AI

AI-enabled web browsers like Dia and Atlas from OpenAI are what’s new in AI this week. What makes them different? They enhance your web browsing experience by adding an AI tool on the side, allowing you to chat in real-time with the web page you're viewing. We’ve talked before about the benefit of adding context to your AI prompts, and these new browser tools are basically doing that step for you. Here’s a great example: We have a customer with numerous Amazon reviews, and they would like to analyze that data. But Amazon doesn't have an export reviews feature. With these new browser tools, you can visit the Amazon review page, open the ChatGPT interface, and ask it to provide a summary of the reviews in a format compatible with an Excel spreadsheet. We’ve all had the experience of searching for a recipe and having to scroll through pages of ads, stock photos, and stories about grandma to find the information you need. This marketing-driven experience may become an artifact of the internet age, fading as these AIs [...]

We’ve heard numerous conversations from experienced developers stating that the new AI capabilities and tools are transforming how they work. But one thing to keep in mind is that there's a lot of coverage of those success stories and a lot of articles extrapolating that for one experienced developer working on their own. The claim is that because AI can bring phenomenal increases in one person's productivity, so therefore, a team of 10 people will see a proportional increase in productivity. However, that might not be entirely true. A Forbes article recently stated that in their survey group, at least 40% of employees received "workslop" in the last month. What is "workslop"? -- that's the term for when AI is used to generate plausible-looking but incorrect output, and a human being passess that work off as their own. It ends up being work that a coworker has to redo. AI is great at creating something that appears, on the surface, to be exactly the deliverable you need. But when you look into it deeper, one problem we commonly [...]

MCP stands for Model Context Protocol. MCP is what allows an AI to be augmented with extra abilities. We talked in a previous post about retrieval augmented generation, adding context to assist an AI task. MCP is an extension of that, allowing your AI to do connect to other tools and tasks like entering sales notes into your CRM database or composing a draft email. The MCP Project describes it this way: “Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect electronic devices, MCP provides a standardized way to connect AI applications to external systems.” MCP is an open protocol that works like an API, which means you can give your AI permission to connect to tools like GitHub or Figma and interact with those tools you are already using in your business. MCP can help you automate workflows and let your business systems work together in new ways that are tailored to your business needs. Want to learn more? Check out our podcast: Episode #4: Is AI [...]

The idea of a robot version of yourself taking over the drudgery of your job while you are playing video games or watching TV is pretty appealing. But can AI really replace an employee? We’re pretty sure the answer is no. At least not yet. When we talk to other business owners, a common theme we hear is that we all have many creative people we enjoy working with who are doing good work for our business. The problem is that they're spending a huge amount of their time doing work that they don’t like. Small businesses especially ask employees to wear a lot of hats, and they don’t always love the hats they have to put on. Often, roles like data entry or content creation go unfilled because there isn’t a person to fill them. These roles can be ideal for an AI “employee”. The simplest solution is to use AI automation to minimize the drudge jobs and help your employees do more of what they like. Using AI to fill the jobs you can’t find a [...]

A recent Gartner report stated that 61% of modern B2B buyers prefer to buy without human contact. That means more and more business owners are negotiating complicated, nuanced business interactions without talking to a person. Imagine you have a landscaping company and you are buying a new fleet of trucks. You can go online and research all of the features and specs. You can see what other field service owners are using and read reviews. You can make a decision: these are the trucks I need. Are you going to call the company to place an order? The trends say probably not. Any business with a sales line has seen call volume drop to a trickle compared to what it was 10 years ago. When you call a salesperson, it will take time, and they will try to upsell you or sell you the warranty. Why spend all that time when all you want to do is type in your credit card number and finish the transaction, because you’ve already made your decision.  As a business owner, your [...]

No matter what field your business serves, there are tasks we all have in common: the need to create content for presentations to generate sales or report to an invested constituency. You have the data and the templates created by your designer, but the perpetual struggle is how to get past that first blank page.  Why not ask ChatGPT to make the slides for you? Realistically, AI isn’t great at generating slideshows yet. It doesn't always understand visual data, like graphs and charts, or how to fit information into a graphic template. It looks for patterns it recognizes and might skip over important details specific to your business. But you can train the AI how to help you get past that blank page syndrome. First, give your AI some key context to help it succeed: upload a slideshow template that you have already created and the notes about what you want to talk about. In our last blog post, we introduced the idea of augmented generation to give your AI access to specific and relevant information. Then give [...]

You might have heard horror stories about companies that have replaced their help desk team with an out-of-the-box AI chatbot system. But much to the frustration of customers, generic chatbots can give generic answers.  Here’s a great example from one of our readers: “I was looking for a replacement part for my sewing machine on the manufacturer’s website. I couldn’t find the information, so I asked their Live Chat Bot a question. The answer it gave me was to go search the manufacturer's website I was already at.” When chatbots are built, they start by feeding the AI all of the information it can get access to, indexing every website, manual, book or pdf it can find. This is a very time-consuming process. So, one problem with training an AI this way is that it can only generate answers based on the information it has at the time. The entire AI model is based on that training data, so if a new piece of information is added, the training must start again from scratch. This is why many [...]