AI Augmentation vs. Automation

Hi there,

A lot of AI use cases fail, not because they lacked vision, talent or tech. But because they started off on the wrong foot.

Let me tell you a story about two companies that had the same goal - use AI to upgrade their sales operations.

This little example will help you understand one of the biggest differentiators that can make or break your AI use case.

Let’s go!

🚨 Last chance to sign up! 🚨

In this upcoming hands-on workshop, I'll walk you through 20+ use cases for ChatGPT in data analytics. LIVE on February 7 & 8, 2024. Recording available when you register. (Subscription required, but you can join with a free trial)

Case Study for AI in Sales: Same ideas - different approaches

Imagine two different companies, let’s call them company A and company B to keep things simple. Both companies had a similar idea: Improving their sales operations with AI. But they took different approaches.

Company A made hundreds of sales calls every week. During these calls, the salesperson would talk to customers about different products and get to know them better. After each call, their AI system would transcribe the conversation and provide a summary with a list of suggested information that matched fields in their customer relationship management (CRM) system, like company size, mentioned vendors, or the next steps. The salesperson could review the suggestions and make any necessary changes. With just one click, the information would be added to the CRM.

Company B tried a different approach. They purchased an AI-powered CRM plugin that crawls through all CRM records, identifies missing fields, and attempts to automatically fill in those fields based on a history of all available transcripts from all recent sales calls.

Pause.

Which approach do you think was more successful?

Here’s what happened:

After launching a prototype on a test user group, Company A realized that the amount of sales fields in the CRM would go up by about 40% immediately, compared to a salesperson who was not in the test group.

Company B, on the other hand, had a different experience. During the prototype phase, when they tested the system on sample data, they found that the AI service could only map 75% of the missing fields out of the box. Of those, about 10% were incorrect and required further manual review. Within the first 6 months, the team realized that if they wanted to scale this system across the entire CRM, they would need an additional workflow to ensure accurate results. At this point, 6 figures had been spent on implementing the solution in the CRM, consulting fees and the cost of running the AI services.

Learnings:

What was the main difference between the two approaches, and why was Company A so much more successful?

Company A made a smart move at the outset. They took their existing sales workflow, a high-value business process, and looked for ways to enhance it with AI without worrying too much about the tools right away. They put people first and relied on augmentation.

Company B, on the other hand, while starting with a similar problem, approached the situation by following the promise of AI automation offered by a tool. They put the tool first and relied on automation.

In fact, this pattern is quite typical between successful and unsuccessful AI projects that I see in my consulting.

Let’s shine a light on this difference even more.

What is AI Augmentation?

In simple terms, AI Augmentation is about using technology - in this case, AI - to help people do their existing jobs better. At its core, it's not about replacing people, but rather empowering them to do a task in less time, with more accuracy, or simply with more pleasure.

AI Augmentation recognizes that business processes can be messy, complicated, and hard to control. People are in charge, and we want to get the best out of them.

There are generally four types how augmentation with AI can happen:

Knowledge augmentation uses data to help people make better decisions by bringing in more information they don't know yet or don't currently have in mind. There are many use cases and examples:

• Retrieve similar customer contracts

• Suggest a forecast in a BI dashboard

• Give recommendations for improving website SEO

• Recommend a discount or pricing package for a new customer

Conversational augmentation is not so much about providing new knowledge or information as it is about helping users arrive at a solution through an interactive conversation. For example, it can be used to:

• Train a sales pitch

• Run a mock interview

• Get sparring for a business strategy

Contextual augmentation is when systems adapt to the user's current workflow context, learning from their interactions and preferences to provide more personalized and relevant information. Company A's use case above is a good example of this, where the AI pulls information from the interaction the salesperson had with a customer. Other examples include:

• Recommend content based on personal user behavior

• Formulate emails that mimic the user's language and tone

Creative augmentation is quite in contrast to knowledge augmentation. This type of AI augmentation gives the user some unusual or unexplored suggestions to work with. It is particularly useful for creative industries or processes such as art, image creation, creative writing. Some typical examples:

• Brainstorm themes for a corporate event

• Create new graphic artworks and logos

• Write catchy marketing headlines

AI Augmentation vs. AI Automation

Automation is pretty much at the other end of the spectrum. The goal is to perform tasks repeatedly without human intervention.

However, as practice has shown countless times, trying to automate everything perfectly can be costly and not always worth it, especially if the cost of maintaining the automation is higher than the cost saved by it.

That's why it's usually best to start your AI journey with an augmented solution that has a lower level of automation and a lower level of integration (yes, you can build amazing use cases with low integration and low automation!). From there, you can take it step by step and progress to more complex solutions. Check this article for more details:

Conclusion

While AI and automation are often conflated, it's generally better to approach a new AI use case by first trying to augment existing processes and workflows - especially if you're just starting out. This will give you immediate feedback, value, and lower costs to get started.

Once you have an augmented AI use case in production with a positive ROI, congrats - you're on the right track in your AI journey!

Next week, I'll explore the concept of augmentation a bit more and talk about best practices for making this augmentation most successful.

Also, if you're particularly interested in the area of data analytics, watch this space because I have an exciting announcement coming up!

Until then, think about: What workflow should be improved in your organization?

Hit reply and I'll help you gauge its readiness for AI augmentation.

See you next Friday!

Tobias