As investments in AI ramp up, so too has the number of AI consultants. A great consultant can be a real asset, helping define valuable problems, build good tech, and enable new ways of doing work. Unfortunately, as with any new concept or tool, many self-proclaimed experts lack the skills, flexibility, or domain understanding to deliver lasting value.

Of course, the right role for a consultant depends on where an organization is and what it’s trying to achieve. Still, a few key principles apply:

1. Start with the business problem

This sounds obvious, but it’s worth restating: AI is a tool, and tools should serve a clearly defined business problem or opportunity.

Any consultant that starts with the technology, the data, or the model is already off track. Don’t get swept up in talk about “the art of the possible.” While it’s helpful to know what’s possible, too many AI firms make their money selling inspiration when they should be focused on results. The real job is moving from the art of the possible to the science of the valuable.

Many organizations skip this step because the pressure to “do something with AI” feels urgent. A strong consultant resists that pull and keeps attention on the underlying business mechanics, including how decisions are made, what metrics matter, and where value actually resides. This upfront clarity can save months of expensive iteration later.

2. Honest assessment of organizational readiness

Before any AI engagement begins, the organization needs an honest look at its readiness. Foundational investments, such as clean data, accessible infrastructure, and processes to integrate insights into daily work, must come first.

If those aren’t in place, the consultant’s first job must be to help remedy (or at least identify) that gap. Otherwise, you can end up with sparkling one-off projects that never scale.

A European grocery chain that one of us has worked with learned this the hard way. They brought in an AI consulting firm to build a sophisticated predictive forecasting model to optimize their supply chain. On paper, the model worked beautifully. But in practice, store managers questioned its predictions, and senior leaders didn’t trust the data feeding it. Ultimately, the company’s existing systems couldn’t convert forecasts into replenishment actions due to supplier contract and store layout constraints. This lack of trust, integration, and usability ultimately dealt the project a fatal blow. 

3. Look for expertise in both technology and use cases

Breadth of expertise allows a consultant to see opportunity across the organization, not just optimize one narrow slice of it. That expertise matters in two dimensions:

• Use case expertise: Consultants often double down on what they know, rolling out the same chatbot framework, predictive model, or automation pattern from one client to the next. There’s nothing inherently wrong with repetition, which can build expertise and efficiency.

The problem comes when the playbook becomes the answer, rather than a starting point. Real value comes from diagnosing where AI can make the biggest difference in your business and then tailoring proven methods to fit that reality.

• Technical expertise: The same issue applies to technology. A consultant that always reaches for the same platform or model architecture probably isn’t tailoring the solution appropriately.

Certainly, organizations should avoid a messy sprawl of tools, but they should not force every problem through the same pipeline. Success with AI deeply depends on matching the right tool to the right problem.

4. Insist on measuring ROI, even when it’s hard

There’s a growing belief that AI’s ROI is too complex to measure. For sweeping transformations, that may be true. Benefits often accrue across systems, processes, and people in ways that are hard to estimate.

But at the project or process level, measurable ROI is essential. A good consultant helps define reasonable, testable expectations for both costs and benefits. Every model, automation, or workflow improvement should have a hypothesis about value that can later be evaluated.

The grocery chain’s ROI for the forecasting project looked extraordinary on paper. But when they couldn’t operationalize it, the real value disappeared. 

Being ROI-focused has two additional benefits: First, it can be used as an accountability mechanism to ensure that AI remains a business tool, not a technical showcase.  Second, it can set the foundation for fair pricing, given that AI consultants should be rewarded for the value they create, not the hours they bill or tasks they complete.

The bottom line

The best AI consultants build systems that work in the real conditions of your organization. They understand your data, your people, and your decision processes well enough to turn AI into an advantage. The rest is expensive theater, leaving you with little that truly changes how your business performs.

About the Authors:
Joel Shapiro, J.D., Ph.D., is a professor of Managerial Economics and Decision Sciences at Northwestern University’s Kellogg School of Management. He leads Kellogg’s Analytical Consulting Lab, one of the school’s flagship experiential courses, and teaches MBA and executive programs on data and commercial impact.

He also served as Chief Analytics Officer at a leading sales software company, where he helped guide the organization through a period of high-growth transformation and a successful exit. He is the founder of The Evidence-Driven Enterprise, a boutique consultancy that helps organizations build the tools, data infrastructure, and leadership mindsets required to make data valuable. Shapiro is a frequent advisor to senior teams, a trusted speaker, and is widely quoted in the business press for his pragmatic take on turning data into results.

Udheep Sai Cherukuri is an MBA candidate at the Kellogg School of Management, where he explores the intersection of business leadership and emerging technologies and serves as Vice President of the Kellogg AI Club. Before Kellogg, Cherukuri spent over a decade leading AI transformation initiatives across Fortune 500 enterprises and global startups, specializing in translating complex business challenges into scalable AI and machine learning solutions.