We're building an AI "employee." Here's what it messaged us.

Hi, and happy Tuesday.

🔔 We’re demo’ing this on Friday - 6th March, 10am PT / 12 CT / 1pm ET.Reply “INVITE" for the invite / recording.

Last week I got an impromptu chat message from our AI agent - without anyone asking:

"I saw you're working on EU PPWR compliance. 4 companies you're tracking have published compliance statements. Want me to pull up the details?"

Nobody prompted it. Nobody scheduled that message. It just… knew it mattered.

It had been building a knowledge base - automatically - for weeks. It named specific companies, connected them to regulations they'd need to comply with, and flagged a patent filing from 11 days ago that nobody on our team had noticed yet.

Think about how most teams use AI right now.

Someone opens Copilot, asks a question, gets an answer, closes the tab. Tomorrow, the conversation starts again from zero.

It doesn't know what project you're on. It doesn't know that the patent it helped you analyze is connected to the supplier your procurement team is evaluating. 

That's not a team member. That's a search engine with better grammar.

Let me make this concrete.

Imagine you lead an R&D team focused on sustainable materials. Today, staying current means your team manually scanning industry news, tracking regulatory changes across multiple jurisdictions, monitoring competitor patent filings, and trying to connect all of it to your active projects.

That work is important but tedious, and it never feels done.

Now imagine an AI team member configured for your domain. Every morning, it's already processed overnight developments. It messages your team channel:

"A research group at ETH Zurich just published a paper that's directly relevant to the membrane project. And two new patent filings related to mono-material flexible packaging from Amcor and Mondi."

You didn't assign that work. You didn't even know those patents existed yet. But now you do, and you can act on them.

When you ask a follow-up question - "which companies are most active in chemical recycling?" - it doesn't search Google. It answers from structured knowledge it's been building for weeks, with entity relationships, timeline context, and connections to your existing projects.

That's not a chatbot. That's a team member who never sleeps, never forgets, and never loses context.

The 4 things that make AI a team member instead of a tool

We've been building this internally at PreScouter, and a pattern is emerging that is surprisingly simple. There are four capabilities that separate "AI tool" from "AI team member":

1. It knows your world.

Not the whole internet. YOUR world. Your clients, your projects, your team, your competitors, the regulations that matter to you. It builds a living knowledge base from every meeting, every email, every report — and it connects them. When you mention "Acme," it doesn't ask "which Acme?" It knows.

2. It reaches out first.

This is the one that changes everything. Most AI is reactive - you ask, it answers. An AI team member has a heartbeat. It's scanning your domain every day. When something changes - a new regulation, a competitor move, a stalled project - it tells you. Before you knew to ask.

Think about the best analyst you've ever worked with. They didn't wait for you to assign them work. They came to your desk and said "hey, I noticed something you should see." That's what this is, every single day.

3. It does the work, then asks permission.

There's a massive difference between "here's information" and "here's what I think we should do - approve it and I'll execute." An AI team member prepares actions - draft emails, CRM updates, team assignments, research scopes - and presents them for your review. You edit what needs editing, confirm, and it executes.

Need to notify three team members about a project change? It drafts the messages, shows you what it'll send, and waits for your go-ahead. You stay in control. But you're reviewing work, not doing it from scratch. That's the difference between managing and doing.

4. It gets smarter about YOUR organization over time.

Every interaction, every decision, every piece of data makes the system more useful. After a month, it knows your patterns. After three months, it's connecting dots across projects that no individual could track. After six months, it has institutional memory that doesn't walk out the door when someone changes roles.

This is the part that compounds. And it's the part no one-off AI tool can replicate.

Skills are the new apps.

Here's the thing I want to be honest about: Microsoft, Google, and OpenAI are all building AI agent platforms. Within a year or two, the infrastructure layer - the plumbing that connects an AI to your chat tool, gives it memory, and has it be proactive - will be commoditized. 

Within a year or two, what I'm describing will be normalized.

But what those platforms won't build is the skill layer that sits on top:

• What YOU look for in patents.

• YOUR domain configured as a living knowledge base.

• YOUR team's patterns of work learned and refined over decades.

They also won't build the skills you want, but could never easily hire for:

• The mental models regulatory bodies use to evaluate your product.

• The due diligence framework a top materials scientist uses to separate genuine polymer breakthroughs from marketing claims.

• How to audit supplier sustainability claims against actual certifications.

• How a top patent attorney evaluates freedom-to-operate risk in a crowded IP landscape.

• How a world-class venture capitalist evaluates new packaging companies.

That's the layer we're building: the operational knowledge that turns a general-purpose agent platform into a useful team member.

The platforms will give you the engine. We're building the driver who knows where you need to go.

See it live this Friday

We're hosting a 30-minute live demo this Friday: AI as a Team Member (10am PT / 12pm CT / 1pm ET)

I'll be demoing a live instance configured for packaging and materials intelligence. You'll see:

• The agent living inside a chat channel (not a separate app)

• Proactive morning briefings generated from overnight intelligence

• Real-time Q&A where the agent answers from its knowledge base, not the internet

• How we're layering capabilities custom to specific domains

Whether you join live or watch the recording, you'll walk away with a concrete picture of what "AI as a team member" actually means in practice - and a framework for thinking about where this fits in your own organization.

If you'd like to join us, just reply “INVITE” to this email and I'll send over the invite - and the recording for those who can't make it.

Talk soon,

Dino