Lots of talk about AI these days. It seems like every day there’s a new stunning achievement in the field of artificial intelligence that is poised to upend the way we work. We’re having these conversations as well and have some thoughts.
Let’s address the macro elephant in the room. AI will disrupt the labor force in ways we haven’t experienced in our lifetimes. Quite literally millions of jobs will be displaced, first in the service and knowledge economy, and eventually in the manufacturing sector. And it will be different than prior eras when technology disrupted the job market due to the sheer speed and recklessness of the rollout. There is far too little attention being paid to the security concerns AI presents. In time, these will work out but we’re going to break a lot of eggs on the way to making this omelet.
In the micro sense, many of us are already gaining a sense of the importance of AI in our daily lives. And many of our customers are wondering aloud how they should be thinking about incorporating AI into their workflows and what it means for their industries. And, quite frankly, what it means to agencies like ours.
This is where things get interesting.
In the knowledge and service economies, many direct use cases are already extremely powerful. The ability to generate high quality research and employ it to generate more thoughtful user experiences is positively staggering. We’re seeing it in our own processes. We are better, faster and more productive today than at any point in our existence. And our clients are the beneficiaries of this activity. Others are wondering whether they can onboard these technologies to bring the work that we perform for them in house. Both are possible, but certain fundamentals remain. I’ll get to that in a moment.
The ability of AI powered solutions to train on data sets to provide specific outcomes that are vastly superior to old methods is here. This is a now thing. But it’s only the beginning. What the tech industry is working toward is a world in which these directed agents can communicate with one another in order to accomplish more complex tasks.
Here’s an example:
Picture a manufacturing company with four distinct systems.
- Sales and order management (CRM)
- Marketing automation systems
- Inventory management and purchasing
- Financial management
In each instance, there is a single agent training on these systems and learning how to automate and intelligently interact with the core data and functions from these systems. The purpose is to transform our communication with each of these systems from a binary relationship (input, output) to a dynamic one where we can simply prompt these systems to produce certain outputs or results. Let’s look at them individually.
- Sales and order management (CRM): Agents now have the ability to update contact records dynamically, generate reports and alerts, and generate tasks and behaviors that would otherwise be done manually.
- Marketing automation systems: Templates can be auto-populated to incorporate content and information about client offerings and connect with prospects or existing customers through social media, emails or dynamic chats.
- Inventory management and purchasing: Agents can automatically generate reports and alerts in real-time to predict shortages, bottlenecks and pricing concerns.
- Financial management: Agent models can retrieve financial information that finance departments and operations can utilize to make purchasing, hiring and debt management decisions.
Now imagine each of these agents communicating with each other in order. The possibilities are mind-blowing.
In each instance above, agents are replacing manual processes. In some cases, they are efficient enough to displace certain job functions. But the business intelligence required to program these systems to collaborate and make decisions normally reserved for humans is where this enters the sci-fi realm.
This is the piece everyone is talking about. And it’s going to happen sooner than most of us expect.
Getting to this moment, however, is going to be rocky and this is where the fundamentals come into play.
If you think critically about the single-agent environment transitioning to a multi-agentic universe of decisions and actions, there are two fundamentals required that every company should be thinking about, which are the top and bottom layers.
On the top we have the decision process. How these actions are programmed and coordinated must be built and coordinated by those who understand the business use case and how these things have traditionally been accomplished. The industry talks a big game about “no-code” overlays, agent building and intelligent prompting, and those are elements. But what this translates into is humans generating the solutions, stitching them together, setting parameters and rules, monitoring outcomes and interpreting the results based upon industry-specific knowledge and experience. This cannot be replaced.
Perhaps the more important layer, however, is the bottom layer. The data architecture. This is what agents are being trained upon.
Content. What information is contained in your knowledge base?
Contacts. How are records stored and managed? Do you have standard naming and field protocols? How are records input, updated and enriched?
Workflows. What are your sales and marketing triggers? What are the inventory replenishment levels? How is pricing and quoting incorporated into business and purchasing decisions?
Integrations. Which systems will you choose to build these workflows? Which information can be trusted and continuously sourced so data are timely and updated?
Parameters. Who sets the hierarchy of decision-making data in the organization?
Permissions. Who has access to the data?
Performance. What are an organization’s success metrics? Who sets them?
Security. Are your core systems—and by extension your agent solutions—secure and industry-compliant?
The overarching lesson, whether we’re speaking about the top knowledge and performance layer or the bottom-line data architecture, is as old as time: Garbage in, garbage out.
AI holds great promise to transform the nature of work, but you still have to do the hard stuff.
-min.gif?upscale=true&width=960&upscale=true&name=giphy%20(2)-min.gif)
-min.gif?upscale=true&width=1120&upscale=true&name=giphy%20(3)-min.gif)
