If you were having a 10K road race, trying to stand up on a tough hill, and suddenly the rules of the race happened, what would you do? What if the drivers start catching the ball in the car and then playing against each other? Will you continue running? Do you know you will put it behind the back of your backpack? Or get in the car, cheer up and compete for the grand prize?
In today’s business, AI is the car that disrupts the way companies operate. Companies still have the option to move forward the way they do – creating remote plans, sticking to processes, and pushing employees to work harder than ever to succeed in a competitive environment. But AI is changing the nature of race. It provides companies with a new car to move faster and provides workers with new routes to narrow down problems. Any business that does not have a steering wheel and instills the power of AI into its workforce will remain behind that long and steep hill.
Embrace the future by becoming an AI manager
In the cockroach lab, we quickly learned that AI can help us do things we never thought of. We have used it for Gen AI search, recommendation systems and semantic search throughout the company.
One of the best examples of how AI can change the workforce process is happening in our education sector. Our team is using AI to accelerate the development of a course that helps customers, partners and our own workforce become experts in the operation of our database product line.
We recently created a course that includes 21 hands-on exercises and 20 slide decks with detailed student notes. Before starting the project, we estimate that using our normal development process – considering how long it takes for a developer to produce an hour of content in an industry standard estimate – this will take three to five months to complete.
So, what happened? Incorporating Gen AI into our existing process, we are able to complete the task within five weeks.
We learned a lot of lessons along the way.
- We are all AI managers. Each of us has the opportunity to use AI to think differently. Each of us should act as a manager, whether we have direct reports or not, because we manage a nearly unlimited supply of intelligence capabilities that we can invest in challenging projects. How much can you automatically? How creative are you? How do you effectively prompt your AI tools, challenge and deploy the new models it generates? You can take advantage of it. You can manage it. You can essentially do as many things as your personal abilities can do.
- Don’t expect AI to do everything. There are some tasks that are not suitable for execution at all. However, you can task tasks to do things that workers shouldn’t do anymore – time-consuming work, but still require a certain level of intelligence.
- Don’t accept its results blindly. Check, check and recheck. Trust technology, but always verify – because accuracy turns assumptions into achievements.
Step by step process of deploying AI for task management
Here are some short summary of AI that helped us up the mountain, much faster than we expected.
- Different models: Different models have different advantages. So, just like manufacturers use the best components when building solutions, models can be exchanged at will if they make sense. We wrote the first draft of exercise using Claude Sonnet 3.5 because it is good at creating compelling prose and instructions. We used Chatgpt 4O & “O” inference model as technical reviewers to perfect the command and ensure technical accuracy in the second draft.
- Reproducible results: When performing highly technical tasks, we want to be able to enforce clear technical limitations and produce structured outputs that support reproducible results. To do this, we provide clear structural requirements and format examples.
- Prompt height technical tasks: Very specific about what you ask for AI to do –
Otherwise it can do crazy things. Any assumptions about input or environmental conditions are clearly stated and the model is required to handle unexpected situations.
- Refining tips: It is important to encourage AI tools to ask clarification questions. The first tip won’t be perfect, so you can expect multiple rounds. Combine any improvements or steps suggested by the model in your basic tips and iterate with AI and your teammates.
- Test, test, test: Consistency check is crucial. One way to measure the effectiveness of a prompt is to ensure consistent output. So we often do tests to make sure we put the same input in and the output remains the same.
Human expertise at the helm: The basic role of AI supervision
While AI removes time-consuming tasks from Workers’ Day, it doesn’t completely remove them from the workflow. Humans still play a key role in our curriculum development and need to be integrated into AI-driven processes to ensure the process is successful.
A good example is how our education team works in a timely manner. Humans are responsible for formulating initial tips, including context from relevant sources. Then, after the Gen AI tool executes the prompt, humans review the tool’s output. This person must be a subject matter expert and can encounter errors early in the process. Teammates continue to work with the tool and iterate until the team meets the prompts to be released.
Although this collaborative human/AI has proven to be effective, it does require context and transition between human management models.
Without humans, teams will be at the mercy of AI tools, which may be unreliable. When we first start with a course project, these tools really do a great job of summarizing or explaining the correct context of the concept. However, they often hallucinate. These models are better in reasoning these days, but humans still need to manage processes. Now, humans can focus on censorship and creativity, not just process management.
In the future, AI agents will play a greater role in this process. Instead of manually collecting context from sources, we use context-making tips to move work between AI models and review and refine outputs, we are developing agents that can perform many of these tasks, and with some help. Agents can autonomously collect and process source materials as environments, generate skills taxonomy and course syllabus, execute our established workflows, and make key decisions to human experts.
in conclusion
While brisk running is great for maintaining a body shape, cars from a long time ago turned human abilities into places to go. AI offers the same benefits in the workplace – helping companies improve processes and produce better results. Those who embrace it and take advantage of its compound efficiency will put competitors behind.
