How We Think About Tech
Tech Philosophy
Data science and AI projects are not one size fits all. Many companies have similar use cases but very different tech stacks and levels of knowledge. This leads to different implementations and solutions. The goal of new technology isn’t to be difficult or disrupt current working processes. It should supplement, enhance, and make things easier.
The success of new technology projects is defined by the value they create. If the goal is too broad or ambiguous, it becomes difficult to determine whether it was a success. It all comes down to measurable impact by understanding exactly what you are trying to solve. The scope should be clearly defined and the impact measurable.
Start Small
When implementing new technology, start with a well-defined purpose and use case. These initial use cases should be small and address an existing pain point or missing functionality. Make sure you can quantify the value of the project.
Bad Example: “I want to build a chatbot for the sales team to use.”
Good Example: “I want to build a chatbot that the sales team can use to practice conversations with buyers.”
-
If we look at the two examples above, both aim to create chatbots and leverage new AI technology, but the first scope is too broad and difficult to measure. It doesn’t provide clear direction on what the new technology will solve. This leads to questions about the value provided, whether monetarily or in time and resources. It is also tough to understand if the chatbot is working well and what should be improved.
The second example has a clearly defined scope. The chatbot will act as a training aid for the sales team to practice conversations with buyers. The outcome can be measured by sell-ins or other sales metrics, and there is a clear path to evaluating the tool. This tighter scope makes it clear what to test and how to evaluate whether the chatbot is working as intended.
Build Fast
You should build quickly and show progress along the way. New technology shouldn’t take months to implement before you learn anything. You should be able to quickly determine whether the technology will solve the issue at hand. This means fast builds, constant updates, and clear targets.
The two pieces here are to build quickly to prove value and to provide updates along the way. Let’s take the example of wanting to build a model to predict equipment failures. You have equipment that provides hourly readings from a number of different customers. This equipment periodically breaks, and being able to predict and remediate issues in a timely manner would be a big win for the company. The team is then tasked with building a model to predict broken equipment. Building a full-blown model may take 2–3 months, but that doesn’t mean value can only come after finishing the model. By breaking the larger effort into smaller milestones, the business can realize value along the way and ensure the project continues to align with business goals.
-
A simple example of milestones and the value they can provide:
Define the target and pull a list of examples to pass along to the business for review. This lets the team start working through outputs while the build continues
Start with the low-hanging fruit and the easiest cases to detect first. Depending on crew capacity and fix ability, the number of examples could provide weeks of work for service teams, buying time for the build
Outline the different drivers and features of the model and coordinate with the business. Oftentimes the customer service and operations teams have lots of anecdotal experience with these issues and can help identify what they see and the patterns they detect
If we treat all three of the above milestones as two-week projects, we can be six weeks into an 8–12 week build and have already provided three different pieces of value and feedback to the business.
Prove Value
The new project should be able to easily quantify the value it provides. By selecting the right project with a well-defined scope, we can measure the impact. Whether that is human hours saved, cost savings, increased growth, or a combination of all three, the metric should be easy to quantify.
-
Difficult to Prove Value:
“We want to implement AI to improve our customer experience and make our operations more efficient.”
This example is too broad and combines multiple vague goals. How do you measure “improved customer experience”? Which specific operations? Without concrete metrics like response time reduction, ticket volume decrease, or cost per interaction, it’s nearly impossible to demonstrate whether the project succeeded.
Easy to Prove Value:
“We want to build an AI system that automatically categorizes and routes incoming customer support tickets, reducing our average response time from 4 hours to under 1 hour.”
This example has a clearly defined scope (ticket categorization and routing), a specific metric (response time), and a measurable target (from 4 hours to under 1 hour). You can track whether the system is working by monitoring response times before and after implementation. You can also quantify value in terms of customer satisfaction scores and support team efficiency.
Expand with Purpose
After a successful project, the initial reaction is to keep building and growing as quickly as possible. However, before diving headfirst and committing to a swath of new projects, it’s worth taking a breather to assess the last project and understand the nuances of the new technology. Compare that against the outstanding project list to figure out which ones are actually a match and come up with a plan for those. Make sure to keep scopes well defined, builds fast, and check-ins constant in order to continue realizing value from new projects. Teams are able to pivot quickly when they fully understand the limitations of a tool, and not just knowledge about a tool.
As you expand, prioritize integration with your current tech stack and team capabilities. The goal is to extend what already works rather than introduce complexity that stretches beyond existing tools and knowledge. Favor solutions that can be implemented with your present platforms, skills, and data pipelines so builds remain small, quick, and scalable. When a new component is required, add it deliberately and only when it enables clear, measurable value without slowing iteration.
-
This example takes learnings from the successful pilot and applies them to a new, clearly defined use case. It maintains a narrow scope with specific, measurable goals. The team already understands the technology and its limitations from the first project, making this expansion more likely to succeed. The value can be quantified through resolution time metrics and customer satisfaction scores.
Successfully implementing new technology comes down to discipline and focus. Start with small, well-defined projects that address specific pain points with measurable outcomes. Build quickly, show progress along the way, and prove value before expanding. When you do expand, apply the lessons learned from your initial success to new use cases that are equally well scoped and measurable. This approach ensures that technology serves your business goals rather than becoming a distraction. By maintaining clear targets, fast iteration cycles, and constant communication, you can consistently deliver projects that create real, quantifiable value for your organization.