Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

4 principles to apply AI in business

Use VURA to find real business value.

[Photo: Getty Images]

Profile photo of Andy MacMillan

CEO · Alteryx

We’re a few years into the grand AI experiment, with mixed results. Almost 90% of businesses use AI in some way, and two-thirds believe AI is driving innovation. On the other hand, 90% of business executives say that AI hasn’t improved their productivity at all.  

Statistically speaking, your company could probably be getting more out of AI. 

Let’s assume you already have the right tools and staff in-house. If AI feels like more of a side project than a core focus, revamp your process, because you need your AI tools to turn raw data into actionable outputs every time. 

Think about a restaurant that automates its weekly orders, saving time and getting the freshest ingredients. Or a law firm that can pull up relevant casework in minutes instead of days. Or a logistics company that optimizes shipping costs based on seasonal sales spikes. That’s how organizations become truly intelligent enterprises. 

To understand what an intelligent enterprise is, consider the 10% of business executives who have used AI to boost productivity. These companies incorporate AI into every workflow where it makes sense, and give every employee opportunities to evaluate the results. Over time, the tools improve, the staff upskills, and the virtuous cycle repeats.  

Build an intelligent enterprise 

At Alteryx, I created an internal framework of guidelines to ensure our AI inputs and outputs are consistent and accurate. Intelligent enterprises use AI tools and guidelines to become more adaptable, innovative, and efficient. I call this VURA: 

  • Visible 
  • Understandable 
  • Repeatable 
  • Auditable 

Picture this: A finance team wants to build an AI agent to handle tax reconciliation, which already has a defined process with data, for potential enormous time savings. However, when the CFO must legally sign the tax documents, and the team can’t explain the agent’s conclusions, guarantee consistency, or provide an audit trail for regulators, it becomes a trust problem. It’s why the four VURA principles must be in place before a project like this moves forward.  

Thanks to this nonnegotiable standard, we can build transparent, practical AI systems using  

guidelines that build trust in both our data and our outcomes. You’ll also need that trust if you want your business to evolve into an intelligent enterprise. 

The VURA framework 

Visible 

Good AI tools should tell you where their data comes from and how they interpret that data at every step. If your AI systems can provide this level of clarity, they’re visible. 

If not, focus on increasing your data’s transparency. An IT administrator can edit file permissions, making specific folders available to your AI tools while restricting others. Your staff should also make their prompts as clear as possible. Specify the exact data the AI should analyze for each query, as well as the target audience and the project’s overall goals. If your AI’s answers are still vague or obscure, consider different tools.       

Every employee using AI should be able to answer two simple questions: “Where did this come from?” and “How did we draw that conclusion?”  

Otherwise, they could be delivering bad information and never even know it. AIs can still miscount inventories, create imaginary documents, or make bizarre predictions.  

AI processes should also be visible across departments, especially when people from different teams are collaborating on the same project. 

Understandable 

One reason AI took off so quickly is its ability to interact in natural language. People can ask questions in plain English and get answers the same way. That’s the “understandable” part of the VURA framework. No one should need a technical background to understand AI outputs.  

You can ask tools like ChatGPT and Google Gemini complicated questions and still get simple responses. Your enterprise AI systems should also be understandable when they explain the logic behind their answers. 

Repeatable 

If your AI gives you some brilliant outputs and some ridiculous outputs, you can’t trust it. Every AI process—from prompts used to results received—at your company must be repeatable.  

After discovering an effective way to use an AI tool, employees should document and evaluate it before sharing it with everyone. That’s important if the staff member isn’t an expert in a given field.  

At Alteryx, for example, we use ChatGPT alongside Alteryx to analyze Salesforce pipeline data: Alteryx pulls, cleans, and structures the data so ChatGPT can analyze it and generate insights. We’ve even used ChatGPT to build the Alteryx workflow itself. The entire process is codified, and any team member can run it consistently every quarter. 

A good AI tool must work properly every time, especially when undertaking changes to important code or legal documents. Have subject-matter experts test new AI processes before making them widely available.   

Auditable 

Having real people involved at every stage is a core VURA framework principle. Upon receiving a report that doesn’t make sense, an employee should be able to see who’s in charge of the project, what prompts were used, and the AI’s chain of logic.  

Every AI tool in your organization has to be auditable. This way, both the AI tools and the people who use them are accountable for their work.  

Raw data, actionable outcomes 

Every part of the VURA framework makes AI systems more effective. Visible tools show where data comes from. Understandable tools put AI in everyone’s hands. Repeatable tools build trust in your outputs. Auditable tools prevent costly mistakes. Put these principles together, and you have a foundation for an intelligent enterprise. 

If your AI systems still feel like experiments rather than productivity tools, it’s time to take a closer look at your data and processes. Every department should provide high-quality data, and every team should experiment with AI workflows. Keep what works, discard what doesn’t, and share progress openly across the organization.  

Just incorporating AI tools into your operations isn’t the same thing as building an intelligent enterprise, as 90% of businesses are now learning the hard way. Becoming part of the 10% won’t happen overnight, but this framework will help you get there. 

 
Andy MacMillan is the CEO of Alteryx.