Freelance Growth, Paid | +100 entreprises transformées | +5 millions générés | Formation Growth

Passionate about the human relationship and the new technologies, I acquired several years of experiences as a commercial. Over the years I have noticed an evolution of the buying process, buyers are primarily looking on the internet before contacting a company. Aware of the difficulties companies have in being online, I decided to help them in this transition.


Jeremy LEPLU

"L'IA vocale c'est le futur de la prospection."

Je vois ça partout sur LinkedIn.

Je suis pas d'accord. Et j'ai des preuves.
J'ai fait +15 démos ces 6 derniers mois. Honnêtement ce n'est pas encore prêt !

Quand j'en parle à des PME ou des boîtes avec 5 à 50 commerciaux.
À chaque fois que je parle de l'IA vocale, c'est la même réaction :
"On sent trop que ce n'est pas humain." "Ça ne fonctionne pas." "Il y a trop de latence"

Le vocal crée de la méfiance. Il force une réponse immédiate. Il met la pression.

Alors que le texte, c'est l'inverse.
Avec un WhatsApp ou un SMS :
Le prospect répond quand il veut
Il peut réfléchir avant de répondre
Il ne se sent pas agressé
On ne sent pas que c'est une IA derrière.
Tu crées une conversation. 

L'IA vocale deviendra peut-être crédible dans 2-3 ans. Les voix seront meilleures. La latence sera invisible.

Mais aujourd'hui ? Le texte gagne. Et de loin.

Si ton objectif c'est la conversion.
Choisis le canal que tes prospects préfèrent. 

Pas celui qui fait bien en démo.

Améliorez vos conversions grâce à l’IA sur Whatsapp | +100 entreprises accompagnées (Sécuritas, Getaround, La Première Brique, N2F, Eurotunnel, Kardinal, Addingwell)

I'm a founding partner at Highline Beta. 

Highline Beta is a corporate venture studio, working with large companies to innovate beyond the core. We help build new ventures internally and externally, and invest in pre-seed/seed stage startups that have a meaningful connection to our corporate partners.

I'm an entrepreneur, investor, product expert and author. I've worked in startups (VarageSale, GoInstant and others) and large companies (Salesforce), with the knowledge and experience to bridge the gaps between them.

In 2013 I co-authored Lean Analytics (http://leananalyticsbook.com), to help startups and large enterprises build better products faster with data. I've spoken globally and consulted with many companies around implementing Lean Startup and Lean Analytics.

In 2010 I co-founded Year One Labs, a pre-seed startup accelerator. One of our portfolio companies, Localmind, was acquired by Airbnb. Year One Labs was the first accelerator to apply a rigorous Lean Startup methodology.

Areas of expertise: venture studios, corporate venture studios, venture building, corporate innovation, startups, business model innovation, Lean Analytics, Lean Startup, product management, design thinking, startup investing, startup accelerators & recruiting.

Ben Yoskovitz

New year, refreshed brand. Highline Beta is approaching its 10-year anniversary (crazy!) and we felt it would be a good time to spruce up our brand and content.

So much has changed in 10 years it's amazing. One can only imagine what the next 10 years will bring...

Throughout our existence, a few things have always remained true:

1. We build ventures and startups with corporates and other partners that believe in the power of "beyond the core" innovation.

2. We invest capital in founders + startups that we help co-create.

3. We move fast, we're scrappy, and committed to maximizing value for our partners.

4. We believe in the fundamentals (Validate problems first! Test and learn!) but we adapt too, leveraging AI more and more each day to work faster AND smarter. We're also going to release more tools for people to use, such as PAIGE (our AI venture innovation strategist) and the Corporate Venture Studio builder (for designing a venture studio). Building stuff is at our core, and personally, I love it, AI has unlocked a ton of opportunity for non-developers.

Let's build better ventures together.

Founding Partner at Highline Beta | Author of FocusedChaos.co

The more I vibe code, the more I learn. What I've realized, ironically, is that you need MORE SKILLS to vibe coding successfully (not LESS), despite the ability for AI code assistants to write code.

How is that possible? Shouldn't the AI handle everything for you?

Well, it handles a lot of the code for you, but it also makes a lot of decisions too, specifically around system design. Every product is a "system" -- components that have to work together, data that has to flow, variables & rules that have to govern things, etc. AI tools often take the shortest path to build something, without necessarily designing the proper infrastructure.

A power tool used poorly, without a plan, doesn't eliminate the need for craftsmanship.

Some of the recent issues I've dealt with:

* Backwards compatibility of data. You decide to change a field or add a field, but what about all the previous records?

* Hardcoded variables. You have a field with X options in it. The AI may by default hardcode those options, which makes it tough to change them, or use them in other parts of the product.

* Rebuilding UI components. AI will rebuild AI components that are meant to do the same thing and look the same -- but make them work and look different (unless you tell it not to).

I've come up with 5 questions that I now ask myself before building something:

1. What is likely to change later? If it might change, it probably shouldn't be hardcoded.

2. Should this exist once or everywhere? If the same value, rule or concept shows up in multiple places, you need to design the system appropriatley.

3. What is the source of truth? Is this coming from config, the database, user input or assumptions buried in logic?

4. What breaks if I change this? The left hand has to know what the right hand is doing.

5. How would I test this? If testing feels awkward or overly manual, the system design is probably the problem.

AI loves the happy path. But if you've ever shipped a product you know users don't always follow it. In fact they often do unexpected things or uncover edge cases. AI building tools aren't well designed to handle edge cases or error management. And if you don't realize you need those things in the system design and then explicitly build those things, you'll have a brittle product.

I don't think you should go crazy on system design before building. The best way to learn is to build. So please: build more stuff! But I'm definitely experimenting with different lightweight "system design sketches" (think: simple Product Requirements Docs) that at minimum answer the questions above, to think through stuff before jumping in.

Skill-wise, if you're vibe coding production-grade software, you need to be a strong product manager, a decent designer, a system architect (sort of) and a code reader (so you can figure out what's going on). Oh, add QA to the mix too. You need a broader range of skills to leverage vibe coding tools successfully, despite the promise of automation.

Sometimes all you need is a pen and paper. Yup, crazy isn't it?

I love building products. It's made even more fun/interesting/challenging now with AI assistants that can help design, code, deploy, etc. 

But building something, launching it (even as a test), and watching tumbleweeds go by, still sucks. 😢 

And most people don't have the intestinal fortitude to realize the product isn't working it, drop it and move on. Plus, they often realize they didn't learn enough either, so they have no clue what's working, what's not, etc.

In the early days of any product / startup, learning is the key.

No learning = no progress.

So that begs the question: What's the fastest way to learn something useful?

Is it jump into an AI code building tool and launch a half-baked product? Maybe.

But at Highline Beta we like to go old school. Bring out the pen and paper...(or at least digital versions) and test sacrificial concepts.

A sacrificial concept is a deliberately rough, low-fidelity version of an idea. Often we start with a single piece of paper or slide that describes a concept (typically with the problem statement written out). There are no visuals. Just words. You put that in front of someone and get a reaction. It works. They're not focused on a UI, or 500 features, or bumping into bugs--they're just reading the problem statement + the concept/solution statement and reacting.

The goal? Provoking a response, not polite compliments. Think of it like bait. You're not trying to win. You're trying to get a reaction. 😀 🤩 😥 🤷‍♂️ (often, it's a shrug!)

We'll show users 3-5 concepts in a single interview (yes, we still interview people! I know synthetic users are a thing, they're not a replacement for now). The goal isn't to sell, but to explore which directions spark curiosity, tension or pushback.

I've seen a lot of teams claim they want to  build something sacrificial, start using AI building tools and then fall "in love"--suddenly they're iterating little things to "perfect" the product, having forgot the goal was to learn quickly and scrap it or double down. That's death by a thousand paper cuts.

Sacrificial concepts are hard to fall in love with, they're low fidelity, nowhere close to a solution and therefore easy to "throw out."

So don't rush to code (even with AI), despite the urge to do so (I get it, the pull is strong). Think about other tools you can use for the key job you have early on when building a new venture: learning. 📚

taplio

The Battle

Jeremy LEPLU

Ben Yoskovitz

Followers

Posts

Engagements

Likes

Performance ScoreJeremy LEPLU's LinkedIn Performance

Performance ScoreBen Yoskovitz's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Jeremy LEPLU

Ben Yoskovitz

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Jeremy LEPLU's LinkedIn Performance

Performance ScoreBen Yoskovitz's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreJeremy LEPLU's LinkedIn Performance