100% practical · 1 day · in small groups
Agent Challenge: your teams build their first AI business agents in one day
Twelve participants, working in subgroups, define, build, and test an agent on a real-world scenario from their field, then collectively select the agents to be deployed. This can be done on your studio (Copilot, ChatGPT, Claude, Gemini, or Mistral), either on-site or remotely, in French or English.
Second step in the agents' journey, already underway at Bristol Myers Squibb (15 Agent Party(134 registered, 4.71/5), Alptis and CNES. 4 professionals trained by NextStart since 2023.
From €2,000 excluding VAT per day · Written proposal within 24 working hours, without obligation.
The Agent Challenge, in brief
Everything you need to decide, in nine lines.
- What it is
- A practical day in subgroups Twelve participants work together to define, build, and test an AI agent on a real-world scenario from their field. They then present their agents and collectively select those to be deployed. No lectures or demonstrations: the agents are tested at the end of the day.
- What you get out of it
- Each subgroup leaves with an agent tested on a real-world business caseand your organization with a matrix of uses to be industrialized as a priority, ready for an Agent Hackathon or direct deployment.
- For who
- The professional teams of large organizations who have already practiced generative AI, ideally after a Prompt Party or an Agent PartyFifteen career fields to choose from, from marketing to management.
- Duration and rhythm
- A day : 45-minute masterclass, 5 hours of group work, a 45-minute prioritization challenge, and 15 minutes of reflection. Second step in the agent training program, between theAgent Party and the Agent Hackathon.
- Where and in what language
- At your premises anywhere in France, or remotely. French or English.
- Price
- €2,000 to €3,000 excluding VAT the day (up to 12 participants), depending on the personalization of the cases with your referents. See price benchmarks
- Which studio
- Yours: Agent Builder (Copilot Chat or M365), ChatGPT agents, Claude's Projects and Skills, Gems of Gemini, Mistral agentsDust, Prisme.AI, or your secure in-house studio. We train within your tenant, without third-party tools.
- Funding
- Eligible for OPCO and ESF+ funding. NextStart is a registered training organization (No. 11756140675): the day is part of your skills development plan, with the agreements and documents expected by your funder.
- Lead time
- Written proposal within 24 business hours, without obligation.
Build your first AI business agents
One day to move from a recurring task to an agent tested on your real cases, then choose together which agents to deploy as a priority.
- Defining an agent use case: input data, task, expected result and delivery format.
- Write the instructions: logical sequence, rules of behavior, safeguards and output format.
- Connect the knowledge sources and tools authorized by your IT department.
- Test on real-world and borderline cases, then present and prioritize the agents.
The four stages of the day
Masterclass, 45 min
In plenary session: what an agent is, what distinguishes it from a repeated prompt, its components (model, instructions, sources, tools) and demonstration.
Agent building, 5 hours
In subgroups: 2.5 hours in the morning to frame and design the agent, 2.5 hours in the afternoon to test and make it reliable.
Challenge, 45 min
Presentation of agents in plenary session, positioning on a shared matrix and selection of agents to be deployed.
Taking a step back, 15 min
Strengths, weaknesses, ethics, regulations, security and additional resources.
In the morning, frame and design the agent
Define the use case
- A recurring task in your job
- Input data and expected result
- delivery format
Prepare the knowledge
- SharePoint files, folders and sites
- Internal document databases
- Web sources and messaging services
Choose the model and tools
- Standard model or reasoning
- Web search and code interpreter
- Connectors authorized by your IT department
Write the instructions
- Logical sequence of tasks
- Rules of conduct and safeguards
- Expected output format
In the afternoon, test the agent on your job
Each subgroup tests its agent on real-world cases within its domain, then on borderline cases: incomplete data, pitfalls, and requests outside the scope of the task. Instructions are adjusted until the desired outcome is achieved. The cases are tailored to your specific context and designed with your experts, across fifteen business domains of your choice.
🎯 Marketing
- Studies and trends
- Benchmarks and targeting
- New offers and visuals
🗨️ Communication
- Multilingual content
- Audiovisual production
- Communication plans
💼 Sales
- Appointment preparation
- Arguments and calls for tenders
- Minutes and reporting
👥 Human Resources
- HR Data
- Policies and frameworks
- Workforce planning and regulatory monitoring
🤝 Recruitment
- Job descriptions
- Announcements and selection grids
- Questions and integration
🎓 Training
- Educational engineering
- Screenwriting
- Evaluation and promotion
📈 Finance
- Data analysis
- Anomalies and VBA
- Dashboards
🏷️ Shopping
- Sourcing
- Tenders
- Analysis and negotiation
🏛️ Legal
- Contract compliance
- Risk areas
- Notes and monitoring
🧪 Search
- Documentary synthesis
- Data mining
- Patents and designs
✨ Innovation
- Group interactive Ideation
- Future-oriented scenarios
- Prototyping and markets
📅 Project
- Performance and risks
- OKRs, KPIs and roadmap
- Reporting and improvement
✅ Quality
- Compliance audit
- Processes and procedures
- Anomalies and corrective plans
🧭 Consulting
- Diagnosis
- Organization and processes
- Recommendations and roadmaps
🎯 Management
- Strategic analysis
- Objectives and dashboards
- Coaching
Introduce and prioritize agents
Each subgroup presents its agent in plenary session: the task addressed, instructions, sources, and test results. The agents are then positioned on a shared matrix, for example, Time Saving / Quality Improvement, to choose which ones to deploy as a priority or to explore further in the Agent Hackathon.
Personalization and pedagogy
Collect
Preliminary survey and understanding of needs.
Adapter
Program design and validation.
Experiment
Kolb cycle: practice, observation, conceptualization and transfer.
transmit
Presentation and distribution of the training booklet.
Tools and methods
- Environments: your tenant and the agent studio validated by your IT department, without third-party tools.
- Studios: Agent Builder (Copilot), ChatGPT agents, Claude's Projects and Skills, Gemini Gems, Mistral agents, Dust, Prisme.AI or your secure internal studio.
- Modalities French or English, in person or remotely.
- Funding: eligible for OPCO and ESF+ funding.
The day is prepared with your contacts and adapted to the tools authorized by your IT department.
A professional and proven training program
NextStart.AI supports large organizations in the concrete and responsible adoption of artificial intelligence.
Bristol Myers Squibb, 2026
15 Agent Party One hour at headquarters, 134 registered participants and a score of 4.71/5: the stepping stone to the Challenge. Program, results and detailed feedback in the BMS France case.
Alptis, April 2026
Animation kit Agent Party designed with Copilot (Agent Builder), then deployed autonomously by 31 AI Champions: the same agent journey, driven by your ambassadors.
The Challenge extends theAgent Party for one hour or half a daybefore theAgent Two-Day Hackathon and the Agent Studio.
What you will produce during the Agent Challenge
Three agents of the NextStart libraryOpen access: each sheet provides the connectors and sources, the triggering frequency, and the fallback procedure in case of failure. During the day, your teams implement them on their own tools.
Spending anomaly detector
An agent who reviews the expenses incurred and identifies the lines to be checked before validation.
Regulatory monitoring
An agent who follows the regulations of your sector and only sends what is truly relevant to you.
Weekly campaign reporting
Each week, campaign figures are consolidated and commented on, ready to be shared.
Price benchmarks
The precise quote depends on the number of subgroups, the studio used, and the customization of cases with your referents.
Eligible for OPCO and ESF+ funding. Written proposal within 24 working hours, without obligation. See the detailed pricing by format and number of units →
FAQ
Who is the Agent Challenge aimed at?
For business teams who want to build their first AI agents on real-world scenarios, then collectively choose which ones to deploy. The scenarios are tailored to your context, across fifteen business areas of your choice, from marketing to management.
Is it necessary to have followed a Agent Party before ?
The Challenge extends theAgent Party The session lasts an hour or half a day, allowing for an initial hands-on experience. However, the day begins with a 45-minute masterclass on AI agents, the content of which is tailored to your teams after a preliminary survey.
What's the difference with a Agent Party ?
The Agent Party The Agent Challenge allows each participant to build their first agent in an hour or half a day. It goes further: a full day in subgroups to define a use case, write instructions, connect source code, test on real-world and edge cases, and then collectively prioritize the agents to deploy. It directly prepares participants for the Agent Hackathon.
What tools are used during the day?
Using the agent studio already available in your organization: Copilot Chat or M365 Agent Builder, ChatGPT agents, Claude's Projects and Skills, Gemini's Gems, Mistral, Dust, Prisme.AI agents, or your secure internal studio. The day follows the recommendations of your IT department and is prepared with your designated contacts.
Does our data remain in our environment?
Yes. We provide training within your tenant, using the tool validated by your IT department, without any third-party tools: the agents built and the sources attached to them remain in your environment.
How much does an Agent Challenge cost?
From €2,000 to €3,000 excluding VAT per day, for up to 12 participants, depending on customization. Written proposal within 24 business hours. Our training courses are eligible for OPCO and ESF+ funding.
Is it taking place in person or remotely, in French or English?
Both options are available: the day is conducted either in person or remotely, in French or English. A training booklet is provided to participants.
What happens after the day?
Each subgroup leaves with its tested agent, its documented configuration (model, instructions, source code, and tools), and the training booklet; your organization also receives the prioritized use case matrix. The selected agents can be deployed directly or further developed in an Agent Hackathon, and then industrialized in Agent Studio.
What happens after my request?
You will receive a personal response within 24 business hours, followed by a 30-minute discussion about your teams, use cases, and agent studio. You will then receive a detailed proposal and the training agreement. You can also call 01 89 16 48 05.
Launch your Agent Challenge
Describe your context in a few lines: teams involved, intended use cases, existing tools, and deadlines. We will get back to you within 24 business hours with a tailored proposal. You can also book a 30-minute consultation directly.
Do you prefer to call or write? +33 1 89 16 48 05 or hello@nextstart.aiYour contact details are only used to reply to you.
NextStart is a training organization registered under number 11756140675. This registration does not constitute state approval.