Stop Building Another AI Feature. Build a Better Integration.

AIAIAIAI

Artificial Intelligence is a word just flowing with smart intelligent conversations and huge expectations. I recently worked with a few companies that came to me asking how they could implement AI in their workflows. I told them just to wait, the technology isn’t ready. The response was well everyone’s using it, so what’s so difficult about it? They found tools that had AI in their name and we tested it.

It was a huge disappointment.

Why?

Expectation.

For most people, when we say “AI” we think big, massive productivity enhancing, growth-exploding tools that will change the way we work and fix the broken way we interact with computers.

Remember when computers came out? The general populous of interviews I watched was saying how revolutionary it will be, humans will be so efficient we will barely have to work. John Maynard Keyes’ 1930 economic essay Economic Possibilities for Our Grandchildren claimed that because of innovations would only be working 15-20 hours a week because of technological innovation.

Where did that go?

Why is AI Still on the Fringes?

Take a look at how ChatGPT got where it is now.

Information.

They vacuumed so much data, both copyright, non-copyright, everything. ANYTHING. It pulled it in, and put it in its database of potential solutions driven by a model that can pick out the nuances of the question a user is asking it.

So why is AI still on the fringes for most people?

Integrations and information.

Just the same as content providers shift TV shows every few months to keep you paying for every networks streaming service, every company that deals with any sort of information wants you to pay for their AI features and be the de-facto company that you go to for your needs.

It can’t happen.

Yes, we have kludgy integrations that can make working with AI within these tools better (I’m looking at your Apple Intelligence), but the difference is in the settings and the information.

With most tools, I can set it up the way I want. There’s a fixed set of settings, with knobs and dials I can change till I’m satisfied with what I am interested in. If it’s unable to do what I need, I can simply find a tool that works the way I want.

Same with information, usually I can import data to a new tool, or add new information to set it up so it has exactly what I need when I need it, and be the only source of truth for that category of information.

With current AI, you’re talking to it in human words. It learns your likes and dislikes like a person would, in long-form words and sentence structure. People then adjust their results based on that feedback, just like the AI learns based on how you teach it. Most AI tools don’t have knobs and settings to adjust. That’s why Prompt Engineering is a real job. The problem is, every company has an AI that they want you to teach, and the people who are trying to use it are not trained prompt engineers, just a user who wants a correct answer, “like AI sales groups promised”.

We don’t have that time or money to teach each and every AI from each and every tool we use all our likes and dislikes.

For instance, I use a lot of tools for work. Each one of them has an AI feature integrated into it. Why wouldn’t they. It’s an important part of the growth for these tools, offering value to their customers.

I also only pay for 1 of them.

Trying to get all these AI tools to work together on top of a kludgy interface with no general settings or personality import while also paying for the same OpenAI model I already pay for is insane and I won’t do it. If I pay for another AI feature, I’ll pay for it for it’s basic generative tooling, and nothing more.

Doesn’t this remind you of the problem we had with Alexa? “What’s the weather like?” was about all it ended up being good at.

What I Want from AI

If companies want to TRULY innovate don’t charge me for another generative AI tool based on the same model I already pay for, but now integrated independently with your tool.

Charge my for an integration to a tool I already have and have taught all my likes and dislikes. Charge me for that calendar integration. Don’t make a new calendar program I have to use to gain all the benefits of AI in my calendar, that doesn’t quite do exactly what I was hoping for.

The future of AI is not going to be in the interface, it’s going to be in how well they integrate with literally every other tool.

Remember how OpenAI got so big? They vacuumed up all the data from all the sources, no matter what format, endpoint, copyright restriction or database.

Why will that be hard? Because most companies base their current metrics on user count and usage counts based on a user interface they created.

If you want your app to truly get on the AI revolution don’t build another slick UI. Build a slick agentic AI integration that is so open I can pass in my preferences from Apple AI, from ChatGPT, from NotionAI, and let your special sauce of a program be able to whip up something so amazing I have to use it. Not spend my time flitting between tools because it’s not quite what I wanted.

I don’t want another tool that can tell me what the weather is, or be another glorified search engine. I want what AI promised me. A new future of productivity.