In partnership with

As it becomes baked into pretty much any kind of tech you can think of, generative AI is becoming abstracted away from the chatbots we tend to associate with the tech.

But while there are plenty of AI-based app-building tools out there already, Zapt wants to help businesses harness the flexibility of raw generative AI by creating a layer on top it to make large language models work for end users in highly specific ways.

Intrigued? Read on…

– Martin

Intercom for Startups

Join Intercom’s Early Stage Program to receive a 90% discount.

Get a direct line to your customers. Try the only complete AI-first customer service solution.

Apply now

Zapt wants to add a layer of simplicity on top of generative AI

Generative AI is gradually making a transition from free-text boxes that require you to know what you want and how to formulate a prompt, to more sophisticated implementations. See the ‘Apple Intelligence’ features coming soon to iOS for example.

Zapt wants to be part of this transition. The startup is building what co-founder David Fox describes as “a platform to allow anyone to build their own generative AI apps.

“We're targeting businesses, to help them optimise their processes, their workflows. Anything that is currently a quite manual task or requires a manual process, we want to allow them to bring generative AI in to speed it up.”

Fox sees Zapt as a way to make generative AI more accessible to more people. The point of it is to create a custom layer above raw large language model inputs.

He says the product can work with any underlying LLMs without the end user having to worry about what’s going on under the hood.

How it works

Fox took me through an example of how Zapt works. 

A business getting set up with the product does have the free-text boxes I mentioned at the beginning of the article. They use this to create the app they want to build for end users among their staff. Those end users then get the benefit of generative AI without the need for any experience with prompts, and a much more specific, hand-held experience.

The process of creating apps is quite an exploratory experience, Fox says.

“What we've been finding is people are not 100% certain what the use case is and what the capability of the platform is. So they start with a broad request or requirement, and they see what the app can produce. Then they try it out, and they're impressed with it. And then they go and tweak, and make changes. 

“All you need to be able to do is write down what you want to change, and the AI does it in about 30 seconds for you. So it means that anyone can do it. It's available to non-technical users, and it's very quick to make those changes.”

In an example Fox showed me, a business had developed a risk assessment app based on a body of data they had about how risk should be assessed.

Once the app had been created and refined, staff within the business could use it to speed up risk assessments via a simple Q&A interface. It’s not super-pretty at this stage, but useful business tools don’t necessarily need to be.

The story so far

Fox has a PhD in machine learning (earned 10 years ago, “before machine learning was cool,” he says). He previously founded Orenda, a fintech startup offering embedded financial services products.

“The idea was to get to a point where it was completely drag and drop, almost like Wix does for websites, but this would be ‘build your own bank’. We were actually doing the dragging and dropping behind the scenes, because people paid a lot of money for these programs and they didn't want to do anything. They just wanted to set the requirements.”

So while Orenda didn’t work out, the concept of connecting microservices together behind the scenes stuck with Fox. 

This led him to found MAPT, a startup we covered here at PreSeed Now in December 2022. That was designed to simplify the way developers build their backend systems.

But when ChatGPT came out, Fox was inspired to combine generative AI with his experience of machine learning and apps that connect things up for users without them having to get involved with the nuts and bolts.

And so he co-founded Zapt, initially as a consumer product. He had been a solo founder at MAPT, but for the new startup he brought in Phil Gold as a co-founder to focus on business development.

“The idea for Zapt initially came from watching people on social media saying ‘if you do this prompt, and then that prompt, you get this really cool output’,” says Fox. 

“And I was thinking ‘no one's actually going to remember what those prompts are; it would be much better if you could have this in an app that has these prompts under the hood.’

“Although we got 15,000 consumers to sign up, mainly via listings on AI app directories, we found that they were just playing around. Their use cases were not well defined, and we struggled to monetise them.”

That led Zapt to pivot to a B2B offering.

“The idea is to be a platform for companies to do all their generative AI,” explains Fox.

“You can't productise anything on ChatGPT, it's not going to work. You're not going to store prompts, you're not going to share prompts, you're not going to layer prompts. It's just too manual a process. Whereas an app with all of that stored behind the scenes and abstracted away becomes accessible and usable.”

Fox says Zapt has deliberately avoided a strategy of pre-baking apps for end-users. However, he says they intend to focus on a few sectors such as education and utilities, and create a few example apps to inspire them.

Fox says customers currently need a lot of hand-holding to make the most of the platform. That is one thing that will need to change for the startup to scale effectively. More on that below…

Subscribe to Premium to read the rest.