Mario Keller

Marketing, AI, Automatisation, Crypto 🚀

[Mario Keller @mario_keller]

Your AI could run on your hardware for example with local LLMs. This is my personal experience: https://t.co/CKLZyYvkg9

[Mario Keller @mario_keller]

Interested in agents that learn, improve, and actually handle your workflows?

We're opening beta access for Getsper.

Read the full story: https://t.co/ZsfeNy6L6u

[Mario Keller @mario_keller]

Building Getsper. We want to hear from you:

- What workflows frustrate you the most?
- What have you tried to automate that didn't work?
- What would your dream workflow automation look like?

Reply or DM. Genuinely interested. https://t.co/ZsfeNy6L6u

[Mario Keller @mario_keller]

Automation doesn't mean replacing people.

It means shifting them from "person who executes steps" to "person who designs and improves systems."

Your team gets smarter work. The workflows get smarter execution.

Everyone wins. https://t.co/ZsfeNy6L6u

[Mario Keller @mario_keller]

A rule says "if X, do Y."
An agent says "I understand the goal. I'll figure out how to achieve it, and I'll get better at it next time."

That's the difference between brittle and resilient automation. https://t.co/ZsfeNy6dgW

[Mario Keller @mario_keller]

Your team spends 20-40% of their time on repeatable workflows.

Most companies have documented these processes.
Almost none automate them.

That gap is costing you thousands per employee per year. https://t.co/ZsfeNy6L6u

[Mario Keller @mario_keller]

Some of my thoughts on MCP Servers for AI Agents. Why AI Agents Are Useless Without Them https://t.co/tclMFqAvUL

[Mario Keller @mario_keller]

@Shivam25mishra True. Notion has more features and better in UX. When it is about working together with AI Agents on some documents, Obsidian brings more flexibility.

[Mario Keller @mario_keller]

We're in alpha. Building Getsper in public.

Learning what works, what doesn't, what you actually need.
Your feedback shapes Getsper.
Join the beta. Try it. Tell us what you think.

We're building this together.
https://t.co/ZsfeNy6dgW

[Mario Keller @mario_keller]

@1980Aktuell @Mrkilometer1987 very fancy ✨

[Mario Keller @mario_keller]

Imagine:
- Daily reports generated (agent improves accuracy each day)
- Lead workflows handled (agent refines over time)
- Content creation assisted (agent learns your voice)

All with agents that improve with each run. https://t.co/ZsfeNy6L6u

[Mario Keller @mario_keller]

This changes who does what.

Instead of employees executing repetitive steps → teams designing workflows

Managers supervising agent improvement → people focusing on strategy

Your team becomes architects, not executors. https://t.co/ZsfeNy6dgW

[Mario Keller @mario_keller]

AI agents as workflow executors, not rules.

We give them:
- Clear inputs & goals
- Access to your tools & systems
- Guardrails (structured workflows)
- Memory (they learn from each run)

Agents improve over time, just like new hires. https://t.co/ZsfeNy6dgW

[Mario Keller @mario_keller]

What if workflow automation worked like hiring a new team member?

You don't hand them a rulebook. You explain the goal, show them the tools, let them learn by doing.

Over time, through feedback & repetition, they get better.

That's Getsper. https://t.co/ZsfeNy6L6u

[Mario Keller @mario_keller]

You can use n8n, Zapier, Make. They're great.

But here's the catch: they ask you to define EVERYTHING through rules.

Your data has variation? Add more rules.
Edge case? More conditions.
Over time?
Rules-based automation breaks on reality.
https://t.co/ZsfeNy6L6u

[Mario Keller @mario_keller]

Every company has documented workflows. Onboarding. Daily reporting. Lead nurturing.

Almost nobody automates them first.
This is the problem I started Getsper to fix.

Read more: https://t.co/ZsfeNy6L6u

[Mario Keller @mario_keller]

You don't onboard a new colleague with a 1000-page rulebook. You share docs, show examples, give feedback, they improve.
Why do we do the opposite with automation?

That's the problem I'm solving.

[Mario Keller @mario_keller]

All automation tools ask the same thing: Define 1000 rules upfront.
Predict every scenario. Maintain endless configs. Update when business changes.
That's not automation. That's management overhead.

What if agents learned like colleagues instead?

https://t.co/PHwj0LGXRL

[Mario Keller @mario_keller]

@jayjay_md231 Oh, I just activated it. Can text me now

[Mario Keller @mario_keller]

I tried everything. ChatGPT. Claude. Mistral. N8N. Zapier. Azure. Python. Custom agents.
Result? 10% real automation.

Everything else still needs me. Minute-to-minute management. Manual reviews. Decision-making.

Here's why, and what I'm building instead.
https://t.co/PHwj0LGXRL

[Mario Keller @mario_keller]

@jayjay_md231 yes of course. Looking forward to discuss some ideas