AI Field Notes – Issue No. 20

Deep Research, Custom CRMs, and the Coming Agent Layer

Deep Research Is Wildly Underused

One of the most underutilized capabilities in AI right now is Deep Research.

These reports aren’t summaries — they’re near PhD-level syntheses that get you up to speed fast and propose real solutions.

You can dig into geo-specific markets, ideal customer profiles, TAM/SAM/SOM, and even competitive positioning in minutes — insights that used to take teams days to compile.

From a marketing and sales standpoint, this changes how fast you can validate strategy and spot opportunity. If you’re not using Deep Research, start. You won’t regret it.

Salesforce’s “AI Is Fine” Problem

Salesforce just reported earnings and assured investors that things are fine. They’re not.

AI — the same thing they’re trying to build into their stack — is an existential threat to their business model.

Salesforce is a classic example of an incumbent benefiting from saying the word “AI” while quietly being disrupted by it. We’ve used Salesforce multiple times — it’s powerful but cumbersome, expensive, and bloated.

Here’s the truth: You will soon build a CRM that’s custom to your company.

You won’t have to shoehorn your processes into someone else’s platform. AI will handle 80% of the build — internal or external teams will fine-tune the rest.

If you’re evaluating your CAC, include your CRM in that math.

Long-term contracts with legacy CRMs often eat up marketing ROI.

Yes, you need a CRM — but custom, lightweight, AI-native CRMs are coming fast.

My advice: don’t lock yourself into a long Salesforce contract.

Agents — What They Actually Are

There’s a lot of noise about “agents.” Let’s define it clearly:

An AI agent is a system that can take action on your behalf — not just give you answers.

Agents can execute multi-step tasks like:

• Research a market → draft a report → send it to your drive.
• Monitor campaigns → pause underperformers → alert the team.
• Pull CRM data → update records → generate next-step recommendations.

We’ve built and tested these internally — they work, but they’re early.

A lot of companies selling “agents” are overselling. True, autonomous agents are coming, but we’re not there yet.

Tread carefully, but pay attention — this is the layer that replaces repetitive work.

MCPs — The Quiet Revolution in Workflow Integration

If Agents are the muscle, MCPs (Model Context Protocols) are the nervous system.

An MCP lets your AI securely connect to your company’s data and tools — Google Drive, Slack, HubSpot, SharePoint, CRMs — and operate within that environment.

Think of it as plugging AI into your actual business.

Example:

A company could connect its sales data, calendar, and project management apps to an MCP-enabled agent. That agent could then brief you on client meetings, summarize open deals, and even draft follow-up emails automatically.

When you connect your data to AI through MCPs, team productivity explodes. It’s not just faster work — it’s smarter coordination across every department.

The Bottom Line

• AI is moving from conversation to coordination.
• Deep Research gives you speed and clarity.
• Custom CRMs give you control.
• Agents give you execution.
• MCPs give you connection.

Together, they redefine how marketing, sales, and operations actually function day to day. Start experimenting with all four — they’re not future tools. They’re right now.