EO AI Productivity Exchange · #1

The Setup
Trap.

Two operators show the stacks they actually run. Tool overload, setup cost, what survived 18 months.

Christoph Erler & Dom RauteMay 11, 2026 · 18:00 CEST

EO Berlin
01 / 17
60 Minutes

What you walk out with.

01
Clarity
Which of two stacks fits you, without burning weeks.
3-5
Hacks
Concrete, actionable. Wins and the struggles behind them.
103
Peers
Where other EO operators stand. Where you stand.
∞
Series
Seat in a continuous 30 to 90 day exchange.

Not the goal: another tool overview, generic theory, polished vendor demos. We show the struggles, not just the wins.

EO Berlin
02 / 17
The Data

Two traps everyone walks into.

Trap 01
Tool Overload
50%
of EO operators say "too many tools"
34%
say "I do not know where to start"
"Which one for what?"Sirah Pallas, EO Rhein-Ruhr
Trap 02
Setup Cost
95%
of AI pilots show negative ROI, MIT 2025
$50
per day Opus burn during build-up
"Weeks burned on setup before any value lands."Andrew, EO Best Guy interview

Most don't escape because they fight the wrong trap.

SourcesEO WhatsApp Poll (122 votes, March 2026) · MIT Project NANDA, "The GenAI Divide", 2025 · Andrew van Aswegen, EO Best Guy interview, April 2026

EO Berlin
03 / 17
5 Setups in 18 Months

Three dead.
Two live.

v1
ChatGPT only, no memory
dead
v2
Notion + Zapier orchestration
dead
v3
Multiple tools, no architecture
dead
v4
Workspace-native + memory layer (Chris)
alive 6 mo
v5
Local-first + 5-tier memory (Dom)
alive 12 mo

Two stacks survived. Both share one principle. Stop maintaining tools. Start maintaining memory.

EO Berlin
04 / 17
Diagnostic

Are you in the Setup Trap?

3 questions. Answer honestly.

01
Do you maintain your setup more than you use it?
More than 20% of your time? Setup Trap.
02
If your setup dies tomorrow, do you lose knowledge or just workflow?
Knowledge? Setup Trap.
03
Could a new team member understand it in 2 sentences?
No? Setup Trap.
EO Berlin
05 / 17
The Hierarchy
MEMORY>TOOLS>MODELS
MEMORY
what your system remembers
TOOLS
what it can execute
MODELS
which AI runs in the backend
90% of executives see no measurable AI impact in 3 years. Not because AI does not work. Because they discuss models instead of building memory.

SourceMcKinsey, "The state of AI: How organizations are rewiring to capture value", 2025

EO Berlin
06 / 17
Demo 01 · Chris · 9 min
01of two

AI inside your existing stack.

Two hacks. Same architecture. Different outputs.
Hack 01
Morning Briefing
Inbox + Calendar + Drive + priorities, one page in 30 seconds.
Hack 02
Meeting Prep
LinkedIn + past mail + CRM + Drive context, on demand.

No new tools. Orchestration on top of Gmail, Calendar, Drive, Notion.

EO Berlin
07 / 17
Demo 01 · How it works

Same bot. Same sources.
The prompt is the lever.

4 Sources
Gmail
Calendar
Drive
Notion
→
Bot
Orchestrator
Telegram-native, voice + text
→
2 Outputs
Hack 01
Morning Briefing
One page. 30 seconds. Read on phone.
Hack 02
Meeting Prep
One page. On demand. 5 min before any call.

Two outputs. Same architecture. Only the prompt changes.

EO Berlin
08 / 17
Demo 01 · Hack 01

Morning Briefing.

Trigger
> morning brief
Voice or text in Telegram. 7am every day, before I open Gmail.
Output, 30 seconds
Inbox top 3: Tim Fischer (Valtis follow-up), Diana, Raphi (RC ICP v2)
Calendar today: 11:00 Robert call, 14:00 focus block, 17:30 EO Exchange dry-run
Drive priorities: RC sales script v3 needs review
Open threads: Reisensburg expose, Stephan Big Decision
5h/ week
Tracked over 3 weeks, March 2026.What it does not do: prioritize by emotional weight. Bot does not know it's your kid's birthday.
EO Berlin
09 / 17
Demo 01 · Hack 02

Meeting Prep.

Trigger
> prep meeting with Tom Debus
5 minutes before any call. Used to take 20.
Output, 30 seconds
LinkedIn last 3: Ferris.ai launch, AI agents for finance, Q1 wrap-up
Past mail thread: 4 emails over 6 months, last topic was MCP integration
Drive mentions: 2 docs reference Tom, including Sales Exchange notes
Suggested opener: follow up on his Ferris launch + ask about MCP rollout
8h/ week
More than morning brief, because more meetings than mornings.Where it fails: meeting with someone new. No history, bot returns thin air. Then I do it manually like it's 2024.
EO Berlin
10 / 17
Demo 01 · What you just saw

Workspace-native, in numbers.

13h/ week
Total saved across both hacks
30s
Bot response time, both hacks
39
Workspace tools orchestrated
2h/ month
Maintenance tax, patching
Honest struggles
Maintenance tax is real, every Workspace update can break the bot
No emotional weight, bot does not know your kid's birthday is today
No long-term memory across sessions, fresh context every time
New contacts return thin air, no history to pull from

No long-term memory across sessions. That is where Dom takes over.

EO Berlin
11 / 17
Handoff

Dom takes you
one layer deeper.

From hack on top, to architecture beneath.

EO Berlin
12 / 17
Demo 02 · Dom · 9 min
02of two

The agent that remembers.

Two hacks. 5-tier memory. Zero monthly cost.
Hack 01
5-Tier Memory
System Prompt > Bootstrap > On-Demand > Search Index > Raw Archive.
Hack 02
Rolodex
107 person dossiers, communication style, last contact, open threads.

Local models. 1,183 indexed files. 2-5 sec queries. $0 API cost.

EO Berlin
13 / 17
Demo 02 · What you just saw

Local-first memory, in numbers.

$0/ month
Memory layer total cost
5,735
Vectors indexed across 11 collections
2-5s
Local query time, no API call
107
Person dossiers in Rolodex
Honest struggles
$50 per day Opus burn during build-up phase
Cold-start memory loss despite the 5-tier architecture
Manual rolodex drift, not auto-synced yet
Setup is real engineering work, not a weekend project

Architecture wins long-term. Setup cost upfront, near-zero compounding cost after.

EO Berlin
14 / 17
Cost Reality

What you pay vs
what you get.

$0/mo
Local-First
Self-hosted memory, local models, 2-5 sec queries.
$500/seat
Glean Enterprise
Hosted enterprise search, vendor lock-in.
$1.5K/mo
Enterprise Stack
Notion AI + Glean + ChatGPT Enterprise + others.

The architecture beats the budget.

SourcesGlean Enterprise public pricing, 2026 · Notion AI + Glean + ChatGPT Enterprise list pricing combined, 2026

EO Berlin
15 / 17
Live Tonight

11 solutions shipping today.

Tool Overload
29%
Team Adoption
18%
Agents in Production
18%
Integration / MCP
12%
OpenClaw Stability
12%
Pace / Keeping Up
10%
Setup Itself
9%
Healthcare
6%
Time / Learning
6%
Security / GDPR DACH
5%
MIT License · fork and remix
github.com/chris1928a/eo-ai-exchange
EO Berlin
16 / 17
This is Day 1
You name the pain. We build the skill.

30 to 90 days. Then again.

10
Pain Clusters
from 103 registrations
11
Solutions
live tonight on GitHub
#2
Event #2
topic picked by your Slido vote
∞
Cadence
market-reactive, not quarterly
Slido open 24h
#3261743
GitHub, live tonight
chris1928a/eo-ai-exchange
EO Berlin
17 / 17
arrow keys to navigate