Here are Five Actionable Steps to Implement AI in Your Business

At the beginning of this year, one of my customers asked me to prepare a list of the top 10 GenAI applications. First, I liked that idea — who wouldn't want to know the must-haves that could drive their business forward?

But I found myself stomped and having to answer an overarching question: What makes an AI application truly 'must-have'? The answer, I discovered, is less about the applications themselves and more about the transformative steps your company can take to implement AI for a competitive edge.

That's why I will share with you my strategic framework: five actionable steps to implement AI in your company. Let’s get started!

Step 1: Leverage AI in your Landscape

Start by maximizing the potential of what's already at your fingertips.

"Do what you can, with what you have, where you are."

In the context of AI, this means before chasing new technologies, ensure you're fully utilizing your existing tools, e.g: Adobe Firefly, Microsoft CoPilot, and Google Duet. From what I can tell, the deeper you explore the field the sooner you realize that it’s mainly the big tech companies that have infused AI in their portfolios today. AI is mainly used to increase User Experience, improve integration, and generate ideas.

This means that we can treat AI the same way we do with any new technology or software: By implementing yet another change in the company. Just because you buy a flashy AI solution doesn't mean your people will jump on the bandwagon.

Step 2: Start Using an LLM like GPT-4 Turbo

The next step in preparing your company for AI is to start using some kind of Large Language Model (LLM) like GPT-4 or the new GPT-4 Turbo, PaLM 2, or Claude. There are also open-source alternatives that you can use if this is what you prefer. Here's how I see it:

I chose to distinguish between Open Source/Closed Source solutions and Backend/Frontend. A cost-effective way for your company is to start by using the backend services that are provided by private companies like OpenAI or hosting your own Open Source LLM.

As frontend options, I thought of two easy-to-implement solutions: ChatBotUI and Anse. On the other hand, you could think about licensing Enterprise Solutions e.g. for ChatGPT. This would come at a higher cost for additional benefits like data encryption and additional dashboards.

There are also free alternatives available like Bard/BingAI, but those might not be easy to roll out for the whole company if the basic infrastructure is not set up (Microsoft Exchange, Google Workplace).

In addition to implementing the AI, the way you interact with the LLMs is important. Make sure to train your employees and give them a basic understanding of hallucinations.

I have a free cheat sheet you can download here that reduces the probability of hallucinations and helps with making sure you use any LLM to its maximal intent.

Step 3: Automate with GPT-4

Let's say you've chosen an LLM that you want to integrate into your company's Enterprise Landscape, you can then leverage existing automation flows or build new ones.

If you have not considered building automations with e.g. Zapier or Make the new LLM could be the best reason to do so. Through the advances in LLMs, complex and repetitive tasks like sentiment analysis on x.com can be done in just seconds. These are my favorite high potential examples for task automations:

• Classification/Sorting: Classify the feedback of your customers (Sentiment) or triage Mails to different teams.
• Interfaces: Automate manual steps e.g. between two Systems since AI can be trained to re-format data strictly.
• Information Flows: Improve the efficiency of the organization by connecting multi-efforts.
• Reports: Reduce efforts by generating weekly updates and reports.
• Alerts: Use as a trigger (like an incoming email) to warn teams or follow up on leads immediately.

Step 4: Implement AI Chatbots

AI Chatbots are popular since they are easy and affordable to create. Here are my 5 top chatbots you can implement in your business:

In my eyes the strongest arguments for chatbots are the positive impacts on customer service:

• 24/7 availability: Enhancing your company's support availability, especially on weekends. Chatbots can fill this need with on-demand messaging, providing consistent responses and information without tiring.
• Omni Channel: Chatbots are designed to function on multiple platforms such as WhatsApp, Telegram, websites, and more. This reduces barriers for interaction and brings your company closer to potential customers.
• Lead Generation: A prospecting bot can assist in engaging leads by asking the necessary and relevant questions. With proactive messaging, promotions and offers can be tailored to these leads.
• Multi-Lingual: By addressing customers in their native language, we enhance trust and satisfaction, leading to higher engagement rates. Moreover, it allows you to tap into new markets without the need for significant localization efforts.
• User Experience: It introduces a fresh interaction method, optimizing time-to-value. Rather than navigating the homepage to retrieve information, customers experience reduced friction and can obtain details through a single query.

Step 5: Monitor and Understand AI Agents

Before diving into the final step, how would you define the term 'ai agent'? There are several takes on what it is, but I think this tweet captures the essence quite well:

In this article, let's focus on autonomous agents and multi-hop agents (1&4). These ai agents chain LLM API calls to achieve a specific goal. But what's the point of it?

The most common issues with LLMs today are their reasoning capabilities: There is no self-optimization and a needed, constant interaction with us (aka 'humans'). Instead of this passive approach, we can actively enhance their reasoning abilities with AI agents. This is achieved by having several agents that engage in conversations, give advice, and perceive their surroundings.

There is a famous paper from Stanford, where 25 generative AI agents were playing a video game and building their own community. Here's a video based on this to get you inspired for real life applications:

For mid-sized companies, the possibilities are gamechangers:

• Continuous Productivity: Picture a workplace where AI agents work 24/7. They optimize processes, detect inefficiencies, and suggest improvements, ensuring businesses operate at peak performance.
• Tailored Customer Experiences You will be able to personalize marketing campaigns, find leads, create dynamic narratives, and drive customer engagement like never before​.​

But we know this isn't just about the present. It's an investment in the foreseeable future. And at the rate things are moving, probably the very near future.

As of now, the closest solution coming to this vision is AgentGPT (Open Source) or custom code projects.

In the most recent OpenAI DevDay Assistant API's were launches together with a lot of new multimodal features like vision, image creation, and text-to-speech. AI Agents, as a natural development of the existing LLM landscape, will only become more powerful, and we should make sure we invest in them.

Conclusion

• Step 1: Use the AI tools you already have before buying new ones.
• Step 2: Start using an LLM like GPT-4 and teach your team to use it right.
• Step 3: Let AI automate repetitive work, so your team can focus on the important work.
• Step 4: Implement chatbots so your customers can get help anytime & in any language.
• Step 5: Keep an eye on AI Agents to make sure your company stays ahead.