👾This AI Solution Saves $20K & Creates Consistent Deal Flow

Welcome to Edition #5 of Agents Made Simple

Another week, another round of new models. This time, we’ll spotlight open-source solutions that run locally on common computers and phones.

Many companies are adding agentic features to their services and workflows. I sketch a turnkey AI solution that creates immediate value. You can implement it right away.

This week’s topics:

• New models from Alibaba, Xiaomi, Microsoft, and Amazon

• Visa, Mastercard, and Yelp are integrating AI agents into their services

• Re-activate leads for consistent deal flow with this agentic system

• Run open-source models and agents locally

• Plus AI investments, community highlights, trending AI tools, and more

AI Agent News Roundup

💥 Breakthroughs

📈 Investments¹

🇺🇸 Cisco launched Foundation AI, a new organization at Cisco Security, dedicated to creating open AI technology for cybersecurity applications, planning to release models, tools, and data built from the acquired Robust Intelligence.

🇺🇸 Visa launched Visa Intelligent Commerce to enable AI agents to shop and pay, aiming to make shopping experiences powered by AI more personal, secure, and convenient on its payment network.

🇺🇸 IBM announced plans to invest $150 billion in America over the next five years to fuel the economy and accelerate its role as a global leader in computing, including investing over $30 billion in R&D and American manufacturing.

🇺🇸 Google announced that its podcast-generating service within NotebookLM is expanding to over 50 languages.

🇺🇸 Amazon launched the first batch of satellites for Project Kuiper into the lower orbit, aiming to build a large network of internet-satellites, competing with Elon Musk’s Starlink.

🇺🇸 Uber announced the deployment of thousands of robo-taxis on its ride-hailing platform across U.S. cities.

🇺🇸 Aurora launched their first driverless freight trucks, running their first routes in the U.S.

How to Build a Lead Reactivation AI Agent System

The Hidden Gold Mine in Your CRM 💎

Most businesses are sitting on a treasure trove of untapped potential: old leads. These contacts who previously expressed interest but never converted represent significant sunk acquisition costs.

Lead reactivation AI systems automatically re-engage these prospects through personalized, multi-channel communication at scale.

Core Components of a Lead Reactivation System ⚙️

An effective lead reactivation AI platform consists of:

1. Conversational AI - Handles natural language interactions across channels

2. Multi-Channel Support - Orchestrates outreach via voice, SMS, and email

3. Intelligence Layer - Scores leads, determines optimal contact timing, and personalizes messaging

4. Compliance Management - Handles opt-outs, recording consent, and regulatory requirements

5. CRM Integration - Updates lead status and syncs conversation history

Technical Stack 💻

The system can be built with varying levels of technical complexity:

No/Low-Code Approach

• Voice: Vapi or Retell for voice conversations

• Automation: Make, Zapier, or n8n for workflow orchestration

• Communication: Twilio for SMS or WhatsApp integration

• Database: Airtable or Notion for lead management

• Analytics: Metabase or Looker for tracking performance

Advanced Implementation

• Custom LLM fine-tuning for industry-specific conversations

• Hosting on Groq or Cerebras for ultra-low latency responses

• Self-hosted open-source models for data privacy compliance

• Custom advanced voices, recognition systems

• Dashboard to easily start, stop, or take over AI conversations at any time

Key Architectural Decisions 💡

When building a lead reactivation AI system, consider:

1. Complete Control - Allow human takeover of any conversation, and when AI encounters uncertainty

2. AI Memory - Conversation context is maintained between interactions and across channels

Value Creation Breakdown 💰

First-order consequence: Eliminates the need for multiple full-time follow-up staff, potentially saving $15,000-20,000 monthly in labor costs while increasing lead engagement rates.

Second-order consequence: Creates consistent deal flow from previously dead leads and builds valuable long-term relationships, increasing annual deal volume.

Tool Spotlight

👾 Run Qwen3 Open-Source Model Locally

Two popular methods: Use Ollama if you’re familiar with the terminal and don’t need a user interface. Go with LM Studio if you’re aiming for a similar experience to the ChatGPT interface.

Via Terminal with Ollama

Ollama lets you run AI models on your computer. It's free and easy to set up.

1. Go to ollama.com and download the right version for your computer (Mac, Windows, or Linux).

2. Once installed, Ollama runs in the background.

3. Open your terminal or command prompt and try these commands:

# Pull a specific Qwen3 model
ollama pull qwen3:8b

# Chat with the model
ollama run qwen3:8b

# List your models
ollama list

ℹ️ Try a smaller version first (fewer parameters) to test if it runs on your device.

The first time you run ollama pull qwen3, your computer will download the model. This might take a few minutes.

When chatting, type your message and press Enter. To exit, type /exit.

Using Ollama as a Server for AI Agents

Ollama starts a server on port 11434 automatically when installed. This lets you build AI agents that talk to the model.

Here's how to use it with Python:

# Simple API request
import requests

response = requests.post('http://localhost:11434/api/generate', 
    json={
        'model': 'qwen3:8b',
        'prompt': 'Write a short poem about AI.'
    })
    
print(response.json()['response'])

For building agents, you can connect Ollama to frameworks like LangChain:

from typing import List

from langchain_core.tools import tool
from langchain_ollama import ChatOllama


@tool
def validate_user(user_id: int, addresses: List[str]) -> bool:
    """Validate user using historical addresses.

    Args:
        user_id (int): the user ID.
        addresses (List[str]): Previous addresses as a list of strings.
    """
    return True


llm = ChatOllama(
    model="qwen3:8b",
    temperature=0,
).bind_tools([validate_user])

result = llm.invoke(
    "Could you validate user 123? They previously lived at "
    "123 Fake St in Boston MA and 234 Pretend Boulevard in "
    "Houston TX."
)
result.tool_calls

ChatGPT-like User Interface with LMStudio

1. Download LM Studio from their website for your computer

2. Install and open the app.

3. Go to the "Discover" tab (look for the magnifying glass icon).

4. Type "Qwen3" in the search box.

5. Find the Qwen3 model size you want.

6. Click "Download" next to the model.

7. Once downloaded, go to "My Models" and click on Qwen3 to load it.

Using LM Studio to Build AI Agents

LM Studio includes a server. This makes it easy to build AI agents. Example with Python:

1. Install the SDK using pip install lmstudio

For building agents with custom tools:

import lmstudio as lms

def multiply(a: float, b: float) → float:
    """Given two numbers a and b. Returns the product of them."""
    return a * b

llm = lms.llm("qwen3:8b")

llm.act(
  "What is the result of 12345 multiplied by 54321?",
  [multiply],
  on_prediction_completed=print
)

Community Highlights

Made By Agents Updates

Trending AI Tools

🦙 Ollama: A command-line tool that enables users to run and manage open-source LLMs locally, offering privacy and offline AI capabilities.

❇️ LM Studio: A user-friendly graphical interface that simplifies running, fine-tuning, and managing local LLMs across various platforms.

0️⃣ ZeroGPT: An AI content detection tool that accurately identifies AI-generated text using advanced algorithms, supporting multiple languages.

More Resources

Blog: AI-driven business automation and practical strategies for growth
AI Tool Collection: Discover and compare the perfect AI solutions
Consultancy: I help you solve your problem or discover AI potential
Follow along on YouTube, X, LinkedIn, and Instagram

See you next time!

Tobias from MadeByAgents

P.S. Was this useful? Have ideas on what I should publish next? Tap the poll or reply to this email. I read every response.