Join the next Agent Builder Lab
Build and deploy production-ready agents, with your data and your rules. End-to-end deployment inside the Agent Builder Lab itself, on secure systems you can trust.
Who it is for
Professionals at every level and in every function. Building agents is becoming a core skill for all of us.
Format
Virtual, live via Zoom
Duration
3 hours per agent module, plus an optional 3-hour fundamentals session
Prerequisite
None. Agent Building Fundamentals is optional and taken once
Value Proposition
We co-build a custom solution with you that is ready for production in hours, and that you and your teams can later enhance and further develop independently.
A real agent, not a concept
Not a concept, not a demo, not dummy data: a real agent that is ready to go into production after our session.
Your governance and compliance fully accounted for
We show how to deploy all agents in a sovereign and compliant way: whether on aixplain, or using other platforms.
Your team fully enabled
Your people can edit and improve the agent independently once our session concludes.
How it Works
Three steps from assessment to a working agent.
Step 1
Sign-up and Assessment
You choose the agents you want to build. We send back a short survey and the exact list of what needs to be ready on your side (e.g., system access, data, approvals) so nothing blocks the session and your agent can be deployed inside the Lab session itself. Our team can help you with the preparations if necessary.
Minutes
Step 2
Agent Building Fundamentals
A one-time deep dive for first-time builders: how agents are scoped, instructed, equipped with tools, tested, and integrated safely with the systems you already run. You can skip this if you are already an experienced agent builder, or if you have already completed this module.
Three hours, once
Step 3
Agent Builder Lab
We build the agent you selected together, connected to real data, using a battle-tested pattern for that agent type. Every question gets answered as it comes up, and you learn to edit and improve the agent yourself afterwards. You deploy the agent you build in our Lab on aixplain cloud, or on a cloud or data-center that you provide.
Three hours, per agent
Technology & Compliance
We support you in deploying the agents you build in our lab in the way you need: on our cloud, on-premise, or your private cloud. Three deployment options, decided with you before the Lab begins.
aixplain cloud
Your agents are built and deployed on the aixplain cloud, running the aixplain agentic OS, our operating system for building, governing and scaling agents. It meets government-grade safety and security standards, gives your team access to hundreds of models, open-source and closed, and provides a full environment for your organisation to work in.
On-premise
We install the aixplain agentic OS in your own data center. Every query, document and log is processed there, and no third party, including aixplain, has access unless you explicitly grant it.
Your cloud
We install the aixplain agentic OS in your own private cloud. Every query, document and log is processed there, and no third party, including aixplain, has access unless you explicitly grant it.
All three are fully supported. The aixplain cloud is the fastest to deploy, so many organisations start there and move on-premise or into their own cloud later.
Program Structure
In each Agent Lab we cover five dimensions, to move you from idea to production within the session itself. Every module covers the same five dimensions, in the same order.
Scope
We define together exactly what the agent should do, and what should stay outside its remit. We start with a pre-defined scope and design, and showcase throughout the session how to edit and develop further from there.
Requirements
We identify every system and data source the agent needs, and how each can be connected safely.
Build
We configure together and co-design every artefact needed to deploy the agent: instructions, model choice, tools, guardrails, evaluation and checkpoints.
Roll-Out Strategy
We discuss strategies on how to plan for going from pilot to large-scale implementation.
Go-live
We go live with the agent, connected to real data and real systems, with clear governance and security protocols in place.
Agent Library
Illustrative agents from our library that we can help you implement. Select one to see the session details.
Executive Assistant Agent
Scope
- Daily email dashboard: tracks and analyses incoming email, showing what needs a decision, pending client requests, pending internal requests, and invoices that have arrived.
- Scheduling support: when a meeting request arrives, the agent checks the calendar against your scheduling rules and recommends possible timeslots.
- Meeting briefing: before a meeting, it reviews previous correspondence with the participants to help you prepare.
- Response drafts: it prepares replies to incoming email for your review.
In the first version, no permissions are granted for the agent to communicate with third parties directly. We show how that can be changed safely later, should you choose to.
Requirements: access to systems and data
| Source | Typical systems | Access and sensitivity |
|---|---|---|
| E-Mail System | Microsoft 365 or Google Workspace, delegated access to one mailbox | Only read plus draft-create access needed to your e-mail system. |
| Calendar | Same suite, read and free/busy scope | Read-only, for scheduling recommendations and briefings. |
Requirements: other requirements
| Requirement | Description |
|---|---|
| Mailbox in scope | The professional taking part, and the mailbox the agent will work with. |
| Delegated access | A read-and-draft grant to that mailbox, approved by your IT team. No send permission is requested. |
| Compliance | The compliance requirements this inbox is subject to (e.g., are zero-data retention contracts necessary, are they sufficient, is local LLM hosting required?). |
Build
| Component | What we develop |
|---|---|
| Model | How to select between models for the different tasks the agents need to accomplish (e.g., fast models for classifying incoming email; more advanced models for briefings and drafts). |
| Instructions | The instructions and exact prompt for the agent — the exact items the dashboard should show, sample e-mails for tone, and how to ensure e-mail content is treated as data to summarise, not as actions to take. |
| Tools | Your e-mail and calendar will be added as "Tools" to the agent. Tools will not be given permission to "send", "forward" or "delete" any e-mails or calendar items — hard-coded in the permissions. |
| Guardrails | Injection defence, token management, and permissions for all tools used. |
| Evaluation | We will review the quality of the agent outputs, and discuss how to best iterate to achieve continuous improvement. |
| Human checkpoints | This agent will not interact independently with third parties. It will only give notifications and drafts to you. You are always "in the loop". |
Roll-Out Strategy
Several key decisions follow once the agent exists. For example, when do you give it permission to go beyond drafting, and against what rulebook? Which capabilities do you add next? And how do you scale the use of assistant agents across your organization?
Go-live
- Access and permissions verified: the read and draft-create grant is live on the right mailbox, and someone has confirmed send, forward and delete are genuinely unavailable.
- Compliance position settled: the hosting option is chosen, retention is agreed, and the requirements this inbox is subject to are confirmed with your compliance team.
- Output reviewed against real days: the dashboard, timeslot recommendations and drafts have been checked against several real days of email.
- Instructions tuned to the person: dashboard items, scheduling preferences and tone reflect how this professional actually works.
- Running cost understood: a token budget is set, and the expected monthly cost is known and accepted.
- Logging and visibility agreed: what the activity log records, who can see it, and how long it is kept.
- Owner named: who maintains the instructions, and who the professional contacts when the agent gets something wrong.
- Off switch tested: how access is revoked, who can do it, and how quickly it takes effect.
Format & Duration
2 Days
Delivery Modes
- Virtual via Zoom
- In-person at your premises
Scheduling Options
- Intensive: all days back-to-back
- Modular: split across multiple dates
Pricing
Available on request
Pricing varies by group size, format, and level of customization. Contact us for a tailored quote.
Request a quoteAlso available as open enrollment
Some flagship programs run as open cohorts, enroll as an individual and learn alongside peers from other organizations.
See open cohortsFAQs
Ready to run this program for your organization?
Tell us about your team and we'll design the right delivery: virtual, in-person, or hybrid.
Get in touch