Joanna Leung

Product Leader | Passionate in solving problems valuable to people with technology |

AI Agent Starter Suite

π—–π—’π—‘π—§π—˜π—«π—§
Pragmatico trains mid-market teams to use AI at work. The curriculum starts with the basics of Claude and ChatGPT, then introduces learners to their first AI agent. As teaching material, each agent required hands-on experience for beginners on day one.

π—§π—›π—˜ 𝗣π—₯π—’π—•π—Ÿπ—˜π— 
Beginners at the start of their AI adoption journey often find AI Agents intimidating. They understand the tools but cannot yet visualize an agent at work.

π—›π—¬π—£π—’π—§π—›π—˜π—¦π—œπ—¦
β€œAs a mid-market employee new to AI, I need to see an agent work on my own tasks within minutes, so I gain the confidence to find use cases for my team.”

If a beginner can personalize a working agent in under 10-15 minutes using only skills they just learned, it will reduce the mental friction of using AI Agents and even building them.

π—£π—”π—œπ—‘ π—£π—’π—œπ—‘π—§π—¦

  • Blank page: beginners lack a starting task to use an agent.
  • Setup friction: connecting extra tools, files, or code drives early learner drop-off.
  • Generic output: an agent that ignores role and habits reads as a novelty, not a work tool to adopt into the learner’s workflow for sustained usage.

π—¦π—’π—Ÿπ—¨π—§π—œπ—’π—‘
Designed ten department-agnostic agents that handle tasks ranging from deep market research to inbox triage.

Each agent ships as a Claude/ChatGPT Project with project instructions and guides users to connect the necessary MCP connectors, the features learners just practiced.

Guided onboarding by design with a seed prompt collecting information from the user through a few questions, then personalizes the agent to their preferences.

No file uploads by default. Each agent pulls context from connected tools such as email. Minimum effort to maximize ease of use.

π—˜π—«π—˜π—–π—¨π—§π—œπ—’π—‘

  1. Learners receive a set of ten one-pagers, each with 4-step instructions to set up each agent, including a model selection recommendation to optimize token consumption.

Each agent ships for both Claude and ChatGPT and runs on the LLM the company subscribes to.

Onboarding Sequence:
2. User copies and pastes a prompt from the one-pager into a new chat in their preferred LLM. The user will be asked a set of questions that the Agent will use to generate a custom Project prompt to personalize their Agent.

  1. User will be instructed step-by-step to create a new Project and copy-and-paste the custom Project prompt, and how to use their personalized Agent.

Feedback Loop:
4. A patch mode lets the learner report what went wrong. The agent proposes the smallest rule fix and returns updated instructions. The learner triggers each fix, so output improves over time.

π—œπ— π—£π—”π—–π—§
Ten agents reached clients in cohorts of up to 200 learners. Each learner left with a working agent for a specific use case. Each agent went live in under 15 minutes on the company’s existing LLM.

𝗣𝗒𝗦𝗧-π—Ÿπ—”π—¨π—‘π—–π—›
Patch mode keeps the feedback loop user-triggered, so learners retune agent rules directly.

Role
AI Adoption Consultant; Builder
For
Employees of Mid-Market Companies
Date
April – May '26
Type
AI Agents, Claude Team, ChatGPT Business
URL
www.pragmatico.ai/
AI Agent Starter Suite
←
Next:
AI Agent Generator
→