We’ve all been in a situation where a senior leader on our team is retiring, and we’re worried about the decades of experience that person has walking out the door with them. Or we reach the end of a long project and want to be sure we capture all the key lessons learned so we have those sights for the next big thing. Today, you might save those learnings in a project file somewhere or present them to the team in a meeting as you turn the page and close out a project. I’m here to tell you that with the rise of artificial intelligence (AI), we have new tools at our fingertips that everyone in our industry should be taking advantage of. There is a better way to transfer knowledge, capture lessons, learn from our past work and deliver better results for clients in the future.
At Gresham Smith, we’ve begun piloting AI-enabled chatbots to assist with knowledge management and transfer. We think these bots have the potential to help us enable instant access to project data which we can use to find problems or trends and deliver better results in the future.
Before we dive deeper, I want to include a brief disclaimer. Without a strong data protection policy in place, you won’t get very far using just any kind of AI tool. At Gresham Smith, we only use custom models or protected channels with large language models, ensuring our intellectual property is protected.
Why Use a Chatbot?
For those who don’t know, the textbook definition of a chatbot—or “bot”—is a computer program that simulates human conversation to solve queries. It’s like a search engine, but better, because it’s more tailored to the specific needs of you and your team. A chatbot can shift through the information you feed it, learn from it, and grow smarter over time.
It also helps to deal with information overload, especially as manuals, policies and procedures are always changing. It’s available 24/7, meaning it’s always there if you have a question, and it allows a team member to get answers quickly. Additionally, a chatbot frees up that more experienced colleague you would normally rely on for answers.
Here’s how it’s done in two simple steps:
Gather the data. Start with the information you already have. Conduct interviews with key stakeholders and develop surveys to capture that data. Lastly, you will want to clean that data for specific uses.
To conduct interviews, first decide what information you’re seeking, and select the right people, as in the people that everyone goes to when they need answers. Schedule those interviews, record and transcribe. AI can help with that part, too.
After you’ve sat down with the key people, develop a larger teamwide or officewide survey to determine where the gaps are. Think high-level and brief. What information is missing? Where are the gaps?
Finally, clean and proofread the data. It’s only usable if the columns in spreadsheets are consistent, complete thoughts are captured and you’re using the most up-to-date information.
Train and refine the bot. Bots aren’t trained overnight. The process requires beta testing, good, old-fashioned trial and error, and debugging. Bots also require attention and oversight. A key point to remember is that a bot is only as good as the data you feed it.
Once you master the process, your old, dusty postmortem documents become living documents, with evolving spreadsheets of data. Over time, you can begin to introduce new data sets to help connect projects and help the bot understand new relationships.
