What Are Companies Really After with AI? Let Me Tell You 7 True Stories

Every spring, lawns wake up. For a company called Rachio, this is peak season — and a nightmare.

Rachio sells smart sprinklers. The moment it gets warm, households everywhere start watering their lawns, and customer-service tickets come in along with the water: devices that won't connect to WiFi, valves that won't open, schedules jumping around. More than a million users, and every problem trickier than the last.

The old solution was primitive: frantically hire a batch of seasonal workers every year. Recruit, train, deploy — and when the season ends, send them all away. Year after year.

Then they changed their thinking: hand customer service over to AI.

The result? One customer-service lead, with one AI system, took on the inquiries of over a million users. Accuracy: 95% to 99.8%. Cost: down 30%.

When I saw this case, three words popped into my head: This is insane.

Honestly, the past two years have seen too many articles about AI, and most are concepts — you finish reading and still don't know what to do. So I went through every case I had on file where companies spent real money on AI — and had the numbers to show for it. Seven in total.

Today, I'll tell them to you. As we listen, we'll do the math.

Shift One: Customer Service

Back to Rachio first.

Its tool is called Crescendo.ai. Note that this is not the kind of dumb bot that says "Sorry, I didn't understand your question." Sprinkler failures require knowing the device: resetting it for you remotely, walking you through WiFi setup step by step. This kind of work used to be something only a trained technical support agent could handle.

What does an "AI + human" hybrid mean? AI charges in front, catching the vast majority of routine questions; whatever it can't judge, or whatever comes with emotions running high, gets passed to a real person. That's how one person can manage a million users, and how that yearly round of "hiring temps" got canceled for good.

On the customer-service shift, AI has already taken the watch.

Shift Two: Writing Code

You might think: customer service is a standard process, easy enough. What about creative work like writing code?

Duolingo tried it for us.

It's the world's largest language-learning platform, with 300-plus engineers managing 400-plus microservices. They wired GitHub Copilot, an AI coding assistant, into their development process. How did it work? Look at the numbers:

Engineers in new repositories got 25% faster; veterans, 10%. The median wait time for code review dropped 67%. What does that mean? Reviews that used to queue up now jump the line. The number of pull requests rose 70%.

There's a hidden benefit too: new hires get up to speed much faster. In an unfamiliar codebase, the AI can read alongside you and generate boilerplate code for you. Before, they'd spend their first two weeks on the job lost; now they get straight to work.

When it comes to writing code, AI is like a tireless junior partner: it covers all the repetitive work, while the hard judgment stays with humans.

Shift Three: Selling

Cutting costs is only the first step. Can AI directly help make money?

Starbucks has a homegrown AI engine called Deep Brew. What it does, in one sentence: send the coupon to the right version of you.

It makes personalized recommendations for more than 30 million members. Based on what? Your purchase history, the current time, even today's weather in your city. Hot day, it pushes iced drinks; rain, a hot latte. You may have just sat down, and the ordering page on your phone has already figured it out for you. What is hyper-personalization? Exactly this.

The math is easy: the more accurate the recommendations, the more frequent the repeat purchases. Since this system rolled out, Starbucks' membership program has grown to nearly 35 million members in the U.S. alone, and same-store sales climbed along with it.

Even better, Deep Brew also took over chores like scheduling and inventory, and store managers finally have their hands free to do what only humans can do: chat with customers, remember the regulars' tastes.

Shift Four: Driving Traffic

The next story hides where you can't see it.

Rocky Brands, an international premium footwear brand, plugged BrightEdge's AI SEO platform into its operations and fully automated the keyword-watching that used to be done by hand.

The AI did two things. First, dig for keywords: using its Data Cube technology, it dug up thousands of high-value terms that competitors had overlooked but real users were actually searching. Second, scale up: automatically generating and polishing title and description tags for thousands of product pages.

Think about it: thousands of product pages, retitled by hand one page at a time — how long would that take?

The results: search-channel revenue rose 30%, overall revenue rose 74% year over year, and new customers grew by 13%.

Plainly put, in the search-traffic game, the contest is now about who looks closer and who moves faster. On that front, humans can't out-watch machines.

Shift Five: Mentoring

In the first four shifts, AI faced outward. The next two turn inward.

Britannia, a consumer-goods giant with 130 years of history. When an old company wants to go digital, the difficulty is: the body won't move. Their employee capability assessments relied on Excel spreadsheets, and one manual round took 10 weeks.

Then they partnered with Edrevel, an AI workforce platform, and moved the assessments into the system: multilingual, covering 18 factories and 431 employees. Assessment time was cut by 75% outright, saving 280-plus hours of labor in phase one alone.

The real change came after: because assessments got faster, they switched from once a year to once a quarter.

Assess once a year, and you're settling the books at year-end — by the time you find a problem, it has been growing for a year. Assess once a quarter, and you correct course as you walk. The first is a retrospective. Only the second is development.

Shift Six: Making Videos

Zoom, the big brother of video conferencing, couldn't escape a problem of its own: the product iterates so fast that internal sales-training videos simply couldn't be shot in time.

Bring in an expert on camera, set up the space, record, edit — several days per video. Worse, every time the product updates, the videos have to be reshot.

Then they adopted Synthesia, an AI video-generation platform. Now the way they make videos is: write the script, let the AI generate an avatar, ship the video. One video, under an hour — 90% faster than before.

Changing content? Just edit the script, and a new version comes out in minutes, no reshoot.

The cost? $1,000 to $1,500 saved per person per month. What's saved is the experts' time — taking days off to appear on camera and read scripts in front of it.

The wildest part is output: one designer produced 200-plus micro-videos in six months. Good grief. On a traditional shooting schedule, you'd have to hire another platoon of designers.

Shift Seven: Watching the Floor

One last topic, a slightly sensitive one: AI watching employees work — is that acceptable?

Affordable Staff is an outsourcing provider. The peculiarity of this business: clients aren't buying attitude — they're buying hours. Whether you actually worked, and how much — clients will ask.

They adopted Hubstaff's AI time tracking and workflow analysis: automatically logging activity status, automatically generating work reports, handing clients "100% proof of work." Managers no longer have to babysit people manually, and time spent on supervision dropped 80%. As a bonus, the system can also recognize from the data who performs best, and coach employees who are temporarily falling behind.

The financial ledger: $928,750 saved per year, and $4.2 million cumulative over twelve years.

On surveillance, I have reservations. Used well, it's called transparency — the people doing the work get seen, and get priced fairly. Used in excess, it becomes the overseer, and hearts go cold. Tools take no sides; the people using them must know where the line is. Affordable Staff's use holds up because "verifiable hours" is exactly what they sell to clients.

The Final Tally

Seven stories told. Customer service, writing code, selling coffee, driving traffic, mentoring, making videos, watching the floor.

Have you noticed? These seven shifts share one thing: what AI takes over is all repeatable work that can be written down as process, and the time people free up all goes to judgment, and to caring for the things only humans can care for.

Whatever can be written as a process deserves to be handed to AI; whatever requires judgment is where humans step in.

Back to that waking lawn at the beginning. The lawn will wake every year, but from now on, Rachio's customer-service lead no longer has to dread the yearly search for seasonal workers. What he leads is a system that never gets tired, never forgets, and is always on call.

And here's my wish for you: that you're the one who puts AI to work.