The Framework Every Leader Should Use Before Investing in AI

If I had only 60 seconds to teach you to become an engineer, I would ensure you understand just one concept: how to solve any problem by breaking it down into smaller pieces and drawing it. As AI has entered the mainstream, I have stuck to this principle when implementing AI tools for my clients. After you read this article, you will be able to do the same for your organization or customers.

I approach every problem with a technology-agnostic lens. Over the next few years, I’m sure companies will waste millions of dollars investing in technologies first and searching for problems to solve second. By contrast, I take a problem-first approach. This means I lead my clients through my simple three-step solution design framework: the SPI Methodology.

This methodology consists of three distinct phases:

1. Strategize

2. Procure

3. Implement

Putting strategic design ahead of making purchasing decisions ensures clarity about the problem you are solving. Procuring technology after defining the ideal solution requirements means you buy the right tools. And implementing strategically with the correct tooling supports effective business transformation.

If you try to procure before your strategy is clear, you waste time and money implementing the wrong solutions. That’s why it’s critical to follow this framework in the correct order.

Strategize before you buy

Strategy in business can mean different things to different people. Under the SPI Methodology, it means defining the ideal business processes that must be facilitated by your company’s staff and systems.

You are looking for gaps in technology within your processes. It is these gaps that AI may be able to fill. If you don’t understand those gaps, it is impossible to effectively procure tools to bridge them. While AI may be a relatively new technology, this methodology is not. I have applied it in my digital transformation work well before the surge of AI innovation.

So, how do you document business processes? You do what good engineers do: Break down the processes step by step and draw a process flow diagram. Then you write plain language functional requirements for each defined step.

To give you a clear example, let’s look at a real business challenge I’ve solved using AI. This business distributes controlled goods. For every client they deal with, they must carry out a thorough check to ensure individuals and organizations are not on a sanctions or debarment list. As the business expanded, the volume of verification required became unsustainable.

This is a high-stakes process. Without mitigating this risk, they could lose their controlled goods clearance and therefore the business.

When we drew out the process, it became clear that the gap that needed bridging was the debarment verification process. Automating this process made sense. Manual verification had become a clerical nightmare for the sales staff due to the timelines and steps involved.

Thus, the functional requirement became clear: The CRM system must automatically perform debarment verifications at each critical deal step, document its findings, block progress if a potential issue is identified and send an alert when this occurs.

Procure only what you need

Once that functional requirement was well understood and agreed upon, we could procure the right tool. But just because phase two of my methodology is labeled “procure” does not mean jumping into a purchase is the right first step. It is not.

The right first step is to evaluate the technology in which you have already invested. Does it have features that could be leveraged to facilitate the newly identified functional requirement? In this instance, that was where we started.

Their existing manual tool for debarment had a more advanced integration method for automating verification queries. We set up the connection between their CRM and the tool. But quickly, we found that the tool was unreliable. It returned a significant volume of false positives and required a complex understanding of matching algorithms.

With that effort sidelined, it became clear that buying a new tool was the right path. When you are procuring technology, you must clearly define the following before you speak to any vendors:

• Your budget and timeline for implementation

• The scope of work required for the new tool (this comes from phase one)

• The stakeholders involved in the implementation

Through research, you then identify which third-party providers you want to involve in the procurement and start having discovery conversations with them.

To exit the procurement phase, you must make the best choice you can about which tool to buy. Choosing the wrong tool or partner is a risk, but not making a choice is also a failure. When change is needed, you must move forward confidently and bravely.

Implement with confidence

Most technology implementations fail out of the gate because most companies rush past the first two phases I have described. But when you have clarity around what problem needs to be solved, and you have invested time into choosing the right tools for the job, the implementation phase is significantly de-risked.

In the example I shared, we implemented an automated LLM agentic query that is tasked with deciding if any results it finds meet the threshold of potential risk. If risk is identified, it updates the CRM and notifies the sales staff immediately. The agentic approach is significantly simpler than the complex algorithmic query approach we first tried.

As you start to think about onboarding AI-first solutions, make sure you put in this critical groundwork to protect your investment of time and money. Do not chase the new shiny object because of a sense of FOMO. AI does not reduce the need for proper engineering-minded problem-solving. In fact, it makes the SPI Methodology even more necessary!

If I had only 60 seconds to teach you to become an engineer, I would ensure you understand just one concept: how to solve any problem by breaking it down into smaller pieces and drawing it. As AI has entered the mainstream, I have stuck to this principle when implementing AI tools for my clients. After you read this article, you will be able to do the same for your organization or customers.

I approach every problem with a technology-agnostic lens. Over the next few years, I’m sure companies will waste millions of dollars investing in technologies first and searching for problems to solve second. By contrast, I take a problem-first approach. This means I lead my clients through my simple three-step solution design framework: the SPI Methodology.

This methodology consists of three distinct phases: