Getting Started with Business AI: A Non-Technical Roadmap

Hi there,

One of the most common questions I get about AI is: "Where do I start?"

AI is a rapidly moving field that can be overwhelming for those without a strong technical background. If you want to quickly answer "What's in it for me?", keeping up with trends and cutting through the noise is a huge challenge. That's why analysis paralysis, the feeling of being overwhelmed and not knowing where to start, is a common barrier to successful AI adoption.

To help you navigate this maze, I've distilled the essential steps from my consulting practice. They have proven effective for both small and large businesses, with a particular focus on B2B.

So let's see how we can make AI work for you.

No one can build an AI roadmap for you

One of the most common mistakes I see business leaders make in the early days of their AI journey is to delegate AI to people with more technical expertise, such as the IT department or data scientists.

For some reason, there's a perception that people with more technical expertise are much better equipped to create a detailed plan for integrating AI into your business. However, AI is not just another tool or software that you can buy like a new CRM. AI is a technology that can be applied across multiple workflows.

And it's up to you to define where this makes sense.

Of course, especially at the beginning, you don't have to do this alone. You can gather a small team, seek out expertise, and get support from IT. But ultimately, it's important to remember that these people are building the AI roadmap with you, not for you.

"But how do I do this?"

Read on!

6 Steps For Starting Your AI Roadmap

The following list of strategies has worked for me in many organizations of all sizes and levels of technical maturity. For maximum success, however, I recommend that you start by focusing on a specific part of your organization.

This could be a business unit, a department, or a product you want to improve. If the results you're getting are too abstract, zoom in. If they're too granular, zoom out.

Step 1: Learn the AI Basics

I know AI sounds like it's for someone who spends more time in front of a computer than people, but the truth is, if you want to grow your business with AI, you need to know what's behind the buzzword.

There are literally hundreds of different ways to explain AI. But I've found the following concept works best for me: AI archetypes. This groups AI capabilities by the type of data they work with, and while you wouldn't find this in a scientific journal, it cuts through the hype really effectively.

In short, the 5 AI archetypes are:

• Supervised Machine Learning, ideal for predicting small tabular data

• Natural Language Processing to make computers understand text

• Audio and voice, often used to turn text to speech and vice versa

• Computer vision to let computers “see” images

• Generative AI to read and create anything

Once you're familiar with these archetypes, read up on some commonly used terminology, the pitfalls of integrating AI too much at the beginning and the 4 Supercategories of Generative AI.

Bonus: If you have a question, ask this AI Mentor GPT, that I've trained on my own materials.

All in all, step 1 shouldn't take you more than half a day.

Step 2: Identify Business Pain Points and Bottlenecks

Now that you have a high-level understanding of AI, you should take a look at our business and, ideally, your business processes. Because it turns out that the easiest way to grow your business with AI is to start by improving the things you're already doing:

• workflows

• user stories

• product features

• etc.

But don't fall into the "improve for AI's sake" trap. If a process is running just fine, adding AI to it will probably just make things worse.

There are two ways to look for problems effectively: Pain points and bottlenecks.

Pain points are problems that you are encountering right now. For example, if you're dealing with a lot of customer support tickets, a low response time or a high number of unanswered tickets is a current pain point.

Bottlenecks are problems that will hurt you in the future. For example, if you run a call center that does not serve customers outside of business hours, this may not be an issue now. But as soon as your competitors start offering 24/7 voicebot-assisted customer support, you won't be able to keep up.

So go ahead and think for yourself: Where are my problems right now and in the future?

Make a list and proceed to step 3.

Step 3: Map AI Capabilities

Once you better understand your problem space, it's time to see if there's a fit for AI. To be clear, while AI is a super horizontal technology, not every problem is a good AI problem. For example, if one of your core business challenges is that you have a poor data culture, it's unlikely that AI technology will solve that.

But how do you find these sweet spots of relevant problems and good fit for AI?

The easiest way to do this is to map the stages of the process to AI capabilities that can support them. It can look as simple as this table:

For each column, we have one of the AI archetypes, such as supervised machine learning, NLP, audio, computer vision, and generative AI.

And for each row, we have a different step of the process we’re looking at.

Now we can just fill in the blanks and see what kind of AI service could support which part of the process. Don’t look for tools that will automatically solve your problem, look for ways AI can augment your workflow.

Fun fact: You could even ask ChatGPT for help.

Shameless plug: Run an AI Design Sprint™ if you want to do this professionally.

Step 4: Analyze Use Cases

Now, only, it's time to look closer at use cases.

An AI use case is simply a match between a business problem and an AI capability that could deliver a positive ROI.

Don't overthink this step. We're still early stage and we haven't touched any tools or data yet, so let's keep it super simple.

Organize your use cases in fact sheets. A use case fact sheet fits on one Power Point slide, and describes your initiative with - at least - the following fields:

• Title: How do you call this use case?

• Description: Explain what it does in simple words

• Business impact: Which needle would be moved? How?

• Feasibility: Initial big guess. Do you have what it takes?

• Owner: Who’s the main contact for this use case?

Here’s how a very basic use case fact sheet could look like:

Step 5: Prioritize and Prototype

Rank your use cases according to business impact and feasibility. This could look like this:

The use cases in the upper right corner (medium to high feasibility and impact) are the ones to look at first.

Now before you get your hands dirty, do one final check:

Rule out all use cases where you cannot build a first prototype in less than 20 days or with a budget of less than $20,000 (US / EU region).

I call this the 20-20 rule.

It essentially eliminates all use cases where you need to collect data or train an AI model from scratch. Those use cases may be great, but they're not where you want to start your AI journey.

Start small, but think big. Prioritize AI projects that promise quick wins and can scale over time, building trust, momentum, and support for your broader roadmap.

Once you have that momentum, tackle more complex use cases.

Step 6: Iterate and Improve

This is the most critical step of all.

Your AI project is very much like an agile software project. Hard to plan, hard to predict. There are many uncertainties: data, technology, people.

Don't get me wrong - you have to have a plan. Just be willing to change it along the way. My former manager told me - planning replaces coincidence with error. That's the right mental framework to have.

But be aware, for most organizations this requires a huge mindset shift. Choose your stakeholders wisely for your first AI project. They must share this experimental culture.

Otherwise, your initiative will tank before you can say Chat GPT.

Conclusion

These are the essential six steps to take an AI project to the win.

Sounds simple, right? And yet it's so hard.

The real challenge is to actually stick to that roadmap, execute it, and not get distracted by other things because your CEO accidentally overheard something "cool" at a conference or your competitor launched their "own" GPT.

So claim your roadmap, make it happen, iterate over and over, and grow your business with AI.

If you need help, reach out anytime.

See you next Friday!

Tobias