If I were advising a company that is already overwhelmed by AI, I would think about it in three layers:
automation, intelligence, and human judgement.
When those get treated as the same job, everything looks like an AI project. That's why it feels like too much.
Automation is the work that should just happen. Move a record. Create a task. Send the update. You already know what should happen, so it happens the same way every time. That's rules and workflow.
Intelligence is the work that has to be understood. An email in messy language. A document that has to be read. A request that doesn't fit a form. Work out what's going on, then the rest of the business can act. That's where AI belongs.
Human judgement is the work that still needs someone to look over it. A payment. An unhappy client. Whether to make an exception. The system can gather the context and lay out the options. A person still makes the call.
Take one messy request and put it in a layer. If it has to be understood first, that's intelligence. If the next step is a known update, that can be automation. If a person still has to make the call, that's human judgement.
I'd build the tender desk I keep doing by hand.
A pack lands as a PDF, an email, and a spreadsheet that don't agree. Devin would pull the sites, the dates, and the gaps into one record, flag anything that would be unsafe to price, and stop there. A person only sees the call that still needs a human.
I've already got the rules and the failure cases. The missing piece is an agent that can run that path on a messy pack without me babysitting every file.
@the_varunrana I used to hate keeping notion organised but just started it again now that agents can do that. What kind of things are you using it for?
Iโm building the platform behind Layerโs production AI workflows for small businesses. It handles client tool connections, permissions, workflow execution, monitoring and recovery in one place.
Iโd use Devin to help me ship new integrations and workflows faster, while strengthening the testing and reliability layer as I scale it across more businesses
Iโm using Grok Bot as part of a system that keeps me on top of whatโs changing in AI without living on X all day.
It scans for relevant updates, pulls out the ones worth paying attention to, and feeds the best signals into a research + content workflow.
Basically: less scrolling, more useful signal.
Iโd have Devin build the client deployment engine for Layer.
Give it a company config and it spins up the workspace, auth, tool connections, permissions, monitoring and human approval gates, then deploys the first AI workflow.
Goal: take a small business from โwe should use AIโ to a live, controlled system in a day.
@GeorgeSiosi@NotionHQ After fiddling around with more โcomplexโ solutions Iโve found myself back to using notion for this too.
Will be following with your challenge ๐ค๐ฝ
@clairevo The more I use grok @bot the more Iโm starting to rethink a lot of my skills/workflows as โjust make a bot for itโ
Is this what you mean?