When Your Data Is Your Coworker

Natural language and how you will…and maybe won’t…do your job

So much has been said already about the power and the utility of AI chat interfaces in how we do our jobs. However, I think the day-to-day nature of where things seem to be headed are extremely interesting for a variety of reasons, not the least of which is the way people are likely to perceive their working environment over time.

As people, we are always looking for ways to reduce what we consider toil and risk. Toil is those things we have to do to survive and thrive in life that really bring us no joy. Risk, of course, is the chance that anything might cause us to lose (or miss the opportunity to gain fairly). The less we have of both, the more (at least in theory) we can focus on what we look forward to doing.

Agentic AI (and AI in general) promises to provide the tools to remove vast amounts of toil and risk from our daily lives. But a key part of this is a new default user interface for our digital systems. And that’s what I want to focus on here.

Talking Instead of Commanding

First, I think we should acknowledge the ways in which natural language conversations with digital systems improves how we complete certain tasks. I’ve noticed this most when I’ve needed to find out what new (to me) technical terminology means, and how it might affect the overall market.

Where I would have once used Google to search the term, then read through a few articles until I was satisfied I got the concept, I now simply ask an LLM to explain the term to me. If there is additional terminology included in the response that I don’t recognize, I quickly ask a follow up question to get that clarified. I even occasionally ask the LLM to suggest additional topics that I should learn about to fully get the value of a new technology or other subject matter.

The key here is that I am having a conversation with an “expert.” Yeah, the LLM doesn’t really know anything about the topic, per se; it is just using the expertise of millions of humans that came before it. But the nature—the psychology—of talking about the topic gets me to a response I am happy with much faster than reading a half dozen articles and trying to put the pieces together myself.

Similarly, when I develop code—or I want AI to help with completing any task—I do so in a “give and take” conversation style. I am not just continually commanding the AI to do things, I am also asking questions, or allowing the AI itself to ask questions where it needs clarification. In that way, the process is much more like working with a coworker than running a machine.

I don’t do a ton of data analysis in my current role, but those I know that are doing so say they are having a similar experience with AI. They are seeing glimpses of “talking” to their data. In other words, rather than a process a lot like setting up a really complicated scientific calculator to run an equation and return a graph, they are having a conversation with a system that can translate their immediate questions or asks into queries and actions, and return the results in easily digestible formats.

The Conversation is Key

This transition is just beginning, so I don’t want to leave everyone with the impression that you should already be talking to your data. There is a lot to be done before usage of these interfaces will be mainstream.

But planning ahead for this change is a good idea. Experiment with different conversational approaches to analyzing your data. Understand the difference between routine monitoring of metrics and exploring for insights. Prepare for ways that your AI models can return graphs and diagrams in addition to text.

If you have been using Jupiter notebooks or similar tools to do data analysis, all of this will be very familiar. The change really is moving from Python (or whatever) as your language of communication to your native speaking language.

The added benefit with AI is that you get help with identifying the right approaches for achieving a specific outcome. You will quickly find that you are not “alone” in solving problems, but that you can quickly assemble a “team” of agents and prompts that can play various roles. Software development, research, even quality assurance and operations can all be managed through AI…if you have the frameworks and automation to make it happen.

You still have to plan, iterate building solutions, and maintain them as they execute, but instead of that requiring a team of people toiling away (there’s that word again), you can utilize the models to reduce or eliminate most of that toil, and improve your risk profile to boot. It’s not “magic bullet” stuff, but it’s a huge step in that direction.

The move from clicking the mouse (Graphical User Interfaces, or GUIs) to speaking with your systems (natural language interfaces) is a big step, but one that simplifies the human-computer interface in a significant way. Perhaps the age of GUI dominance is coming to a close.

Starting the Conversation

So, if you are a data scientist looking to change the nature of exploring your data, how do you get started? Well, I suggest a few things:

Don’t Build Another Data Lake, Use What You Have

Whatever you do, don’t build a ton of new data infrastructure and replicate data yet again to experiment with AI. You’ve been through this before—the expense, the pain, and the risk it introduces into your business practices.

Instead, there are great technologies (Kamiwaza being my highly biased favorite) that will allow you to direct your questions (aka the inference processing for your prompts) to models that sit close to your existing data sources. In fact, you may eventually even be able to shut down a data lake or two because AI orchestration can run such inferences across multiple models (and data sources), returning a single response.

Reuse Proven Applications And Agents

Finally, I think it’s critical to find communities and vendors that give you a starting point for your conversational interface (and agentic AI) journey. Kamiwaza, for example, has an App Garden where a number of proven solutions to common business problems are readily available to deploy on day one. And that list of solutions is growing continuously, with new solutions appearing every couple of weeks. You can also add new solutions to your own private App Garden, and soon the growing Kamiwaza community will add their own contributions. With each new innovation in the App Garden, Kamiwaza becomes an increasingly go-to resource for using AI in your company.

Bring Down The Cost of Experimentation

The truth is, very few enterprises know exactly how AI will change the nature of their businesses. Everyone is experimenting with the technology, looking for lasting and differentiating value. Experimentation is a “gotta have” if you want to get ahead of your competition. And, while it may seem like putting the cart before the horse to purchase an AI orchestration platform (or even install a free open source platform), having the right tools to make AI usage and agent implementation cheaper and easier is a great investment. In fact, it's likely a better investment than paying consultants on a project-by-project basis.

In the End…

What matters after all is said and done, however, is that you study the ways that conversational, natural language interfaces can boost the outcomes at your organization. It doesn’t make sense to replace something that works really well today just to get AI involved. However, for many functions that require dashboards or notebooks or even old-school database query pages, these interfaces may improve the flow of work for those that rely on them. And agentic AI combined with these interfaces may completely change the way the job gets done—for the better.

What are your thoughts? As always, I write to learn, so let me know in the comments if you disagree, or if you are using conversational approaches to software development, data analytics, or any other function within your organizations. There are so many opportunities, it would be fun to hear what experiments you are running today.