Enterprise AI Use cases

“The future is already here – it’s just not very evenly distributed.“ - William Gibson

There's one thing more exciting than tomorrow's advanced AI capabilities. Distribution.

And by distribution I really mean broad adoption of AI technology. AI will power apps in every industry that will make everyone's life better.

The AI boom we're experiencing is really about one thing: ease of access. Generalised capabilities mean the realm of AI is not limited to experts, data scientists and ML engineers anymore. Even the most inexperienced developers can build AI powered applications with basic API calls.

But what sort of apps are people creating? What are the use cases relevant to your industry? I see these sorts of questions a lot.

There's 3 types of use that generative AI unlocks that I think most businesses will want to start with.

• Get - Conversational Information Retrieval
• Create - Asset Development
• Do - Actionable Intelligence

1. Get - Conversational Information Retrieval

ChatGPT brought chatbots to the forefront of everyone's mind in 2022. It felt better than google search but was missing specific context for most commercial use cases. The next question on every business leader's mind was:

"How can I tailor this awesome experience to my customers and employees?"

There is a wealth of data trapped in most businesses. Product engagement metrics that are hard to understand, buried confluence pages or gargantuan policy documents documents Conversational AI helps surface data that is relevant to the individual interacting with it. This means it can take on a new role as a completely new type of user interface.

Private AI is at the centre of advanced information retrieval due to the confidential nature of the data being processed. An organisation's data remain's private and is not used to train, tune or augment commercial or OSS models without consent. Moreover, strict access controls can be put in place to ensure that a model only has access to information that the user requesting said information can access. Cutodial control of a Private AI enables organisations to control its outputs to meet string content guidelines, remove bias and harmful hallucinations.

Customer Conversations

A conversational AI should be able to answer customer specific questions about a business' product, services or industry. Make it easier for customers interacting with your business to educate themselves.

Example questions:

• Product features - "Can I use an offset account with this loan?", "What other options are there?"
• Find and explain documentation - "How do I authenticate with your REST API?"
• Actions - "What's your returns policy?"
• Digest terms of service - "Can I use your API for commercial use?"
• Live status - "When is the next doctor available?"
• Triage problems - "I was overcharged this month, how can I get this resolved?"

Employee Conversations

Likewise, in large enterprises employees can struggle to find answers to domain specific questions or find internal documents. The answers to specific questions may be buried in confluence pages or obscure strategy documents.

Example questions:

• Policy - "Can I expense my home internet?"
• Status - "Will feature XYZ be in next month's release?"
• Company directory - "Who should I talk to about ABC?"
• Information - "What is [Internal Project Name]?"
• Direction - "I saw our company has a new AI strategy, what are the top 3 things I should focus on as a network engineer?"

2. Create - Asset Development

Organisations will want to take full advantage of generative AI to create new assets for their organisation. Generative AI is at the centre of any asset development a company does. The creation process is fraught with IP and copyright risks. Organisations need to be assured that the model's they use do not infringe on others IP while creating a walled garden for their own creation process.

A lot of creative tasks will be augmented by Generative AI models. Lowering the barrier to entry for beginners and accelerating experts.

• Code development
• Content generation
• Articles
• Powerpoint slides
• Brand marketing
• Test development
• Product Prototyping
• UX design
• 3D Modelling
• Legal contract creation / redlining

3. Do - Actionable Intelligence

What's the point of all this data if you can't use it to generate insights to support business decisions or automate tedious processes? These applications will analyse historical and real-time data to forecast future trends, identify opportunities and assist data-driven decision making. New tools that help people and machines 'do' more things, faster. Operational support from a LLM, will either completely automate or assist human operators with digested insights from real data.

This "Do" category further extends the notion that a language model is the next UX people could interact with. Let's see how simple information retrieval could be extended to "do" something.

"Show me the 3 cheapest flights to Europe in the next 2 months"
Flight 1 - SYD to LHR - 12/4/24 - $880
Flight 2 - SYD to CDG - 23/4/24 - $950
Flight 3 - SYD to LHR - 2/5/24 - $982
"Book the second flight"

Here we assume the user is already logged into their account and the system has access to payment details.

This retail example is simple in comparison to some of the huge automation capabilities we can expect. AI agents, specialised predictive models and machine vision will come to dominate high impact applications.

Examples:

• Auditing
• Quality Control
• Predictive analytics to support business intelligence
• Customer operations
• Automated Decisions - e.g. Loan decisions
• Manufacturing
• Supply chain monitoring
• Logistics
• Reporting
• Bookings
• Project Management