How to set up an AI research workflow without needing to be an AI expert

Where to start: delegate what's repetitive, keep what needs your judgement

Start by handing over the tasks that are repetitive and checkable (literature triage, first-pass protocol drafting, structure prediction) and keep the ones where your judgement is essential. The hard part isn't picking a model. It's knowing which steps you can hand over safely, where your data sits while you do it, and being able to show your work.

This is a live session that walks through that loop end-to-end, using peptide design as the worked example. You leave with a setup running against your own research questions, and a record of every step it took.

This first webinar is limited to 15 seats to give the opportunity for a specific, hands-on experience.

When | 9 September 2026, 6pm CET
Format | One hour, plus 30 minutes of Q&A
Seats | 15
Hosted by | Molecule AG and PeptAI
Cost | Free, including $20 of compute credit

Sign up here

You're already an expert in your field. You don't need to become an expert in a second one.

AI is everywhere, and it's easy to feel like you're watching the world speed by while you’re on the sidelines. A new model every week. A new tool every day. Newsletters telling you why the thing you haven't tried yet is already outdated.

If it feels like you missed the start, the good news is that you didn't. This isn't a race with a finish line, and the people a hundred paces ahead aren't running away from you. Plenty of those hundred paces were spent going the wrong way: wrong tools, abandoned subscriptions, setups that never quite worked. Starting now means you get to skip some of the detours.

How to decide what to delegate

This is where most people get stuck. Some research tasks are genuinely well suited to handing over, e.g. anything repetitive, anything where the output is easy to check, anything you'd otherwise put off. Others are not, e.g. decision making, validating source of truth documents, and proof-reading. We will cover the logic behind the architecture and how we approach these handoff moments between the technology and the human.

What goes wrong, and how to avoid it

Common mistakes, pitfalls, and the setups that look impressive in a demo and fall apart in a lab. This is the part that saves you the most time, because it's the part you'd otherwise learn by losing a month to it.

Keeping yourself in the driver's seat

An AI-augmented workflow should make you faster without making you dependent, and without leaving you unable to explain how you got a result. We'll cover how to structure the loop so that you stay the one making the calls.

That includes keeping a record of what an AI agent actually did — which step ran, on which inputs, and whether a person or a machine did it. Provenance for results is the difference between something you can publish and something you can only describe, and it's easier to set up at the start than to reconstruct later.

The worked example: peptide design

The PeptAI team will walk through the architecture they use to develop novel peptides. It's also modular; protein design is the starting point because it's what PeptAI does, but the structure is adaptable to e.g. small molecules or antibodies. Depending on your use case, you can see which parts are specific and which parts are yours to swap out.

What you'll walk away with

A starter pack

​A checklist of tools and subscriptions needed to get started, plus $20 compute credit pre-loaded into your account.

A working tech stack

​A working setup you can point at your own research questions the same day. Starting with protein design, but with modularity for other use cases.

A shortcut forward

​Inherit the lessons instead of stumbling through common pitfalls on your own, saving time and money.

An open channel

A private group with your cohort and the instructors. Somewhere to ask questions, share what worked, and get updates after the session ends.

Who this is for

People with a research question who are curious about AI, haven't yet built an AI workflow, and would rather start with something running than with more reading. You do not need to write code, and you do not need any background in machine learning.

Alternatively, if you are familiar with automated workflows, but have an interest in scientific research and peptide design, you are welcome to join to learn about the other side of this equation.

What about Molecule Labs

Molecule labs is a secure data room for research, with controlled access and data provenance built in. Decide what to make publicly available, and what to keep private for you and your team.

In practice that means three things. Your files stay encrypted and you decide who can open them.

Every step in the workflow writes an entry to an audit trail, so the record of what you did, when, and by whom exists without you maintaining it.

And if you want to share your work publicly, it’s as simple as setting permissions to public, so you don’t have to figure out how to make your own website.

FAQ

Do I need to know how to code?

No. The setup is assembled from existing tools, and the session is built for people who haven't done this before.

Which LLMs are best to use for protocol generation and data analysis?

We'll cover the current options and, more usefully, how to tell whether one is working for your particular case rather than in general.

How do you validate AI-generated targets experimentally?

We'll cover where validation belongs in the loop, what a reasonable checkpoint looks like, and how to decide how much bench time a computational hit has earned.

How do you keep a record of what an AI agent did in your research?

Every step writes to a provenance record automatically i.e. what ran, when, on which inputs, and whether a person or an agent did it. The audit trail exists without anyone maintaining it manually. A combination of tools, including Molecule Labs, is best suited for this job, and we will show you how they work together.

Sign up here