Business Resources

The Two Paths to Integrating AI in your Business

Justin Fineberg
Co-Founder & CEO
December 6, 2023

There are two primary categories for integrating AI into your business:

  1. Boosting Internal Productivity - In other words, figuring out how to make your employees and workflows more efficient, so you can deliver better results faster.
  2. Enhancing the Customer Experience - Exploring what you can offer on top of your existing products and services to make your customers even happier.

Focusing on internal productivity will help cut costs, while improving the customer experience will help you earn more.

Internal Productivity

In my view, it’s easier to start with internal productivity. Greater AI adoption within your team or even by yourself will lead to a substantial increase in overall productivity. The right AI strategy will turn everyone in your organization into a top performer.

Of course, figuring out the best AI strategy can be tricky. But, I think it’s important to not overthink it. Just starting to use AI tools and sharing them with your team can pave the way for in-depth AI integration across your organization. It’s crucial for everyone to have that “Aha” moment with AI.

At CassidyAI, we’re all about this mission. While we’re still slowly rolling out early access, we’ve been documenting our process publicly because we believe everyone can benefit from the thinking we’re putting into how to build features that truly speed up internal workflows.

We showcased our plugin feature that allows you to bring an AI assistant, custom-trained for your business, virtually anywhere on the internet. It’s been a game-changer for sales, customer success, and marketing teams.

(Check out the demo here. You can also signup for the Cassidy waitlist here)

The point is this: When integrating AI into your business, consider the workflows and tasks that currently exist and explore how AI could be introduced to enhance them. Never assume AI can't help--even small steps towards incorporating AI eventually become meaningful time savings.

Customer Experience

On the other hand, you can drive significant customer retention and growth by thoughtfully using AI in your products and services. This approach requires creativity, as it is not as simple as purchasing an off-the-shelf solution for internal workflow optimization.

AI-enabled features in your products and services could include personalized recommendations, an AI chatbot, or simple AI features that make it easier for people to use your product. The catch is that these features must be unique to your business.

For instance, Shopify customers struggled to optimize their stores properly, so their AI sidekick was developed to do it automatically. Another example: I recently spoke with a personal trainer who introduced an AI chatbot for fitness-related questions because clients struggled with guidance on days they weren’t seeing the trainer.

When considering AI implementation, think about the biggest pain points for your current customers. Why do people churn? Is something stopping them from using your product or service more often? What are their unmet needs?

It’s within these questions, you’ll find opportunities to enhance the customer experience using AI.

Read More, Learn More.

About AI

Which AI Model Should You Use?

Justin Fineberg
min read

Which AI model should I use?

This is the #1 question I get asked.

Well…why not all of them?

Over the last year, 3 AI models have emerged as frontrunners in the AI race: OpenAI's GPT-4, Anthropic's Claude 2, and Google's new model Gemini.

But not all of these models are the same. Each has their own strengths and weaknesses. That’s why I believe you shouldn’t just be using 1 model. You should be using every model. 

Let me explain why.

GPT models are reasoning engines

“GPT models are actually reasoning engines, not knowledge databases,” wrote Dan Shipper, CEO of Every. “Even though our AI models were trained by reading the whole internet, that training mostly enhances their reasoning abilities, not how much they know. And so, the performance of today’s AI models is constrained by their lack of knowledge”

Dan is spot-on. While most people are trying to use GPT models as encyclopedias—and have even run into fines in court from thinking GPT delusions were true—, they should be using GPT models as an assistant to speed up workflows and reasoning on projects

While GPT-3.5 quickly generates text, it lacks the nuance of GPT-4. GPT-4 can only handle 32,000 tokens, but it has extremely powerful reasoning and understanding skills. For most tasks, it pays to use the more expensive GPT-4 or GPT-4 Turbo if possible with 128,000 tokens.

One biophysicist Jeffrey Perkel wrote in a Nature article that using GPT-4 as an assistant lowered his time spent on some coding tasks from days to “an hour or so.” Content agency owner Randy Ginsburg said, “One of my favorite ways to use ChatGPT is as both a copy and developmental editor. It quickly catches any typos or spelling mistakes and often suggest new angles for the article that I previously haven’t thought of.” Notion engineer Linus Lee uses GPT-4 to come up with travel recommendations. 

Perhaps as GPT models increase in power, their ability as a knowledge database will increase too, but for now, it looks like GPT-4 will have to be content being an assistant.

Claude 2’s giant context windows

Do you want speed or quality? Ah, the everlasting question in business.

Claude 1 generates text in seconds, but the tradeoff is that its output is a much lower overall quality. “For use cases emphasizing fast turnaround over output sophistication, Claude 1 excels” explains’s Nathan Thompson. “But those needing higher accuracy, reasoning, and consistent quality are better served by Claude 2 or other models like GPT-3.5 and GPT-4.”

Because Claude 2 can handle 100,000 tokens (2-3 hours of YouTube transcript or an entire novel) and you need to pay for GPT-Turbo (can do 128,000 tokens), summarizing long pieces of content has become one of the main competitive advantages of Claude. 

Creator Riley Brown uses Claude to summarize 2-hour YouTube videos in seconds. AI strategist Christian Ulstrup tweeted about how he used Claude to summarize 5 of his meeting transcripts and give him follow-up questions to think about. Writer Darren Broemmer uploaded a large dataset to Claude and was able to ask questions about the data.

PaLM 2 → Gemini 

Since May of 2023, Google Bard has been running on the PaLM 2 model. 

Things weren’t going great for the tech behemoth. PaLM 2 has an 8,000 token context window and still isn’t getting massive adoption despite its ability to connect to the Google ecosystem. “Bard sucks and is pretty comparable to open-source alternatives,” tweeted Dan Shipper. 

Well, it looks like things are changing.

Two weeks ago, Google released a demo for AI model Gemini that took over the internet. 

If you missed it, the demo—which felt like a piece of sci-fi—showed Gemini's ability to reason and interact with the real world from identifying a rubber duck to creating games from scratch.

Reports show that the Gemini model is currently beating GPT-4 at 30 out of 32 categories. Google DeepMind CEO Demis Hassabi says "Each of the 50 different subject areas that we tested on, it's as good as the best humans in those areas.”  

Unfortunately Google won’t be integrating Gemini into Bard until 2024, but we’re counting down the days. Let’s hope they don’t pull the plug on this project 😂

Get access to all models: 

If 2023 was the year of everyone testing out GPT-4, then I think 2024 is going to be the year of everyone experimenting with new AI models. 

This is the beauty of competition, right? We're going to keep seeing better AI models released into the world. While I’ll always have a soft spot in my heart for ChatGPT, every model has its own purposes. GPT-4, Claude 2, and now Gemini each bring something unique to the table. 

That’s why you need to have access to all of them.

By using Cassidy, you can access all of these models in one workflow. No more switching websites and logging in and out of apps. No lost time or data. You can switch between GPT-4, Claude-2, Gemini, and even more open-source models to build assistants and workflows for everything from marketing to coding to product management. 

Ready to try Cassidy? Sign up here.

All Posts

About Cassidy

Meet Cassidy: Generative AI Solutions for Your Whole Team

Justin Fineberg
min read

Only 2% of adults in the US find AI “extremely useful”.

Does this mean AI isn’t useful?

No, of course not.

The problem is that most people don’t know how to use AI effectively in their jobs. They don’t know the best prompts to use, how to build AI-assistants for specific tasks, or have workflows they can use as templates.

The truth of the matter is that in order for AI to be useful in the business world, businesses need to do more than just handing their employees a ChatGPT account and saying “Good luck”. For 98% of people, that’s not enough to go on.

To boost productivity, we believe every business must have:

  1. A thorough prompt library
  2. AI-assistants for different tasks
  3. Workflows customized to specific business processes

But this is an huge amount of work for a business to do by themselves. So how are businesses supposed to actually adopt AI?

Introducing Cassidy:

Cassidy is revolutionizing how businesses adopt AI.

Anyone can use Cassidy to create powerful AI-assistants and workflows trained on your company’s data without having to write a single line of code.

That means you can use Cassidy for:

  1. Marketing: Craft on-brand content using Cassidy’s knowledge of your writing styles, social media content, and marketing goals
  2. Sales: Craft on-brand cold outreach using Cassidy’s knowledge of your sales pitch, process, and customer profiles
  3. Customer Service: Deliver accurate responses using Cassidy’s access to your support knowledge.
  4. Engineering: Make better technical choices and clarify codebases by training Caassidy on your tech stack, engineering methods, and architecture.
  5. Product: Drive product ideation and roadmap collaboration with Cassidy’s thorough knowledge of your team and UX.
  6. HR: Enable employees to quickly search and access HR documentation for answers with Cassidy’s assistance.

This is how businesses actually adopt AI:

Here’s how to get your employees to adopt AI in 3 easy steps:

  1. Find the AI-curious in your company. Find the people who are already testing out AI tools and are excited by an AI-enabled future.
  2. Enable them with a platform like Cassidy to build custom AI-assistants and workflows. Every business is different, so each business needs customized assistants and workflows. With Cassidy, you can ****train your assistants on your data wherever it lives (Notion, Google Drive, Slack, and more).
  3. Make sure your whole team has access to these tools. Don’t just buy 10 ChatGPT seats and pray people use it. Give everyone in your company access to a platform like Cassidy. That’s how you’ll see true automation com alive in your company.

If you want your business to actually adopt AI, join our waitlist here.

Welcome to Cassidy.

All Posts

About AI

OpenAI Fiasco: What did we learn?

Justin Fineberg
min read

I'm going to recap what happened over the 5-day span, then tell you what I learned from the entire ordeal.

November 17: OpenAI announced CEO Sam Altman was fired. OpenAI President Greg Brockman also leaves. The world goes nuts. No one knows why Altman was let go. Rumors fly around Twitter. Did he do something bad? Did AGI scare the board? No one has a clue.

November 18: OpenAI's investors are furious. Investors put pressure on OpenAI's board to reinstate Altman. The board agrees to reverse course and resign, but nothing happens.

November 19: Emmet Shear, the co-founder of Twitch, is announced as the new interim CEO of OpenAI. Altman tweets a selfie at OpenAI's HQ wearing a guest pass.

November 20: Altman and Brockman announce they are joining Microsoft to lead a new AI research team. Microsoft CEO Satya Nadella says anyone from OpenAI can join. 700+ of OpenAI's 770 employees threaten to resign from OpenAI. Hundreds of OpenAI employees tweet "OpenAI is nothing without its people."

November 21: OpenAI announces Altman is back as CEO. A new board is announced including ex-Salesforce CEO Bret Taylor, former US Secretary of the Treasury Larry Summers, and Quora founder Adam D'Angelo. Brockman returns as President. We are so back.

What did we learn?

For a moment, we all thought AI as we knew it was coming to an end.

Because countless AI startups are reliant on OpenAI's models, founders were getting worried if OpenAI goes down, would their startups go down too? It was a scary few days for a lot of people.

This is why ultimately, I believe businesses are going to want to use many different AI models depending on the task. You may want to use Anthropic's Claude for one purpose, OpenAI's GPT-4 for another, and even X's new Grok model for something else.

This is why we built Cassidy to support a bunch of different models, not just OpenAI.

All Posts

One Platform. Unlimited possibilities for all.
Join the waitlist.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.