How to start with AI in your company: a checklist

Care to share?
How to start with AI in your company: a checklist
19:41


Over the last few years, the artificial intelligence (AI) craze has taken the business world by storm and there’s no sign of stopping. Everyone wants a piece of this pie, and there’s a good reason for this. AI solutions can foster innovation, streamline processes, and make a significant difference in businesses’ bottom lines. 
 

Tempted by these benefits, companies are massively developing AI projects. However, few are doing it well. It’s a bold statement, but statistics speak for themselves: a staggering 85% of such projects fail. 

This article is designed to help you and your company become the other 15%. It’s a comprehensive checklist of areas to research, preparations to make, and best practices to establish. Prepared by our AI veterans, it provides everything you need to know to approach this transformative project the right way. At each step, this checklist connects the three areas that must be in harmony for the project to succeed: technology, business, and users. 

So, without further ado, let’s jump right into the checklist. Make sure to also download the accompanying cheat sheet that you can keep with you throughout the project.

Download the PDF cheat sheet

1. Define your synergy clusters 

As the first step on this journey, you should take a thorough bird’s-eye view of your business. Consider the most high-level aspects of your company, such as:  

  • Your main income and expense streams. 
  • Key silos and their touchpoints. 
  • Primary business goals and strategy. 
  • Your digital agenda. 

While this might seem banal at first glance, it’s a key element of this process. Such a bird’s-eye view will allow you to pinpoint synergy clusters within the company. These are the areas where crucial silos and processes intersect and, hopefully, synergize with each other. In other words, synergy clusters are the areas with the biggest impact on the business’s top and bottom lines. 

Why is this exercise so important? Because you also want innovation to have as big of an impact as possible. The AI solution should be implemented in one of those synergy clusters: either one that provides the best infrastructure and data foundations for AI or will benefit the most from innovation. In the ideal scenario, you’ll find a synergy cluster that fulfills both of those criteria at once.  

CTA

2. Evaluate your digital potential 

When you’ve settled on a synergy cluster with the most potential, it’s time to dig deeper into it. Unlocking the transformative potential of AI should begin with a comprehensive evaluation of the current state of the cluster. At its core, this involves assessing your infrastructure and existing processes and a thorough look at the data that you’ll be working with. 

Approach the project the right way 

This process should start with analyzing your existing resources, tools, and processes to identify areas where artificial intelligence can be integrated effectively and make the biggest impact. This is not merely about adding AI to your existing workflow; it should be a complete reimagining of how your business will operate with AI at its core.  

Assess your AI readiness 

Next, gauge your organization’s readiness for an AI transformation. In other words, evaluate your data quality, infrastructure, and the skills of your workforce. Understanding the current state of your data and systems is essential in two ways. First, it makes it possible to identify gaps and areas for improvement. On the other hand, it also uncovers the opposite: high-quality data sets that can provide solid foundations for training AI models. This is called data gravity. 

Trace your data's journey 

Understanding the journey of your data is pivotal to establishing a solid foundation for AI implementation. The technical term for this process is data lineage. It involves tracking the flow of data from its origin through various stages of processing to its final destination. This process helps ensure that your data infrastructure is robust, reliable, and ready to support AI applications.  

In practical terms, you should begin by mapping out your data sources, storage systems, and data processing workflows. This will help you identify any bottlenecks, inconsistencies, or vulnerabilities in your data pipeline. For example, data that’s stored in your customer relationship management (CRM) system often originates from various sources and is later distributed to multiple endpoints. Is it consistent and reliable throughout the whole process? 

global-business-network-2023-11-27-05-24-20-utc
Stock photo from Envato Elements

Find the weak links in your chain 

Manual data inputs are often the key bottlenecks in data processing. These points of human intervention can introduce errors, slow down processes, and hinder the scalability of AI solutions. It’s crucial to identify and address these manual touchpoints to be aware of the challenges that your data presents.   

3. Spotlight on value

We’ve already established that the best use cases for artificial intelligence need to make sense in terms of your company’s needs, infrastructure, and available data. There is, however, another piece in this puzzle: financial considerations. The investment must simply be worth it: it must either increase profits or reduce costs.  

As a rule of thumb, you should prioritize use cases that promise the greatest return on investment (ROI) with minimal resources. This will help you build a compelling business case for AI adoption and “sell” it to the rest of your company. 

To find those use cases, you need to align AI opportunities with your company's strategic goals. A part of this job has already happened while defining your synergy clusters, but there’s still more to do. 

Download the PDF cheat sheet

Define goals and metrics 

You finally have all the pieces to properly define the goal of the project. Such a vision should be built on a deep understanding of the synergy cluster that you’ve selected. In other words, how exactly does it influence the value created by your business? Where should this value be increased or introduced? 

To address this, you should engage with stakeholders from different departments to gain insights into their needs and challenges. Do it in a structured and deliberate way:  

  1. Decide on the high-level goal or goals of the exercise.

  2. Translate the goal into precise discovery questions. 

  3. Based on those, write scripts for interviews. Contrary to popular belief, these aren’t the same as discovery questions, and treating them as such is a big mistake.

  4. Validate the questions. Will the answers be practical for you? If not, rewrite the questions.  

Answers collected during interviews will let you define clear objectives, success metrics, and potential risks of the project. This will help ensure that the AI solution is tailored to address precisely defined pain points and deliver measurable benefits. The interviews will also lay the groundwork for the ideation phase and help you assemble the right team for this task. 

Survey the market 

External research should be just as big a part of your preparations. Explore the market and see how others optimized the synergy clusters that you’ve chosen for your project. Look at the immediate competition within your industry, but also cast a wider net. There’s a chance that a solid solution will work regardless of industry boundaries given small tweaks.  

scott-graham-5fNmWej4tAA-unsplash

Stock photo from Unspalsh

Market research can inspire you and help fill in the gaps in ideas that you’re developing on your own. It can be the glue that binds all the loose dots that have emerged so far into a cohesive vision. Just as importantly, it can prevent you from wasting time and resources on needlessly reinventing the wheel. 

4. Workshop solutions

You might have noticed that we’re already far along the checklist, but still haven’t mentioned defining the actual solution. This is deliberate. When you think about the end product too soon, it can distort your research: you ask questions designed to confirm your idea rather the objectively assess the situation. 

Now, however, the time to brainstorm ideas has come. Organize workshops and make sure to invite as many diverse stakeholders as possible. By doing so, you’ll be able to lean on their differing perspectives, expertise, and insights. At this stage, there are no bad ideas; you’re going for quantity over quality. 

Empower stakeholder co-creation 

Co-creation with stakeholders is an excellent way to validate the business value. By workshopping with people who will work with the AI tools on a daily basis, you’ll define the precise roles and tasks that they perform within the cluster. This ensures that the solution aligns with their needs and expectations and responds to real pain points. You’ll be able to pick up on differences in their approaches and priorities and address them effectively to reach common ground. Simply put, you’ll make sure that your bird’s-eye vision is aligned with experiences on the ground level. 

high-angle-view-of-diverse-business-people-discuss-2023-11-27-05-03-13-utc
Stock photo from Envato Elements

The workshop setting also facilitates organic research. While brainstorming ideas, participants will naturally look for inspiration or data that supports their vision. This can be the last missing piece that completes your puzzle.  

Lastly, it’s about enabling the satisfying feeling that your team's voice matters. Early and active engagement with stakeholders helps build a shared vision and commitment to the AI initiative.  

Analyze the results 

As a result of the workshops, you’ll end up with a loose collection of ideas and perspectives. They might seem chaotic at first, but you’ll soon begin to see patterns and fairly well-defined paths forward. Consider them from the perspectives of added value, required effort, and risks. The one that strikes the best balance between the three should be your choice.  

If possible, validate the assumptions that the chosen idea is based on. Do it with as little effort as possible: as a general rule, it’s cheaper to test initial assumptions than the finished solution. Ensure that the solution meets the defined objectives and performs reliably under real-world conditions.  

5. Develop a PoC

It’s finally time to get started on designing and building your AI solution, and a proof of concept (PoC) should be the first “product” that you deliver. It’s a critical step in demonstrating the feasibility and potential of an AI implementation. A PoC allows you to test key aspects of the solution in a controlled environment, identify any technical or operational challenges, and refine the approach before scaling up. 

AI-Use-cases-covers mock-up_logo

Set the stage 

Start by defining clear objectives and success criteria for the PoC. This will help you measure its effectiveness and make informed decisions about further development and investment. Remember to be realistic: you aren’t building a complete product but a rough demonstration. It doesn’t have to look pretty or come with a number of features. If you only focus on one aspect and get it to work, it’s already a success. 

Build the solution 

A well-designed PoC should focus on the most critical components of the AI solution. These will vary from project to project, but aspects such as data integration, algorithm performance, and user experience are fairly universal bets. Collaboration with a cross-functional team that happened during the previous stage now pays off in a big way. It ensures that all relevant perspectives are considered and the PoC is tested in realistic scenarios.  

Showcase results  

Showcasing the capabilities of your AI solution through a PoC is the logical conclusion. Use it to build confidence among stakeholders and secure buy-in for further development. It is, after all, the first time they see the tool in action and not just on paper. Analyze the PoC results to demonstrate the tangible benefits of the AI solution, such as improved efficiency, reduced costs, and enhanced customer satisfaction.  

Make sure to communicate the findings clearly and transparently. Highlight successes, but also talk about lessons learned. A PoC is, in a way, an experiment. Even if it's designed and conducted perfectly, sometimes it doesn’t go as planned. That’s the whole point of experimentation, and there’s no shame in admitting this. By addressing concerns and outlining a clear path forward, you can pave the way for full-scale implementation and long-term success. 

6. Expand the vision

We’re back to planning, but not for long. With the PoC results as your guide, the next step is to create a detailed roadmap for navigating from initial concept to full-scale execution.  

Create a robust plan 

A well-structured plan should outline the key milestones, deliverables, and timelines for the AI initiative. It should address the resources, skills, and technologies that will be required at each stage of the implementation process. 

Your roadmap should also include a detailed plan for scaling the AI solution across the organization. Identify the key enablers and potential barriers to scaling, such as data quality, infrastructure readiness, and workforce capabilities. Develop strategies to address these challenges and ensure a smooth transition from pilot to full-scale deployment.  

email

Preserve stakeholders’ trust 

Selling the expanded vision to stakeholders can be more difficult than before. A PoC trial run is noncommittal and there’s little risk involved. With a full implementation, however, you’re going all in. To address the potential obstacles, use data and insights from the PoC to support your case and demonstrate the feasibility and value of the solution.  

7. Launch iteratively

We’re at the very last stage of the checklist. You’re nearly ready to develop and launch the AI solution that you’ve been conceptualizing, planning, and validating.  How to do it the right way, though?  

Start modestly 

When venturing outside of your business’ comfort zone, being too ambitious rarely pays off. There are simply too many unknowns. In most cases, starting with a minimum viable product (MVP) is the right way to go. An MVP is a simplified version of the solution, which includes only the most essential features. In other words, it’s the bare minimum that will complete the job that it’s designed for in real-world scenarios. 

There are a few benefits to starting with an MVP: 

  • Simplicity: To put it bluntly, there are fewer moving parts and things to break. During the initial phases, the solution will be new both to users and administrators, so you don’t want to add unnecessary complexity to the mix.  
  • Live feedback: An MVP enables you to test the solution in a live environment, gather feedback from users, and make necessary adjustments before rolling out the full version.
  • Pivot opportunities: There’s a big chance that the results of the initial rollout will surprise you. For instance, users can end up using the solution in a different way than you anticipated or request a feature that you didn’t plan. When starting with an MVP, it’s much easier to pivot and accommodate their feedback. 

Keep improving the solution 

Launching iteratively is a cycle of continuous improvement. Start by deploying the MVP to a small group of users or a specific business unit. Monitor its performance closely and collect feedback to identify any issues or areas for improvement. Then, use this feedback to make iterative enhancements to the solution and gradually expand its scope and functionality. Of course, you should prioritize features that deliver the most value in the shortest time frame; or, in business terms, provide the highest ROI. 

You might think that your work ends at this point, but this is not the case. Successfully maintaining your newly implemented solution means that this cycle keeps going. Your business is constantly evolving, which means that even the most advanced solution will become outdated at some time when left on its own. Continuous monitoring and enhancements will ensure that you’re always one step ahead of these changes and your AI solution will keep being effective. 

8. Keep your team fluid

You might have noticed that an adaptive team has been an unspoken theme of this checklist. It is, however, also worth stating plainly.  

Implementing artificial intelligence the right way must be an interdepartmental and interdisciplinary effort. There’s no such thing as a set-in-stone team for the entire project: it should evolve in accordance with the current needs and stage of the process. Make sure to include different points of view that can keep challenging one another in a constructive way. At some stages, you’ll need to lean on one perspective more than others. For instance, when assessing financial feasibility, the business side will likely have the final say, but when researching needs and pain points, it’s the user’s voice that matters the most.   

KUL09929

That said, there’s one role that needs complete information about the project: the people who will be responsible for implementing and maintaining the solution. It’s crucial that there are no surprises waiting for them because this can present significant practical problems as well as undermine the broader trust in the solution. In practical terms, you should ensure that there’s always a person with continuous knowledge of the project and make them responsible for passing this information on when needed.  

Final thoughts 

Implementing AI solutions in your business is more than a technological upgrade. It’s a strategic transformation that can redefine the company’s future. However, value rarely comes easily. The path from concept to implementation is complex and requires a thorough understanding of your digital landscape, robust planning, and continuous engagement with stakeholders.  

By following a structured approach outlined here, you mitigate the risks inherent to innovative projects and lay a solid foundation for AI-driven growth. You also foster a culture of continuous improvement that’s essential for keeping pace with the evolving technological landscape. After all, AI implementation is not a one-time project but an ongoing journey that demands adaptability and foresight. 

Next steps 

We realize that this checklist might be a lot to take in now, and then later implement during the project. Innovating is rarely a straightforward process. Because of this, it might be worth leaning on the knowledge of seasoned experts to guide you through the initial steps.  

That’s exactly what our AI Sprint is for. It’s a structured workshop designed to help you assess the AI potential in your company, find the best use cases, and develop a realistic roadmap for implementation. If you feel that your company is ready to take a decisive but well-measured step toward the future, make sure to get in touch with us.

Keep the guide with you

This checklist is designed to accompany you throughout the entire project, but we know that coming back to an article again and again isn't very convenient. If you'd rather have the key working questions for each stage in a handy PDF format, you can download it through the form below.

 

Published July 12, 2024