Hey Cortana

(27-May-2029)

“Hey Cortana, restore the Clinic database to 10:43 AM yesterday.”

I don’t remember when humanity gave control to the machines. It was a gradual takeover, like black mold inside your walls. It was slow. You didn’t notice until it was everywhere you look.

But here I was, a former operational database administrator, asking a machine to do the work I used to do.

“OK. I understand you are asking for the Clinic database to be restored to 10:43 AM yesterday. Should I schedule that for you now?”

“Yes, schedule it to start in one hour, thanks.”

“OK, I have scheduled a restore of the Clinic database to start in one hour.”

Backups were out of our hands even before The Girls arrived. AWS and Azure managed backups for years. This allowed operational DBAs to do restores without the headaches of configuring backup jobs. Backups, replication, disaster recovery – all part of the database as a service. No human needed.

In the decades before the arrival of The Girls, the number one task for any operational DBA was the recovery of data. If you couldn’t recover data, you couldn’t keep your job. This fact was still true today.

Despite The Girls being able to perform vital operational DBA tasks, I still had a job. The Board felt more comfortable having at least one human around. To keep an eye on things, I guess.

The Board were also the ones to first refer to, and thereby name, Cortana and all the female-voiced AI in the office as The Girls. In their mannish minds, this was somehow more evolved than “Our Girls”.

“Hey Cortana, send an email to all business stakeholders. Tell them the restore will begin in one hour, so they should log off the system.”

This is the human element the Board liked. Cortana may not always know exactly who to contact about a restore. In time, we will build the rules to help her understand. And she was learning already, keeping track of what databases and what teams to notify. It was as if I was training a human. Except I wasn’t.

“OK, I understand that you want me to send an email to Albany Steakhouse to tell them the restore will begin in one hour. Shall I send that now?”

Learning can’t come fast enough, I thought to myself. But this was my mistake. We had converted our email lists when The Girls arrived. As part of their training, we configured email lists tied to a server, database, or application name. This would help us interface with Cortana and speed up the learning process.

“No.”

“OK.”

“Hey Cortana, send an email to the Clinic server to remind them that the restore will begin in one hour, so they should log off the system.”

“OK, I understand that you want me to send an email to Clinic server to remind them the restore will begin in one hour. Shall I send that now?”

“Yes, thanks.”

“Email sent.”

Database restores should be easier than this, I thought. In the days before The Girls, this restore task would have been completed already. But the voice commands make restores a lengthy process.

Restores weren’t even necessary anymore. But we did them anyway because humans were still used to having humans do certain tasks. The Girls could reload specific chunks of data as needed. All you had to do was ask. But the Board wasn’t comfortable with the idea of data flowing in and out so easily, all done by AI. How would the business ever know if the data was correct?

A while back we tried to build markers into the schemas to identify good datasets from bad. But that solution turned out to be a mess. If you’ve ever seen Wikipedia editors arguing over an entry, then you can imagine what it was like when two salesmen argued over a sale.

I stood up to get another cup of coffee. The next hour I planned to spend with coffee and a magazine. I would rather surf the internet, but the Board monitors usage. I didn’t want to risk my employment on a few Hulugrams. Reading magazines wasn’t monitored, yet. It would be another six months before the Board installed cameras. The Board was under pressure from Wall Street to have employees working at full capacity.

I was about to nod off when the phone rang, startling me and causing me to spill my coffee on my pants. It was Brad, and he wasn’t happy. Brad was on the Data Analytics team. He yelled at me, saying the restore didn’t work. Half asleep I hung up the phone and got to work.

“Hey Cortana, what is the status of the restore?”

“I don’t understand.”

No, of course you don’t, because you aren’t a DBA, but I’m forced to use you if I want to have a job.

My job depended upon two things. First, that the Board wasn’t comfortable turning everything over to The Girls. Other companies were willing to take risks, but not ours. The Board wanted a human to be here, and I was lucky that they selected me. Many of my operational DBA colleagues were finding it tough to get work these days. Some were forced into dreadful roles such as network administration.

The second thing is that I had the skill to talk with Cortana. I took a class on AI Interfacing two years ago. It taught me to speak efficiently with voice-activated devices. I knew my mistake and corrected myself.

“Hey Cortana, what is the status of the Clinic database?”

“The clinic database is in an online status.”

That’s the info I needed to know. The database wasn’t in a restoring state. The restore either was complete or did not start. I wasn’t sure of which. But before I could ask Cortana, Brad appeared on my screen. He was using YouFace for a video chat. Brad only uses YouFace when agitated. Brad told me that the daily 6AM data load from yesterday wasn’t there. And that it looked like the database was from two nights ago.

How odd, I thought.

I decided to check Cortana’s work. But you can’t just talk to Cortana like a human. If you do, you won’t get the answer very fast. You need to know how to ask the right questions. This was my specialty, some called it my superpower.

“Hey Cortana, what was the last point in time restore for the Clinic database?”

“The Clinic database was last restored to the point-in-time 10:43 AM UTC yesterday.”

That sounded correct, it was what I had asked for. But Brad says the data is from two days ago. I sipped my coffee and told Cortana to start the clock.

“Hey Cortana, start the downtime clock.”

“OK, I have started the downtime clock.”

A year ago someone wrote a program to calculate the exact cost, to the second, that an outage was costing the company. The Board then decided to have downtime clocks installed on each floor. Part of my job was to tell Cortana to start the downtime clock whenever there was an outage. The Board wanted everyone to see the severity of the outage in a way everyone could understand: money.

Instead of showing time, the downtime clock showed dollars.

Seeing the numbers on the wall didn’t make me work any harder. Or faster. Or smarter. I knew we weren’t losing money. The clock was a façade, an old-school scare tactic put into place by managers who motivate people by yelling at them.

Brad had told me about the program last year. Brad said The Board did a side project at the same time when building the downtime clock. After the clock added up the dollars, The Girls would take the final tally and adjust the billing for the upcoming month. The service fee increase got buried inside the monthly statements. Few customers would notice because everyone pays bills through auto-pay. For those that do notice, if they complain, they get adjustments the next month and a “Hey, we’re sorry” email.

So, we won’t lose money with the outage, but we won’t make money either. The following month the adjustment is reversed. Well, unless there were more outages.

I didn’t need a clock to remind me to focus on the problem. I was so focused I went to take another sip of my coffee and forgot I finished it earlier. Annoyed that Cortana couldn’t deliver a coffee to my desk, I got up to get more.

Damn machines are supposed to save us time, I thought.

Time.

The mug fell out of my hands, shattering into pieces. I realized my mistake. As the Bruumba came to clean up my mess I threw myself into my cubicle chair.

Ever since The Girls came into service it was clear they were excellent at doing what you asked. The hard part was matching what you asked for with what you wanted. The Girls didn’t make mistakes, they did as we asked.

Despite our advances as a society, we still needed time zones. Ask The Girls what time it is and they report the time based on their home location. But the Cloud has no concept of time zones. All data in the Cloud is UTC.

I messaged Brad. “Figured it out, will be back online in less than 15 minutes. Sorry for the delay.”

Then I went to work on fixing the problem.

“Hey Cortana, restore the Clinic database to 4:43 PM UTC yesterday.”

“OK. I understand you are asking for the Clinic database to be restored to 4:43 PM UTC yesterday. Should I schedule that for you now?”

“Yes, thanks, starting now.”

“OK, I have scheduled a restore of the Clinic database to start now.”

Mad at myself for not realizing my mistake sooner I went back to my magazine. The restore would take a few minutes. When it finished, I would message Brad.

At least I still have a job, I thought.

I knew the Board would get comfortable with The Girls running everything and I will get a new job. But there’s no need for me to speed up that timeline by making silly mistakes because of time zones. You don’t need The Girls to do everything. Many DBA tasks, and DBAs, are replaceable with a few lines of PowerShell.

I flipped the page of my magazine, looked at the time on my phone, and wondered how long before my car would drive itself here to take me home. I looked at the downtime clock and did some math. This outage was going to raise the bill for our customers by an average of seven BaCoins.

No one would ever notice such a small amount. And no one, not even the Board, would think it was a human error. They review the log files on a periodic basis. When they do, the multiple restore attempts will be flagged for a follow-up review. And that’s assuming Brad doesn’t say anything prior to the log review.

I sat back in my chair, wondering what network administration was like as a career.