SUBSCRIBE
The time has come to move on from chatting with artificial intelligence (AI) to putting it to work for your business. The technology is rapidly changing, and many of us are still learning.
The barrier to deploying useful AI agents is not technical skill but clarity of thought. You must break the work into focused chunks to translate AI’s capabilities into meaningful outcomes.
During its early days, my landscaping company leveraged communication as a competitive advantage. It was our strategic response to one of the top complaints among landscaping service buyers.
Our phones were answered by a live person. They recorded and relayed information to sales and design teams and, if necessary, to an experienced manager with the authority to act.
The “click-now-to-buy” world in which we live has created new expectations. Many people no longer expect or want to speak with a live person. This shift presents an opportunity for AI agents that promise to act with the speed and accuracy that human beings cannot match. That promise is real, but it may require changing how you think about the technology.
Train it and test it
Companies that win with AI agents will plan for outcomes and develop the learning systems to attain them. Give AI access to resources with clear instructions and let it do what it does best.
Some examples of what it can do include:
- Qualifying
- Onboarding
- Categorizing
- Comparing
- Filtering
- Rejecting
Once you know your goals, get started. Give it access to data and tools to learn with a low-stakes pilot project, such as categorizing old leads. Each iteration is an opportunity for focused training. When it misses the mark, try to understand why errors were made, and don’t hesitate to push it to do better.
Then circle back to your instructions and revise them as necessary, keeping track of revisions. The progress you make in-house will minimize extra work and rework with real leads and clients.
Straight to the source
Here are the most common data sources and tools:
- Text strings
- Calendars
- Case studies
- Websites
Trigger words will prove valuable for making important exceptions. Be sure the agent assigns the same meaning to them as you have, along with specific action steps to take when encountering one.
For example, if a lead uses the word “emergency” or “storm damage,” the agent should be instructed to immediately escalate that to a manager rather than just scheduling a quote.
The early stages of training and troubleshooting AI agents can be overwhelming. Remember, you are building systems that will self-correct and improve with time.
Know the security risks
The landscape of AI agent security is currently like the early days of the internet. The superhighway has been built, but we’re still figuring out where the guardrails need to go.
HubSpot, Jobber and similar enterprise platforms are managed environments. They do use large language models like Claude, ChatGPT and Gemini, but these platforms act as a buffer. You should learn more about the extent of their protections regarding your intellectual property.
Ultimately, the transition to AI agents isn’t about replacing the personal touch that built your business. It’s leveraging it with powerful digital platforms to scale and ensure its relevance for changing customer needs.
