How I use AI every day

I use AI every day, and I feel like I've barely scratched the surface of what it can do. I'm probably further out on the edge than most people, and I don't think most people are using it the right way. This month I want to share what's working for me, and I want to hear what's working for you, because we're all figuring this tool out at the same time.

I think AI is going the way electricity and the internet went, from a novelty to a utility in every home. It isn't perfect. It makes mistakes, sometimes with total confidence, and it needs someone who knows the work checking it. If you understand how it works and where it breaks, and you stay in the loop, it can turn a week of work into two or three hours, and a lot of the time the result is better than what I would have done the first time.

I don't like AI slop either. If someone pastes my question into ChatGPT and sends me back the answer, I could have done that myself. Lazy input gets generic output, and it shows. When I give it the detail, access and context it needs, the result is worlds better, and it still takes my fine-tuning to get it right. It isn't a shortcut around the work; it does the small steps faster, so I can do ten times as much of it.

Building Websites, Including the Engineering

The first thing I'd show anyone is the websites. I rebuilt Starlight Horizon, the boutique hotel I run on Canyon Lake, 20 Tate, our rental homes in Round Rock, and our investor site with AI. The design is the part you see. The part that used to need an engineering degree is underneath: analytics that show which ads actually bring bookings, tracking that tells Meta who booked so it can find more people like them, forms that drop a lead straight into our email list, and speed testing on every page. Work I would have paid a developer for over several weeks now gets done in a day or two.

I'm still the one deciding. I describe what the page has to do, the AI builds a version on a test site, and I go through it on my phone and my desktop and mark up what's off with screenshots. Some pages take five rounds. When a fix works it goes into the build rules, so the next site starts where the last one finished.

A Filing System That Stores and Finds Everything

Every deal, property, client, contact, tax year and decision I've made lives in one searchable place on my computer: almost 1,500 Markdown files, each carrying several pages of detail. Each property I own and each client I work with has its own knowledge base, checklist and list of open opportunities, stored as we go and pulled back up at a moment's notice. Every work session starts by pulling the context it needs from these files automatically, so I never re-explain a deal, or re-explain to an AI how my business works. The AI already knows the full history of a project, every interaction I've had on it and how I tend to think and make decisions, so its advice and its work are specific to that project and to me.

Written procedures for the team became an easy byproduct. Our Starlight Horizon onboarding guide is 53 pages of procedures for the team (this copy has the private details blacked out), and it came out of what the filing system already knew.

Property Reviews From 3 Hours to 10 Minutes

Before a call with our property manager, I used to review our 12-month income statement, rent roll and leasing numbers at a high level, and that was all I had time for. Now the AI also reads the work orders, the full general ledger, utility bills, the website's booking analytics, guest messages and the other 90 or so reports in the system, and pulls out the needles in the haystack that would have turned into bigger problems if we hadn't caught them early. At 20 Tate it went through all 66 utility bills from the last 12 months against the books, explained exactly why the bills had gone up, and flagged a few bookkeeping and account setup issues for our manager to clean up while they were still small. At Starlight it checked every septic invoice against our signed contracts and caught charges above what we'd agreed to, which the vendor refunded.

Documents That Look Designed

I can now put together a 50-page proposal with real graphic design: timelines, org charts, full photo pages, charts built from the actual numbers. Laying that out by hand used to be days of tedious work in design software, and most of it I wouldn't have attempted. Now the layout takes hours and the design is better than I could do on my own.

Images are the same story. Most people can spot a ChatGPT image by now, it has a glossy, generic look, and I think it looks terrible too. That look comes from typing one sentence and taking whatever comes back. I describe the scene, the light and the exact size the page needs, look at a few versions, and refine the one that's closest. Every image in this letter was made that way.

Catching What a Seller's Numbers Hid

In September I was underwriting an event venue. The seller's summary showed earnings of about $353,000, and $100,665 of that was owner draws added back as if they were a business expense. The owner had never been paid a wage through the business, so there was nothing to add back, and the same dollars were being counted twice. At the multiple the venue was priced at, that one line was holding up roughly $260,000 to $300,000 of the asking price. The AI rebuilt the earnings from the raw monthly statements, and I checked the math by hand before I believed it. AI didn't change this review; I would have done it myself. At the level of detail it needed, it would have taken me five or six hours, and the AI did it in 15 minutes. That's what I mean when I say it blows me away, it has dramatically increased what I'm capable of.

Marketing Insights I Couldn't Find on My Own

Every week I run one review of how the marketing did at Starlight Horizon, the boutique hotel I run on Canyon Lake: the website, ads, Google search, email, pricing and every booking in the reservation system. What blows me away is the connections it makes across all of it. In October it found that 86% of people who searched for dates and saw our cabins left without picking one, which pointed us straight at the page to fix. The same review caught that our Thanksgiving Friday and Saturday were sitting empty at about 25% above the market's typical price, and recommended bringing them down to around market. I wouldn't have found either one in the time I have, and I can check every number it gives me against the source.

Writing the Playbook Once

The biggest time saver is a quiet one. When I do something more than twice (writing this letter, underwriting a deal, building a website), the steps get written down as a playbook the AI follows the same way every time. This letter has one. Quality stops depending on whether I'm having a good day, and every fix goes into the playbook, so the tenth run is better than the first.

97 More Ways I'm Using AI

Most days I run five or six AI sessions at once. I hand one a chunk of work, move to the next and hand it another, and by the time the sixth one is set up, the first is ready for my review. It works as if I am delegating work to a small team.

I don't have space to share details on all these use cases, but here are a bunch more:

• Underwriting apartment and hotel deals
• Rebuilding a seller's earnings from raw statements
• Running investor return splits inside the actual model
• Mapping drive times from a property to job centers
• Pulling county appraisal and tax records
• Reading federal securities filings to size up other sponsors
• Confirming a Texas business is real and who owns it
• Checking a professional's license before I hire them
• Catching contract clauses that break because of how our entities are set up
• Translating a contract into plain English before I sign it
• Reading our full insurance policies to see what's covered when we have a claim
• Building insurance claims around what the insurer needs to see
• Tracking every deal on one pipeline board
• Drafting pass emails to brokers
• Building lender target lists
• Building lead lists from Google Maps data
• Spotting Texas businesses that just opened
• Writing these letters with every number sourced
• Quarterly letters for each property's investors
• Pulling market data from primary sources
• Prepping for property manager calls
• Turning call recordings into transcripts and action items
• Calculating the hotel's quarterly occupancy tax
• Assembling credit card dispute packets
• Reviewing new-hire paperwork for gaps
• Splitting and filing signed documents
• Checking vendor invoices against signed contracts
• Reconciling utility bills against the books
• Tracking county permit and inspection deadlines
• Writing staff onboarding guides
• Following up on group booking leads
• Writing guest policies for the booking engine
• Auditing room holds in the reservation system
• Comparing our nightly prices to the market, date by date
• Finding where guests drop out of the booking process
• Tracking bookings back to the ad that brought them
• Catching a tracking pixel that counted every sale twice
• Reviewing ad creative before it runs
• Auditing the Meta ad account
• Studying what people search before they find us
• Building email campaigns in each brand's voice
• Cutting social reels from our own footage
• Building photo carousels
• Color grading video
• Scheduling social posts
• Researching what's trending for boutique hotels
• Making images for letters and ads
• Turning still photos into short video clips
• Building 3D models in Blender
• Upscaling old photos
• Keeping a brand kit for each business
• Designing websites
• Writing website copy that converts
• Auditing a business's existing website
• Testing page speed
• Finding images that break in some browsers
• Writing the structured data search engines read
• Merging duplicate pages without losing search traffic
• Setting up website tracking tags
• Sending contact forms straight into our email list
• Drafting privacy pages for an owner to approve
• Editing WordPress sites safely
• Putting sites on a test server for review
• Before-and-after screenshots for clients
• Plain-language explainers for business owners
• Proposals with custom design
• Full-screen flowcharts of how a business runs
• Building pricing pages
• Comparing competitors' websites
• A daily log of what got done
• Planning my day and the next one
• Stress-testing a decision before I make it
• Scoring new business ideas against my goals
• Thinking through a job offer
• Planning a negotiation before a hard conversation
• Logging real estate professional hours for taxes
• Organizing years of tax returns
• Keeping health records organized
• Summarizing YouTube videos
• Turning books into a reference library
• Learning new software
• Researching people before a meeting
• Turning business cards into contacts and follow-ups
• Writing reconnect emails
• Drafting replies and texts in my own voice
• Checking my own writing for AI slop
• Fact-checking claims against sources
• Formatting spreadsheets to one house style
• Building dashboards
• Tidying my notes every night
• Connecting apps so they hand work to each other
• Gathering public web data at scale
• Researching vendors
• Turning a voice memo into a first draft
• Having one AI check another's work
• Writing down how I want things done so it improves every week
• Brainstorming this letter

A Small Side Venture

Building these sites for my own businesses turned into a side venture. If you or someone you know needs help with a website, business automation or marketing, take a look at Tanel Website Studio. I'd be happy to help.

What We Are Looking To Buy Next

We're always looking at properties: 20 to 80 units, $2M to $6M, value-add or stabilized, within an hour of Austin, multifamily, boutique lodging, or a cashflowing business. If something crosses your desk, pass it along.

Ryan Tanel