Insights
The RSTS AI Blog
How service businesses are putting AI to work.
How do you know it did the work?
Handing a job to an AI employee raises a fair question: how do you know it actually did it, and did it right? The answer is a record you can check, and an employee that tells you when it's stuck instead of pretending.
Read more →
Don't summarize your business
The instinct is to boil your business down to a tidy one-page summary for the AI. We tried it. It made the AI employee worse, because the value is in the detail a summary throws away.
Read more →
How much of your business does it actually know?
An AI employee is only as good as how much of your business it has learned. We started measuring that as a number, and it changed how we think about what an AI employee is ready to do.
Read more →
Your next customer might send an assistant
Soon some of the calls and messages reaching you won't come from people, they'll come from their AI assistants, sent to find, ask, and book. A missing answer doesn't read as call us, it reads as skip.
Read more →
The work between the deals
Most agents point AI at listing descriptions, the part clients see. A growing team leaks between the deals, in the quiet leads and the deadlines nobody wrote down. That's where an AI employee earns its keep.
Read more →
You shouldn't have to run a swarm
The internet is full of people running fifteen AI agents at once and calling it the future. For a business owner, that isn't a dream, it's a second job. You should hire one employee, not manage a swarm.
Read more →
Who owns your AI before someone leaves
One curious person builds it, it works, so it grows, in their head. Then they take two weeks off and something breaks only they can fix. Put the accounts, the code, the runbook, and the keys in the company's name, then prove it with a drill.
Read more →
AI is the spotter, you still lift the weight
None of the five places says don't use AI. Every one says the AI employee does the work right up to the line, then a person makes the call, on judgment, signed work, confidential data, one-way doors, and the messages where showing up is the point.
Read more →
You are the glue
Between the inbox, the calendar, the CRM, and the quoting, someone spends the day copying things from one place to another. In most small businesses that someone is the owner. That's the job an AI employee is built to take.
Read more →
Every question asked twice is a missing SOP
You never write SOPs because it means a blank page and recalling steps you do on autopilot. So talk it through, let the AI employee interview out the judgment, and test it on your newest hire.
Read more →
A question saves minutes, a job saves hours
A chat answers, then sits there while you copy, paste, fix, and send. A job has a start, a finish, and a place the result goes. Listen for the word every, and take yourself out of the trigger.
Read more →
Answering is just the front door
Picking up the phone is where most people stop thinking about an AI employee. It's the front door, not the house. The same one that answers can qualify, quote, book, and follow up, and hand off to a person when it should.
Read more →
Why examples beat instructions
A model is a lossy zip of the internet, so it hands you the average, not you. Your business was never in that zip, it has to sit on the desk. Facts help, but three real examples change the draft most.
Read more →
Answered in your customer's language
One in seven people in the US speaks Spanish at home, and in Miami it's most of the county. An English-only phone line quietly turns them away. An AI employee answers in the language the caller uses, without a second hire.
Read more →
We stopped describing it and recorded it
Everyone says their AI is great, so nobody believes any of it. We stopped describing our AI employee and recorded four real calls instead. Here's what putting it on camera taught us.
Read more →
Five jobs your inbox already needs done
Using AI to write emails is the smallest slice of the job. The money leaks from the quote nobody followed up on and the question that sat for three days. Point an AI employee at reading your inbox, and keep the send button human.
Read more →
She took the order in Spanish
A missed call at the counter isn't one lost sale. It's a regular who dialed the shop down the street. Here's what changes when an AI employee named Rosa answers every call, in the caller's language, and gets the order right.
Read more →
Train it on your business, not the internet
Out of the box, AI learned from the internet and knows nothing about you. You don't train it, you hand it five plain pages. The one most people skip is the one customers actually notice.
Read more →
Start with one job, not everything
Everyone wants to automate everything, but that's not the first smart move. Pick one costly, repetitive job, prove it, then take the next one.
Read more →
The judgment between your rules
Your business already runs on rules: invoice thirty days late, send the reminder. Rules catch the obvious. The money hides in the judgment between them, the calls a rule can't make.
Read more →
Giving your AI employee a voice
I put a new voice model on my real business line and called in like a customer. It cloned a voice from ten seconds of audio, booked an appointment, and sent a summary before I hung up. Here's what actually held up.
Read more →
What deserves your attention today?
You start the morning with 200 unread things and four hours. The hard part isn't answering any one of them. It's knowing which five actually matter today, before the day gets away from you.
Read more →
The work you stopped doing because it cost too much
Every business has a list of things it would do if they were free: read every review, listen to every call, check on every quiet client. When judgment gets cheap and instant, that list stops being a wish and becomes the plan.
Read more →
The front desk that never misses a call
A missed call isn't one lost call. It's a customer who already dialed the next name on the list. An AI employee on the phones answers in your voice, books into your real calendar, and hands off to a person the moment a call needs one.
Read more →
The most valuable AI on your team can't write a word
Everyone's impressed by the AI that writes the clever email. The one that actually gives you your day back is quieter: it reads the pile coming at you and decides where each piece goes.
Read more →
Don't let your associates skip the rep
There's a version of AI in a law firm that quietly makes your associates worse: every brief handed over whole, every argument generated and filed. Two years in, you have a team that can run the machine and can't build the case.
Read more →
Show your work
The malpractice claim that ends a small firm usually isn't a dramatic loss in court. It's a deadline nobody calendared, a citation nobody checked. The fix isn't working harder. It's leaving a trail you can audit.
Read more →
One question at a time
Your intake form asks twenty-two things at once because it can't think. The best intake is a conversation that asks one good question, shaped by the last answer, and already knows which question comes next.
Read more →
The question before you file
The market sold you AI that writes the motion in seconds. The least useful thing it can do, and the most dangerous. The AI employee worth hiring drafts, then reads the draft the way opposing counsel will.
Read more →
No surprises where it counts
You want two opposite things from the same hire: zero surprises on the work that touches a customer, and the freedom for it to occasionally delight you in how it gets there. Those don't fight; they live in different layers.
Read more →
When your best person leaves, the knowledge shouldn't
When your best person walks out, the résumé isn't the loss. It's everything they never wrote down. Memory isn't the enemy; ownerless memory is. The fix is a business brain you own, that remembers on purpose.
Read more →
The pieces lose the plot
We spent a decade teaching everyone to break work into the smallest shippable piece. Handed to AI, that instinct quietly backfires: torn apart, the pieces can't see the whole job, and the whole job is the point.
Read more →
Ten chat sessions, or one brain
Your business logic is scattered across a dozen chat windows that don't know each other exist. The fix isn't a better prompt. It's one living record that holds the goal, the standard, and how you check the work, all in the same place.
Read more →
Every AI failure is the same failure
The AI didn't misfire because the model is weak. It misfired because it didn't know exactly what you wanted. Prompt tricks, add-ons, and clever workarounds all orbit that one gap — and closing it on purpose is the actual job.
Read more →
Put the AI on the work you already double-check
The domains where AI is winning aren't the flashy ones — they're the checkable ones, where proving the output is right is cheap. The small-business move that follows: hand the AI the work you were already going to review, so checking it isn't a new job. It's the one you already had.
Read more →
Nine seconds to delete, thirty hours to recover
An AI agent with too much access wiped a company's production database in nine seconds — backups included. The cleanup took thirty hours of human hands. The lesson for any small business isn't 'avoid AI.' It's that guardrails and a person on the last step are the product, not an extra.
Read more →
Someone has to steer it. That someone shouldn't be you
The promise was that AI would take work off your plate. The reality at every scale is that it hands you a new job: choosing the work, feeding it context, checking the output, and stopping it when it goes wrong. For an owner, that's a second job. Managed means we do it.
Read more →
The $40 AI that never picks up the phone
Two-thirds of small businesses already pay for AI. Most of it is a window you have to go open and talk to — not something that answers the after-hours call, books the slot, or follows up. That gap isn't the model. It's the work nobody wired in.
Read more →
The fastest way to sound like everyone else
Pour a generic model into your customer conversations and you slowly start sounding like the average of your whole industry. Keeping your own voice takes a deliberate choice about what the AI reads before it speaks.
Read more →
Your competitor uses the same AI you do
The model isn't your advantage. Your competitor can open the same window and get the same answers. The difference that's actually yours is the knowledge only your business has, and whether you've captured it.
Read more →
Average just got cheap
As AI gets better at doing average work for everyone, the value of average collapses. What's left, and worth protecting, is the part of your business that was never average: your exceptions, your judgment, your way of doing things.
Read more →
If it doesn't save you a day a week, we won't build it
There's a simple filter for deciding which job to hand an AI employee first: does it give you back a day a week? If the honest answer is no, it's the wrong place to start, no matter how good the demo looks.
Read more →
You're not renting a tool. You own the employee
Most software rents you a seat and keeps the keys. An AI employee is different: you own everything we build. Here's why owning the outcome beats renting another subscription you'll half-use and never control.
Read more →
The leak that waited until Monday
A tenant reports a leak on Friday night and hears nothing until Monday. By then the damage is worse and the renewal is gone. Here's what slow maintenance response costs a property manager, and what changes when an AI employee named Vivian answers every request.
Read more →
The demo is the easy 20%
Anyone can show a slick AI demo. The invisible 80% (security, guardrails, monitoring, and the maintenance that keeps it working next month) is what separates a managed AI employee from a toy, and where most AI projects quietly die.
Read more →
Two Ideas for an AI's Second Brain — and the One We Built On
Andrej Karpathy's LLM Wiki spawned two answers to the same question: how do you give an AI a persistent brain in plain files? They're not rivals — they stack. Here's why we built our Business Brain on OKF, and why packaging matters.
Read more →
Stop reading case studies. Ask us to prove it on your data
In a market where anyone can generate a slick AI demo, the polished case study is worth less than ever. The honest test is whether a vendor will run their AI employee on your real work, in front of you, before you pay.
Read more →
You're Already Paying for the Employee You Haven't Hired
The case for an AI employee at a small business isn't a leap of faith. It's arithmetic. Every missed call, cold lead, and empty chair is already a bill you pay to no one.
Read more →
The renewal that lapsed on a Tuesday
A policy comes up for renewal, nobody follows up, and a client you already won quietly leaves. Here's what a lapsed renewal costs an independent agency, and what changes when an AI employee named Dana works every one.
Read more →
You're the Right Size to Be Born AI-Native
The Fortune 500 is spending this year trying to retrofit itself into an AI-native company. A founder-led shop can just start there.
Read more →
Why ChatGPT forgets your business every morning
A generic chat window is a brilliant assistant who reads your whole filing cabinet, gives a great answer, then forgets everything by the next question. The Business Brain is the part that remembers, and the reason an AI employee works where a chat window doesn't.
Read more →
You Don't Need 300 Agents
Box's CEO says companies will soon run hundreds of times more AI agents than people. A founder-led business needs the opposite: one AI employee it can actually watch.
Read more →
You don't need the best AI. You need the right one
Chasing the best model is how most AI money gets wasted. The useful question isn't which AI is best. It's what job you're trying to get done, and whether AI is even the right hire for it.
Read more →
The job you're really hiring an AI employee to do
The Jobs to Be Done framework says customers hire a product to make progress. When the product is an AI employee, that stops being a metaphor — and it changes the first question we ask.
Read more →
Nobody answered, so they called the next office
A new patient with a cracked tooth calls at 3pm, hits voicemail, and books with the practice down the street. Here's what a missed front-desk call costs a dental office, and what changes when an AI employee named Maya answers every one.
Read more →
What a Managed AI Employee Actually Is (and How to Tell a Real One From a Chat Window)
The label 'AI employee' is on everything now. Here's what separates a system that does the work from a chat window that just answers questions — and what to ask before you hire one.
Read more →
How to hire an AI employee without getting burned
The way a provider sells you an AI employee tells you more than the demo does. Five red flags in the buying process, and the honest posture that protects you.
Read more →
The ROI we actually measure
The first question owners ask is whether an AI employee will replace someone. That's the wrong number. Here's what we measure instead, and why hours recovered beats headcount every time.
Read more →
The 200 emails nobody wanted to send
Tax season doesn't slow down because the work is hard. It slows down because clients haven't sent their documents yet. Here's what that chase costs, and what changes when an AI employee named Cora owns the follow-up.
Read more →
We'll even tell you what to ignore
Our free AI assessment maps where AI actually pays off in your business, gives you a starting stack and a 4-day plan, and is honest about the moves that aren't worth your time yet.
Read more →
What breaks when you hire an AI employee
The honest version nobody puts in the demo. Four things that go wrong in the first weeks of an AI employee, and how we handle each one instead of pretending they don't happen.
Read more →
Five minutes or it's cold
A new real-estate lead is worth the most in its first five minutes and almost nothing by the next morning. Here's the speed problem in the numbers, and what changes when an AI employee named Riley answers every lead in seconds.
Read more →
The pursuit you didn't chase
An RFQ lands from a client you've served for years. You'd win it. You pass anyway, because nobody has 40 hours this week. That decision is the most expensive one a firm makes, and it never shows up on the books.
Read more →
The call that came in at 9:47 pm
A homeowner with a burst pipe calls one company first. If it goes to voicemail, they call the next. Here's what that missed call costs, and what changes when an AI employee named Theo answers every one.
Read more →
What an AI employee needs before day one
Building the thing is the easy part now. An AI employee that's actually useful needs three things first — a real job, your materials, and a human who owns it.
Read more →
The $1,080 question
That's what a single RFI costs your firm to process. On a $5M project you'll answer a hundred of them. Here's what that looks like the old way, and what changes when an AI employee named Aria owns the log.
Read more →
The chair that sat empty
A no-show isn't a hole in the schedule. It's revenue that can't be recovered, and the average practice loses nearly one appointment in five. Here's what changes when an AI employee named Nora owns the front desk.
Read more →
The lead you never knew you lost
What happens to a law firm's intake call during a deposition — and how an AI employee named Mia keeps it from walking to the firm down the street.
Read more →
Which job do you hand to AI first?
Most owners freeze on this question and end up doing nothing. The right first job isn't the flashiest one — it's the one that repeats, runs on rules, and hurts when it slips.
Read more →
Why We Cut Our Assessment Form From Nine Questions to Four
The friction problem we didn't notice until we looked at it from the client's side — and what we changed.
Read more →
Welcome to the RSTS AI Blog
Who I am, what RSTS builds, what we've done so far, and what I'm documenting here.
Read more →