Five AI Habits to Build in Your Company

As we start the new year many of us have taken to adopting new habits with the intent of being better versions of ourselves. These might be working out more, reading more books, focusing on family time, or spiritual growth. As we look the AI revolution in the eye, the most distinctive quality we’re seeing is the universal transformation of how work gets done. Working with over 100 companies on AI strategy, I’ve seen that some have adopted better than others. In some companies we’re seeing an intentional focus toward aiding employees in making the journey. In others we see laggards who have only blocked or mitigated perceived threats, rather than enable AI outcomes. Let’s take a look at five AI habits that effective companies have implemented to enable every employee to be more productive.

Here is a sneak peak of the five habits or activities. There are certainly more habits to learn, but these are five that I’m seeing used at scale and in increasing number:

Number 1: Having an AI agent research and prepare information or an asset

Let’s assume for a moment that you have a big meeting coming up with a potential client. How do you prepare for that meeting? How do you ready information that helps you gain insight quickly and effectively? Imagine that every time you enter a meeting you have a “packet” prepared of the most critical information necessary for that meeting.

Let’s take an example of preparing for a financial consult:

1. Information on the client and their family
2. Understanding of their financial wellbeing operationally
3. Run several analysis against their retirement savings
4. Prescriptive actions based on the POV of the company

Another example might be a conversation with a customer you produce goods for:

1. Run a report on their financial performance over the past quarter
2. Summarize any leadership changes
3. Describe what makes the company special based on their public data
4. Predict demand based on what we understand about their previous/future performance
5. Summarize a prescriptive forward plan and pricing strategy

Here is an example of asking a question in preparation for a meeting about store performance. You aren’t needing to go interpret a Power BI report. You are simply asking the question and getting the data immediately. Of course, this required the data to be wired up, but it is entirely possible and practical to make this kind of capability available to your decision making teams for diverse sets of data.

Imagine that you have an intern that can get all of this information ready for your next meeting. In some cases it is simply a matter of using what is already off the shelf. In other cases you might need to interface with a business system to gather this information. What I can tell you is that there are companies doing every one of these things today. The ability to leverage AI skills to accelerate your salesforce, customer service, research, etc. cannot be underestimated in the context of preparing you for your conversations.

Number 2: Delegating management of post-meeting action items

This habit is the most obvious to me, as I use it every day and have now come to rely on the regular output of meeting summarization and action items vs. needing to direct energy there myself. This isn’t just one activity but a group of activities that build on each other.

Take a good look at the above. The first two I use pretty often. The third, my colleague Tabatha Frozena taught me from a call we were on recently. The fourth, I use every day but was really impressed when I used it for a board meeting that I’m part of. We had a conversation at that board meeting about a person taking on the secretary role and the suggestion was for a person to take it on, “if he could use AI” and we all agreed. The results were astounding and it enhanced the ability for that person to collaborate in the meeting.

The hot tip is now moving from summarization into managing the action items coming out of the meetings. That movement is a continued enhancement on taking the notes.

Finally, the idea of answering questions from aggregate. I have a client I had worked with over several meetings and was working on a roadmap with them. I was working on a recommendation but wanted to clarify a point. With all the meetings Copilot transcribed I was able to ask a question of the aggregate of those meetings without knowing where a specific point was brought up. This enabled me to quickly find a specific datapoint without wasting an hour looking for it or -likely- having to ask it again. These simple techniques dramatically change the way that we’re engaging personally and with customers.

Number 3: Asking an AI agent to take regular action based on a regular situation

The ability to truly delegate to an AI agent is based on how much we trust it to perform an action or more-so our understanding of how to do the delegation. The most simple form of delegation is to take a regular action that we understand. This is where capabilities like Copilot Actions are becoming so powerful. Imagine every time a meeting summary is created you want to move the transcript to the customer’s Teams site and assign action items in ToDo for the required individuals.

You can see something building here… for individuals to learn how to delegate a regular task to an AI agent, they need to get comfortable working with an AI agent. At the risk of using the “don’t fall behind” scare tactic, I do want to say that there is that risk. The problem with some companies is that they are so wrapped up in legal protection and perceived security controls that they are irresponsibly holding their employees back from productivity. The growth of AI skills and adoption is always a balance between adoption and governance. You need both to exist in order to gain ground. You can’t have employees automating tasks without governance, but to still be in a position where you can’t use internal-grounded data is a bit miss as well. I see the same theme with cost justification. Some companies are overly onerous about having to prove the cost justification of leveraging AI that they miss the aggregate performance improvement. Thankfully, tools are emerging to help us with that story and enable evaluation of real business outcomes, such as Copilot Analytics:

Number 4: Running Processes Autonomously

The next step on something like Copilot Actions is to take long-running processes and build autonomous execution of the activity. This doesn’t mean it is without oversight or a safety net, but it means that where possible the next phase of AI-enabled RPA will have significant gains in our organizations. Think of building AI agents that act like pseudo-employees to perform regular tasks. Then, think of entire AI driven org charts of AI agents that perform activities at the service of their partner-humans in the business.

A way to think about how this is advancing is to imagine you are training a new intern on a task that they will perform regularly. What would you do? You’d walk through it on your computer, describe it as you go, and provide the intent behind the action. This is the difference between what legacy RPA was capable of and an intern. The legacy RPA was not able to understand the intent or take in other descriptive content. The modern AI-enabled RPA, like what is possible in Copilot Agents is to automate processes and think more like a human in the way they are executed.

The opportunity to automate regular processes is key to what we do every day and will open the door to not only more productivity, but redirected energy. I know this can be scary but think of the graphic below. You job in 5 years will look a lot different than it does now. Perhaps not in the outcomes, but in the how you accomplish those outcomes.

Number 5: Delegating a complex set of tasks and then reviewing output

This has some similar characteristics to the above but takes it a bit further. The best example I can think of is what is happening with development. We’re slowly seeing the ability to delegate code generation, remediation, and build activities to AI agents that act like true development partners. Imagine a future where this is true…

Think of a huge set of code refactoring that needs to be done and image assigning an army of humans to do that basic task. If you had 1,000 stored procedures to refactor, what if you could create an AI agent that ingested that and then turned it into a rules-engine rule that fits your new architecture? This is where AI-enabled development is going.

I’ve noticed interesting pushback among some developers to use AI as part of their development workstream. In some cases the developers see it as obvious and they are already using it. For some, they are hesitant because of one thing or another. It is incumbent on every dev leader to lead their development teams into this new generation. To discover what is possible and keep challenging the norm. The productivity gains are too significant to ignore and the opportunity is before us.

What to do now?

So, what to do with this information? If you’ve started, keep it up. If you haven’t, now is the time to engage and start the movement toward this journey. If you haven’t started with strategy, start there, but ultimately you need to create two lanes. A first lane for enablement and another for product engineering (or the factory)

Thanks for reading! Remember to follow all the content coming out of the AI Leadership Weekly newsletter and see you next week!

Nathan Lasnoski