cycle time improvement, error reduction, cost avoidance, decision quality. Not about activities. How many models were trained, how many integrations were built, how many dashboards were designed. Those are activities. Outcomes are what actually changed about how the business operates.

If the consultant is focused on activities, they will optimize for visible delivery. If they are focused on outcomes, they will optimize for organizational impact, which is usually what actually matters.

What a Consultant Should Ask In Initial Conversations

You can evaluate a consultant's quality by the questions they ask. The good ones ask harder questions.

Poor consultant questions: What tools are you currently using. What is your IT infrastructure. What is your current AI maturity.

Good consultant questions: Walk me through a high-priority workflow in detail, including how it actually works today and what the pain points are. What has changed about how you operate in the last two years and why. When you have tried to implement operational change in the past, what has worked and what has not. What decisions are currently bottlenecked on specific people. What information do you wish you had visibility into that you don't currently have.

The good questions force you to think. They also give you insight into how the consultant thinks about problems. A consultant that asks about your workflow tells you they care about your specific reality. A consultant that asks about your change history tells you they understand that implementation is as much organizational as it is technical. A consultant that asks about bottlenecked decisions tells you they understand where real value concentrates.

Red Flags to Watch For

Three things should make you skeptical of a consultant choice.

Proposing technology before understanding the workflow. If the consultant's first instinct is to recommend a specific platform or tool, they are working from a sales template, not from diagnosis. Real consultants understand the workflow first, then pick the technology based on what fits.

Selling speed as a feature. "We can have you AI-native in four months" is not realistic. "Here is what we can realistically accomplish in different timeframes" is honest. Beware the consultant selling speed. You will get implementations that move fast and fail slow.

Blaming the client when implementations do not work. The consultant that says "your organization was not ready" is either making excuses or was not competent enough to recognize readiness gaps before proposing implementation. A good consultant understands organizational readiness as part of the diagnostic work.

Questions to Ask References

If a consultant gives you references, ask specific questions.

Not: Did this consultant do good work.

Instead: Does the implementation still produce value two years later. What would you have done differently. What was harder than expected. What parts of the roadmap changed and why. Would you hire them again for the same type of work.

The reference that gives you honest feedback, including what did not go perfectly, is more valuable than the one that only praises. It means both the consultant and the reference are being real about what happened.

Frequently Asked Questions

Should we choose based on cost, or is cheaper always a red flag.

Cost matters, but it is not the primary criterion. A consultant that is significantly cheaper than comparable options is often cheaper because they are cutting corners on diagnosis, or planning for a faster timeline that is not realistic. That said, paying more does not guarantee better results. Look at what the cost covers. Is it for genuine diagnostic work and detailed planning, or is it for impressive sales and marketing. The first justifies cost. The second does not.

Should we hire a generalist or a specialist.

A specialist in the area you need most help with is usually more valuable. A good specialist will partner with other specialists for the domains outside their core competency. That partnership model usually produces better outcomes than a generalist trying to handle everything internally.

What if we already have an implementation partner and are unhappy with progress.

The conversation to have is whether the unhappiness is about the partner's competency or about the roadmap being realistic. If the roadmap was not realistic upfront, even a great partner will struggle. If the roadmap was realistic but the partner is not executing against it, that is a different problem. Having a readiness assessment or operational audit done by a third party can help diagnose which situation you are actually in.

How do we know if the consultant's timeline estimate is realistic.

Realistic timelines account for discovery, organizational change, and unexpected gaps that almost always surface during implementation. They are longer than optimistic timelines. They also have built in flexibility, with review points where learning can inform adjustments. If the timeline feels aggressive, it probably is.

Should we do a pilot project first, or go straight to full implementation.

A pilot project is valuable if it is genuinely treating the workflow as a learning opportunity, and the pilot is designed to surface gaps and inform the full implementation. It is a waste if it is designed to prove the consultant's approach works, because that is not the point. The point is to learn what your specific situation requires. If the consultant is resistant to a pilot as a learning tool, that is a red flag.

The Bottom Line

Choosing an AI consultant is choosing a partner for a 12 to 18 month journey. The criteria that matter are not the most impressive portfolio or the fastest pitch or the broadest capability. They are: understanding of your specific situation, honesty about sequencing and constraints, track record in mid-market, implementations that actually stick, clarity about limitations, and focus on outcomes over activities.

The consultant that gets those right will spend more time asking questions than talking about their solution. They will be honest about what takes time and what you can expect along the way. They will be able to point to sustainable implementations and explain why those worked and why others did not.

That consultant is worth investing in. The ones that are slick and impressive upfront are often the ones that deliver disappointment downstream.

Team at Navon works with mid-market businesses through the entire AI implementation journey, from readiness assessment through sustainable operations at scale. Start the conversation.