Anthropic Co-founder and CEO Dario Amodei (Photo by Chance Yeh/Getty Images (GETY) for HubSpot (HUBS))
Getty Images for HubSpot
Anthropic’s recent announcement to offer tools to help Claude handle specific tasks in specific industries throws open the question of how the AI industry will be structured. So far, a few artificial intelligence engines from big companies have powered multiple specialized applications provided by small companies. But perhaps those large companies will soon sell all the artificial intelligence services that any company can need.
AI Industry Market Structure
AI large language models are now dominated by OpenAI (ChatGPT), Google (GOOGL) (Gemini) and Anthropic (Claude). Many small companies provide specialized services across a wide range of industries and tasks. Typically a user logs in to a website designed for specific users; that company processes the user input to create a request that is sent to one of the big three LLMs; then the company formats the LLM’s response to serve the end user. The small company has domain-specific knowledge that might include industry practices as well specialized databases, perhaps even including the end-user’s own data. The end user need know nothing at all about AI. I described this industry structure two years ago.
Economic factors drive industry structure. Some industries, such as aircraft manufacturing, are dominated by a few very large players. Other industries, such as restaurants, have a multitude of small competitors. Key issues are market size, economies of scale and ease of entry or exit, with other factors playing minor roles.
The recent product announcement could upend the current market structure. One example that Anthropic provided is, “Turn a folder of lawsuit documents into a chronologically organized exhibit set with descriptive titles and strategic importance assessment.” I performed a Google search for “AI for lawsuit document review” to see what offerings already existed. There were four sponsored results, suggesting companies trying to capture that business. The first non-advertising result was a year-old article entitled, “How AI Enhances Legal Document Review,” on the American Bar Association’s website, explaining how a number of companies were providing the service. I don’t know if Claude can match these companies’ services. A cursory review suggests they offer a broader array of services, more closely matched to attorneys’ needs, than does the Claude plug-in, but I’m no attorney. However, Claude’s direction is clear, and its capabilities will almost certainly broaden.
Although legal services received the most media attention, consider all the business functions that can be improved. An electrical contractor uploads blueprints for a new building and an AI service prepares the bid. A hospital employee sends a list of next week’s surgeries to an AI purchasing agent for all the products that will be needed. A supermarket chain connects a weather forecast help determine next week’s produce needs by store. Right now, all of these decisions require very specific understanding of the industry and the task. But AIs are very good at learning.
The advantage that propels the small, specialized companies today is intimate understanding of a particular function. A mix of people, industry veterans and AI-conversant programmers, can develop exactly what customers need to do individual tasks. In the past, that knowledge was acquired through years of experience. AI could, potentially, bring down the cost of entry to where the AI giants can offer highly specialized services as well as the broader LLM results. There is also the possibility that the big AI companies will offer preferential access to their own specialized services, to the disadvantage of the independent specialized companies.
Business Choices For AI Services
The easiest AI choice will be by companies with hundreds of people doing essentially the same job. Think customer service representatives or sales people. The business can start with an off-the-shelf app, customizing as needed. More challenging is when a business has only one or two people doing a particular job. Today they should go hunting for an app that is specific to the particular job, but works across different companies. Most challenging is a job that is very unusual.
AI consultants might be the best choice for companies with one-off needs. Current tools may be modified, sometimes easily, to fit the particular job. But many good employees don’t have the knowledge to identify the base app and perform any necessary adjustments.
These are early days of artificial intelligence. Services available change quickly. The best business advice today is probably to not lock a company into one provider. Those currently using a service company, who are satisfied with the relationship, should probably review services and price annually. Something better may come along in terms of quality, and companies often drag their heels on cutting prices when customers sound satisfied.
Companies that wonder if they could benefit from a specialized service should try experimenting. See what an AI-powered service can do, and also see what can be automated using one of the inexpensive chatbots.
Most importantly, look at the work done by employees. Any office work that is repetitive and tedious should be a prime candidate for AI.
Comments
Log in or sign up to join the conversation.