AI Assistants vs AI Agents

The future of AI in improving small business efficiency is the emergence of AI Agents. You may already be using AI assistants, like Siri, Alexa, Co-pilot and other assistants throughout the day. Assistants are reactive. That is, they must react to a prompt. Agents, on the other hand, can operate independently. This is because they are able to use a set of work instructions connected within the computer via APIs (Application Programming Interface). This is a means for the agent to interface autonomously with other applications. All the heavy lifting of regular business functions can be coded into a series of steps that run autonomously or on-demand.

The Power of Agents

Agents can respond to high level commands. For example, suppose that sales of your product is under-represented within a certain age bracket and you think you are missing growth opportunities. You can instruct an AI agent to “Plan a marketing campaign targeting Millennial prospects in California and update it monthly with new prospects”. The agent will take all the steps that a human might take and will generate a steady stream of potential new customers.

Marketing – An Example

To expand on how an agent functions, there are a lot of tools available to execute a workplan. Though the use of agents is still in the early stages, the structure will look familiar if you’ve ever run a prospecting initiative. Here is a typical outline.

• Input “Find Millennial prospects in [industry/location]”.
• Search Use web/API tools to gather leads from company sites, LinkedIn, or import a list from your CRM or select from a list that you have contracted for.
• Filter Sift the data for quality and relevance (e.g., company size, years in business, job title and add them to your CRM).
• Output Create ad copy, provide direction for creatives, and articulate a targeting strategy and frequency.
• Launch Initiate the campaign using an Advertising API (Facebook, Amazon, and Google all have advertising APIs). Alternatively, as a safeguard, you could export the data for review prior to launch. You could also experiment with A vs B headlines to see what approach gets the best response.

This series of instructions can be designed to pause at any step for review or adjustment before carrying on to the next step.

Pros, Cons and Managing Risk

This sort of flexibility comes with risks. These are some of the identified risks followed by ways to mitigate them.

• Spamming leads, poor messaging
• Recursive code can get stuck in infinite loops
• Access to APIs could entail security risks or unauthorized access
• Scraping or profiling people can violate ethical practices
• Agents can produced biased outputs

To reduce the risk of these things happening various strategies are usually employed.

• Add human-in-the-loop checkpoints
• Set limits (timeouts, budget caps, etc.)
• Periodic audits to track what’s going on
• Apply privacy safeguards and limit access
• Regularly review and test outputs

Other Uses and Classifications

The presence of agents is just the latest evolution of AI that is moving from the experimental stages to more commercially available plug-and-play options. AI is at the stage where it could be used to make hotel reservations, respond to customer service rankings (like Yelp), do lead nurturing, handle customer returns and so on. These are all basic business processes which are not too complex, but could free up some talent to be used at a higher level.

Agents can also be classified according to their planned structure and function. Reactive agents respond to inputs without memory (chess-playing). Model-based agents use internal models of the world (self-driving cars). Goal-based agents take actions to achieve predefined goals (marketing). Lastly, learning agents adapt to environments over time (performance improvement).

Keep Your Eyes on the Horizon

There are already out-of-the-box solutions (Auto-GPT, AgentGPT, OpenAgents) which require someone on-site to set them up. There are also custom agents that require custom development to integrate into your existing workflows. Lastly, there are Enterprise wide tools starting to appear. Two examples are Salesforce for customer services and Zapier AI agents for automating workflows. It’s worth keeping your eye out for these developments. The AI market is moving fast and putting some smart development tools into your business can give you some first mover advantages. This is especially relevant when it comes to marketing and customer service, two areas that help drive business growth and develop loyalty. Ready or not, there is likely an AI in your future. The time to start planning is now.

Photo Credit:

https://www.istockphoto.com/photo/ai-or-artificial-intelligence-chat-program-or-software-knowledge-and-learning-gm1470405426-501269844

Small Business Resources welcomes questions from inquiring minds looking to improve their business outcomes using best business practices combined with available technologies. Submit any questions you would like us to explore on your behalf to contact@sbresources.com.