Using AI To Grow Our Business

We have implemented AI across multiple businesses within our portfolio

We're building AI into everyday work across our portfolio. But we're not trying to replace people. Instead, we're giving our teams the tools they need for sharper planning, faster quality checks, and much cleaner analysis.

We're sharing the high-level strategy and the exact prompts we actually use day-to-day. These are the workflows that are working for us in practice.

Where We're Actually Using AI Today

Planning: Turning them into clear, actionable plans with proper timelines and risk checklists that people can actually follow.

Quality Control: Think of this as your second pair of eyes. We're running sanity checks on everything from marketing copy to financial models and client proposals before they go out the door.

Portfolio Reviews: Before big meetings, we're getting roll-up summaries that highlight the wins and misses, spot the red flags we might have overlooked, and create those "explain it to me like I'm new" briefs that keep everyone on the same page.

Real Estate Underwriting: We're generating comp summaries, building sensitivity tables, creating DSCR snapshots, and pulling together risk notes on our key assumptions. Basically, all the analysis work that used to take hours.

Content & Communications: First drafts for scripts, emails, and outlines get handled by AI, then our team jumps in to add the human touch and make it actually sound like us.

Prompting 101: The Rules That Actually Work

Here's what we've learned after months of trial and error. The best practices are surprisingly consistent across all the major providers (OpenAI, Google, Anthropic, Microsoft):

• Be really explicit about what you want, who it's for, and how you want it formatted.

• Show the AI what "good" looks like with a short example. It's like training a new employee. One solid example is worth a paragraph of explanation.

• Don't expect perfection on the first try. Ask the model to improve its work, then add more specific constraints until you get exactly what you need.

Simple and direct instructions tend to work better than over-engineered prompts. Spend more time curating the best context (examples) to get better responses.

Sample Templates We Actually Use

Weekly Portfolio Roll-Up

You are a chief of staff. Summarize these updates: <<<NOTES>>> 
Deliver: 8-bullet exec brief (wins, misses, blockers, decisions needed). 
Flag any numbers that look inconsistent and ask 3 follow-up questions.

Marketing QC (Pre-Publish)

You are a fact-checking editor. Review this asset: <<<COPY>>> 
Task: list factual claims, potential legal or compliance risks, unclear claims, and weak CTA.
Deliver: fixes + a tightened 120-character headline and 40–70 character preheader.

Real Estate Comp Digest

Role: acquisitions analyst. Dataset: <<<COMPS/NOTES>>> 
Task: produce a table: address | type | date | $/sf | cap rate (if provided) | notes.
Then 6 bullets on comp relevance and outliers. If data is missing, state it.

DSCR Snapshot + Sensitivity

Role: underwriter. Inputs: NOI [ ], annual debt service [ ], rate [ ], term [ ].
Compute DSCR, then sensitivity: DSCR at ±50 bps and ±5% NOI. 
List assumptions, risks, and "what would change my mind."

Guidelines for Your Team: What's Safe and What's Not

We've found it helpful to give our people clear guardrails so they know when to lean in and when to pump the brakes:

Green-Light Tasks: These are your bread and butter AI wins. Planning outlines, meeting briefs, proofreading, brainstorming idea lists, and those first-draft summaries that get your thoughts organized.

Proceed with Caution: Anything involving numbers or factual claims needs a human double-check. The AI is great at structure and format, but you still need to verify sources and make sure the math actually adds up.

Hard No's: Never put confidential data, personal information, or anything you wouldn't feel comfortable emailing to a competitor into these tools. If you wouldn't share it externally, don't share it with AI.

How to Keep AI From Making Stuff Up

This is probably the biggest concern we hear from people, and rightfully so. Here's our checklist for keeping the AI grounded in reality:

Set clear boundaries: Tell the AI to only work with what you've given it. Something like "Only analyze the text I provided. If you're missing information, just say so" works wonders.

Make it show its work: Ask for sources on any claims. "Cite where each fact came from. If you're not sure about something, say 'I don't know'" keeps the AI honest.

Structure your inputs: Use headings, bullet points, and clear separators. It's like organizing a filing cabinet. Better organization leads to better accuracy.

Always fact-check the important stuff before you publish or make decisions based on AI output.

Keep records of what you've checked, especially for finance, legal, or compliance work. Your future self will thank you.

The Big Picture Insight

Here's what we've learned after implementing AI across dozens of workflows: it works best when you're asking it to accelerate things you already know how to do well.

Think of it as a really smart intern who's great at taking your expertise and existing materials and turning them into cleaner plans, better-organized drafts, and more thorough reviews. Start with transformation tasks like "summarize this," "reorganize that," or "identify the risks in this proposal" before you ask it to create something completely new from scratch.

Your Action Plan for This Week

Ready to give this a try? Here's how to get started without overwhelming yourself or your team:

1. Pick just one recurring task from the examples above and test out the prompt exactly as we've written it

2. Track one specific win: Maybe it's 30 minutes saved, catching an error you would have missed, or just having a clearer starting point for your work

3. Share what worked with your team. We've found that these small wins add up fast when everyone's learning from each other

The goal isn't to revolutionize everything overnight. It's to find those moments where AI can give you back a little time and mental energy to focus on the work that actually needs your human expertise.

Get the Complete Toolkit We Use

Want access to our full internal prompt library and the one-page guardrails document we give to every new team member?