How we launched a startup with AI in every workflow

Introduction

Good evening everyone, I'm Stefano Liviano, I'm CEO and co -founder of Rappen.

Rappen is a new 2026 startup that is focused on groceries and we help thousands of people in Switzerland to save on their groceries.

But today I'm not here to talk about my app, but I'm here to talk and show how we actually put together

there are literally in every workflow AI to launch our startup.

So let's start straight away.

I will

Core Idea: AI as the Whole Company

start with a bold statement.

AI isn't one tool.

AI is the whole company.

Building the Product with AI

So I'll start first with

the app front -end and back -end and as well the website.

Development Stack and Guardrails (Front-End vs Back-End)

So we use mainly cloud code,

cursor and of course anti -gravity visual studio codes whoever like whatever you

prefer and basically we have only one rule in this stuff the rule is that

everything that is front -end you you can just build and vibe code as much as you

want but everything that is back -end I'm lucky enough to have very true true very

good technical developers that actually check each line of code that goes in the

back end so given this we basically built some agent within our workflow of the app the app

Automating Data Ops: Normalizing Retail Offers and Ensuring Quality

collects offers from seven different retailers and of course you can imagine that is a huge

amount of data also different between each others and you need to normalize them streamline the data

and make sure that anything is missing otherwise the app will look very bad the data quality will

be very bad and this is actually a full -time job this is called data ops okay or data quality so

basically this job in our company is totally automatized we have several ai agents in a

workflow that check the quality of each offer that gets in our app and make sure

that people see only what is true and we have a mismatch rate of way less than one percent

Instant Multilingual Translation for Switzerland (Including English)

Connected to this, there is also a job that couldn't be done in the past.

And this is like the translation, not because there were no translators.

So imagine like you had even in the past the best team of translator ever.

What we are doing all internally is getting the offers.

And as you know, Switzerland has mainly foreign official aqua, including Romancio, but we

We have mainly German, French, and Italian.

That is actually the language in which all the supermarkets

publish their offers, right?

But we don't find the offers in English.

But you can imagine that even with the strongest translator

team a few years ago, you couldn't have done this.

Instead, in five minutes, literally,

since the offer get out, we translate all of them

also for the English -speaking population.

that is actually like the immigration here is around 25 plus percent so

Receipt Scanning and Spend Tracking with Vision LLMs

another feature that we build also thanks to AI is the scanner spending

tracker so very soon in our app you will be able to just take a photo of your

receipt even a bad one and AI will basically get old data you will get what

you bought when you bought if it was in offer with supermarket what time so right

Right.

And basically these to think even this few months ago,

few years ago, like would have been a mess.

Like imagine building an OCR pipeline for the most technical of you without AI.

And basically you need to set up everything in a database,

match every categories in the, in, in the receipt.

This could be like very,

very, very, very hard work.

And, uh,

Of course, you will have a lot of not categorized items.

For example, supermarkets usually put new items, right?

Instead, with some very light and cheap model nowadays,

Vision LLMs, basically you can automate this workflow

using just one LLM that recognizes your SIP

and as well classifies and categorizes everything.

thing another big use that we did is actually like the content and visual for

example this world presentation the slides that you're seeing were built by

cloud code in around five minutes okay I just took my time to give the prompt to

say what I wanted to speak about and that's the result it's not bad it's not

the best but actually like it saves you a lot of time rather than building

slides and moreover we use as well Gemini and as well HLGPT

especially in the last week after the update and now as well the cloud

designer for our website and socials.

Using AI for Business Operations

Market and Competitor Analysis at Speed

Another use of AI that I do almost on a

weekly basis is the market and competitors analysis and market and

competitor analysis is something that back in the days would have take for new

startups around two three weeks maybe one month if you wanted to do very

carefully now we can take up to an afternoon so apart from the like very

simple stuff that is deep research you can use deep research on chat GPT cloud

code and whatever perplexity I will give you another example on how to use it you

can do something that before would have taken an immense amount of time that is

is actually check your competitors.

For example, in my case, I'm launching an app.

I'm sure there are some competitors on the app

in the Play Store.

So what they do is identify my competitors,

and then I go on the review section,

and there are tens of thousands of reviews

in all the different languages.

What you can do in literally a few seconds, few minutes,

you just copy paste all the comments you give to an LLM,

and you immediately will know what to develop,

which feature are missing which feature the people like the most and like the

least and so on you can basically understand so much about our your market

in few seconds that before would have take you months and another very

Legal Workflows: Terms, Privacy, and Incorporation (With Human Review)

important usage that I found very useful in the last months is definitely the

terms and privacy and the legal in general like you can build like legal

document for example terms and privacy documents or even like use document

that is with a legal advisor usually like imagine you are a new startup you

starting like in a totally new market you don't have a huge budget of course

most probably you cannot afford a very good legal advisor so I'll do just a

small disclaimer here like AI doesn't substitute legal advisors yet right but

the point here is that if you are using AI you can avoid at least 95 % plus of

the problems that would be legal problems for your companies my personal

workflow is for example I have a cloud code agent that is actually trained on

every part of my company especially on the legal side and basically this agent

create for me documents and whatever that is connected to legal then I put

whatever the agent created to chat GPT for a review then I put it back to

cloud and after the last review from cloud it's good enough for human review

that is me in this case and of course especially legal documents always read very carefully

Synthesizing User Feedback to Pivot Faster

everything and last but not least i really like ai to visualize and pivot fast at least with my

product so basically um i will do another example on visualizing pivot fast in our app we launch a

survey that is uh basically was asking for ideas from the user what they like what they don't like

about the app and so on and we receive hundreds of answers in all the four

different languages I don't speak all the four different languages here in

Switzerland but I just copy -paste all the reviews I got and instantly I had a

roadmap of where to pivot fast where to pivot first where like what to build and

And what is easier, and trust me, if you are a first -time start -upper like I am,

it's basically a game -changer because AI has way more experience than you on this.

The New Reality: AI Shifts the Bottleneck

So the world's pitch here is not about just how useful is AI

and how easy it is to found a start -up nowadays in 2026.

Actually, the opposite.

Building the product is not the hard part anymore, okay?

Okay, so if you're building a product nowadays with AI, it's super easy, okay?

But now what is super hard and became super hard is finding the product market fit and getting funds on the marketing.

1AI didn't make startup easier, it just moved the bottleneck.

That's what I'm trying to say.

SEO Deep Dive: Strategy, Demo, and Early Results

And today, I will do a small demo and a deep dive on actually the SEO.

Basically, I choose this topic because it's something that every business needs, even

physical business, not just apps.

I will just show our strategy that in less than three months has already brought some

interesting results that I will show you now.

What the First Three Months Look Like (Search Console Signals)

Google, this is just the data from Google search, so this is not our website data, it's

just what Google brought to us.

Google brought to us 1067 strangers, 71 % search

for non -branded content.

The peak day, I know it's not much, but was 45 research

that led to clicks into our web page and our growth from beginning since

beginning of February is around 15x.

The interesting signal here is

not the numbers we are just at the beginning usually seo takes years to build but usually

most of the three months old startups get 80 percent branded traffic it means like the people

are looking just for the specific branded name in my case rappin and usually these are friends

investor family and whatever instead with rap and with our startup we flipped the ratio we got 71

21 % of non -branded clicks and only 29 % of branded clicks.

What Changed in SEO: Winning by Answering Questions

So how do we earn this click?

Basically in short, by answering questions.

And here I will show you a very quick demo as well.

So let me first explain what changed in the SEO game since AI is out.

Before, as you can see on the left, in order to rank on the first page of Google you need

to be a very established brand and work on the SEO for several years, invest a lot of

money in sponsored content and so on, right?

Nowadays you need to answer questions.

It's easy as that.

If you bring value to Google, Google is built to answer questions from people.

If you bring value to Google and to the users in general, to the people, Google will pick you up

to be the answer, actually, right?

why this has changed?

Because

since ChatGPT launched the web research directly from the chatbot,

basically,

the LLMs are accounting for a bigger and bigger part of all the global research.

So Google had to adapt as well, and as well you have AI search in Google, you have a short snip at the beginning of the page with AI, and you have as well Gemini.

Workflow Demo: Export SEO Data → Agent Writes Articles → Publish in Four Languages

So let me quickly show you actually how do we work with SEO.

So for example here, we have created an article on one of like Costco retail, it's called Allegro here in Switzerland.

and this is just one of the article we wrote and we have an agent of course

trained for this with all our data now lower strategy and as well while

building the article the agent has all this data okay that's the most important

part of the strategy don't make the agent guess exactly like this meta title

titles metadata and so on okay you don't need to make it guess you just need to

give all the data you can get in order for the agent to pick the right words

and the right meta description in order to rank faster on Google so as you can

see here we we just I'm just showing my Google search results and here you can

see of course more than 1 000 researches with all different keywords you can see as well the pages

who perform the best and you have this button here export on top of the page you just click that you

do the same for being webmaster take seconds download you download all the data about your

your SEO, you paste in your agent,

and your agent actually builds you the article.

In our case, we have the app in four different languages,

so this is built in all four different languages.

Once this is done, I just copy -paste,

copy -paste to Visual Studio Code with Cloud Code,

and here it looks like, of course, it knows what to do.

It's connected to our webpage code,

and you just link an image,

because I want an image in my blog,

and you can link this document we just produced and voila like this you have

four different pages in four different languages that are optimized for your

SEO ranking was very easy right so basically we thanks to this strategy we

we managed to grow very fast just to put everything in perspective we start in

February we have 1 ,000 clickers of now just in the last 28 days we go to 700

hundred clicks so this stuff is working as of now for us and that's why I wanted

Closing Thoughts: Small Teams Can Compete at Scale

Two Quotes on the Speed of Change and Business Transformation

to share with you and I want to keep to finish my presentation citing two very

influential voices in the AI world the first is the CEO of Microsoft Satya

Nadella and I will cite him now unless your rate of change keeps up keeps up

with what is possible you are going to get schooled by someone small being able

to achieve scale because of these tools.

These are dedicated to my competitors

but there is a very interesting thought about this.

We are three people, me and

two technical founders, that we are all working our jobs, we are all three

part -times and what we built is a very compelling product that is competing

with our competitor in Switzerland that have more than 100 employees here in

Zurich and just of course we don't have their budget we don't have their team we

don't have their history of seven plus years in the market but what do we have

here we have a three people team that built a product with more features than

them better than them in my opinion this you will judge as well and actually we

can offer to the partners and to the potential clients and users a better

a product for less, we need to pay just three people, we don't need to pay 100, right?

So that's where the competition is going.

And connected to this, I will continue with a second citation about Eric Kutcher, that

is a chairman from McKinsey.

This is the most complex business transformation we have ever seen.

But it's 80 % business transformation and 20 % tech transformation.

CEOs who sit and wait, their companies aren't going to exist.

I know it sounds like a very bad read like this, but he doesn't mean your company is going to get destroyed.

He means your company is going to get destruct, destruct.

So how?

I was doing before the example about the SEO between CharGPT, OpenAI and Google.

Google, as you can see, Google business model was always the search, right?

CharGPT entered in a market that was almost a monopoly, right, from Google.

And what happened, Google had to change its whole business in order to compete with CharGPT

because AI changed the game, totally.

It was not only about tech, it was about product and what the product can offer.

and this is true that now you can basically build products that even one

year ago you couldn't build and I want to finish my speech saying that if a

small team build an app in few months with four languages and tens of

thousands of offers with AI in every workflow this stuff was not possible 12

months ago okay and who knows if in 12 months will be still possible to compete

because the people who will start today of course would be ahead of the curve

and even six months ago if you think about cloud code was not so good

everything was not so good that you can delegate even like full tasks like to

Conclusion: Ask How You’ll Be Disrupted—and Adapt Fast

agents right so my final message here is if you have a business or you work for a

business or you're building any kind of business the first fault you need to

to have once you get home is, how is my business

getting disrupted?

Which features are now possible that were not possible before?

That's the first.

And second, you need to adapt fast.

Otherwise, some small team like mine

will pick up these small features

and going to compete with you.

And the competition gets very fast nowadays.

So thank you very much for your attention.

If you have any question, I'm here for minutes,

minutes or we can hang out later and I'm happy to talk with all of you.

Thank you very much.