How I Underwrite Deals by Text Message Using Manus AI

This guide walks through everything about how I set it up, the workflows I’ve tested, what actually works, and where it falls short. If you’re in acquisitions, brokerage, asset management, or development, there’s something here you can use this week.

The Problem With Manus AI CRE Underwriting Before Skills

I’ve been asking AI tools to underwrite deals for months. And the outputs were always different every single time. Ask Manus to “underwrite this deal” on Monday, and you get an Excel file with five tabs. Ask the same thing on Wednesday with a different OM, and you get three tabs. Different layout. assumptions. and formatting. One has sensitivity tables. The other doesn’t. One calculates IRR. The other skips it entirely. Run it 10 times, and you get 10 different structures. You can’t compare deals side by side. can’t hand these to an LP and say, “Here’s our standard analysis.” You definitely can’t build a repeatable process around outputs that look different every time.

This is the core issue with using generic AI for underwriting. The intelligence is there; the AI can extract data from an OM, make reasonable assumptions, and build a financial model. But without explicit instructions on your format, your tabs, your formulas, it just guesses at structure. And it guesses each time differently. That’s what Manus skills fix. And when you combine skills with agents, you can trigger that standardized underwriting from a text message on your phone.

What Manus AI CRE Underwriting Agents and Skills Actually Are

Two features were released in February 2026. They work independently, but the real power lies in combining them.

Agents

Agents connect Manus to Telegram. You text Manus like you’d text a colleague:

• “Hey, underwrite this deal.”

• “Pull rent comps for this property.”

• “Generate a construction cost report for this market.”

You can attach files, send voice messages, and queue up multiple tasks at once. Manus works in the background and texts you when each task is done. Current limitation: Telegram only. WhatsApp and iMessage are coming, but there is no exact date yet. This is what allows Manus AI CRE workflows to run tasks like underwriting, rent comps, and market reports from a simple text message.

Skills

Skills are explicit instructions for specific tasks. You upload your template, your model, your process. Manus learns it and replicates it every time.

Think of it like writing an SOP for a new analyst, except the analyst never forgets a step, never changes the format, and works at 10x speed. Skills also get smarter over time. Manus saves “knowledge-based scenarios” from your interactions, so the more deals you run through a skill, the better it understands your preferences.

Combined: Build an underwriting skill once. Trigger it by texting your agent an OM. Get back a standardized pro forma in your exact format. From your phone, you’re doing other things.

How I Built My Underwriting Skill (Step by Step)

This took about 25 minutes. You only do it once.

What You Need Before Starting

• Your actual Excel underwriting model (all tabs, formulas, and sensitivity tables).

• A Manus account with skills and agents enabled.

Step 1: Start the Skill Builder

In Manus settings, go to Skills → Build with Manus.

I typed

“Help me create a skill. Anytime I ask you to underwrite a deal, use my exact model attached here. It should use this exact template with updated numbers from each deal. I will usually send you an OM and ask you to underwrite. You will go ahead and underwrite the deal and return this exact model with the updated numbers for that specific deal. Please ask me questions if you aren’t clear.”

Then I uploaded my Excel model.

Step 2: Answer Manus’s Questions

Manus returned five specific questions:

1. Sensitivity tables: Recalculate dynamically? → Yes

2. Target price solver: Implement a solver to find the maximum purchase price? → Yes

3. Analyst notes: Auto-generate notes? → Yes

4. Unit mix: Support up to six unit types? → Yes

5. Renovation premium formula: Always use the formula approach? → Yes

Step 3: Let Manus Build

Manus initialized the skill and completed tasks like:

• Mapping every tab in the template Excel file

• Cleaning up example files

• Copying the model template to the skills directory

• Creating reference documents for model specifications and OM extraction

• Writing the full skill instructions

After 10-15 minutes, the skill appeared as “Multifamily Underwriter” in my library.

Step 4: Test With a Real OM

I attached an OM for 163 8th Place, texted:

“Please underwrite this deal using our new multifamily underwriter skill.”

The Manus AI CRE Underwriting Text Message Workflow

From my phone:

1. Attach OM in Telegram

2. Text Manus: “Please underwrite this deal for me using our new multifamily underwriter skill.”

Manus:

“I’ll underwrite 163 8th Place using the multifamily underwriting skill. This should take about 5-10 minutes.”

Behind the scenes, Manus:

• Extracted key data from the OM

• Researched market comps

• Built the full pro forma in my template

• Ran sensitivity analysis

• Solved for the target purchase price

• Generated analyst notes

• Packaged everything and sent it back

Even when a subtask hit a snag, the agent self-corrected and completed the underwriting.

Result:

• 5.4% IRR at full purchase price

• Recommendation: Pass

• Excel model matched my exact format with all tabs, formulas, and deal-specific numbers

All from a text message while I was doing other things.

Other Manus AI CRE Underwriting Workflows I’ve Tested From My Phone

Simultaneous Market Reports

Texted:
“Prepare a report on construction costs in Los Angeles.”
“Also prepare a report for construction costs for apartments in St. Louis.”Both reports returned within minutes. Parallel execution saves hours.

Voice Message Rent Comps

Sent a voice message:
“Pull rent comps within 2 miles of 3105 Bellevue Avenue, Los Angeles.”Full Class A multifamily rent comp analysis returned as a downloadable report.

Rendering and Video From a Description

Voice message:
“5-story podium building in Silver Lake. 4 floors of residential over 1 floor podium. About 30 units.”Two minutes later: architectural rendering.

Texted:
“Can you turn that into a video?”A few minutes later: video walkthrough delivered.

What Manus AI CRE Underwriting Does Well

• Phone-native execution: Work from properties, meetings, or driving

• Parallel processing: Multiple tasks executed simultaneously

• Standardized outputs through skills: Every deal returns in the same format

• Proactive agent behavior: Self-corrects and notifies you when tasks finish

• Voice input: Fast verbal commands instead of typing detailed prompts

Where Manus Falls Short (Honest Take)

• Telegram only – WhatsApp/iMessage not yet available

• Accuracy requires verification – Always review assumptions and outputs

• Skill building takes investment – Initial setup takes time and iteration

• Task management – Tracking multiple tasks requires attention

• Cost scales with usage – Complex workflows consume credits

How Manus Compares to Claude and ChatGPT for CRE

Workflow: Write prompt in ChatGPT → Feed to Manus or Claude → Use Manus for phone access/parallel execution, Claude for deep analysis.

How to Get Started by Role

• Acquisitions/Underwriting: Start with an underwriting skill

• Brokers: Start with comp analysis and market research

• Asset Management: Build skills for reports and budgets

• Developers: Use rendering workflow via voice messages

• All roles: Start simple, then scale to complex workflows

Common Mistakes When Using Manus AI CRE Underwriting

• Forward an OM to your Manus agent

• Text “underwrite this.”

• Geta  standardized pro forma in 10 minutes

Build one skill this week: underwriting model, comp report, or property report template. Spend 25 minutes upfront and never manually set up that output format again.

Take Your CRE Workflow to the Next Level