Astute Technology Management

Not every task that could be automated should be. Before handing a piece of work over to an AI bot or a script, it's worth sorting what you're looking at into four categories.

We found a recent article in Entrepreneur magazine enlightening, where Meghna Deshraj provides a framework for making that determination.

In the article, she outlines four “buckets” that businesses can use to categorize each task: judgment, relationship, repetition, and coordination.

Tasks that are judgment calls carry real accountability, strategy, or negotiation. Automating these doesn't just risk a worse outcome; it removes the ownership from decisions and degrades the chain of accountability.

Relationship work includes sales conversations, partnership building, moments that depend on trust. These lose their value the instant they're handed to an AI script, even if the script can technically do them faster.

The next two categories—repetition and coordination—are where automation genuinely makes an impact.

Repetitive, high-frequency, and stable tasks, like data entry, scheduling, and templated communication, are exactly where automation tends to pay off, often more reliably and consistently than a person doing the same task for the hundredth time.

Passwords are a “cybersecurity 101” measure, and many organizations are getting them wrong.

The idea that passwords need to be an impossible-to-remember line of nonsense characters is over 20 years old. It started in 2003 when Bill Burr, a manager at the National Institute of Standards and Technology (NIST), proposed the idea.

He hoped that it would prompt people to create stronger passwords than they were creating at the time.

The opposite ended up happening. This complexity meant that people started using predictable patterns to satisfy the password requirements that computers could guess with ease.

Since then, our understanding of security has evolved a great deal, and Mr. Burr has backtracked on that idea, saying he was “barking up the wrong tree.”

Length is now considered more effective than complexity—longer passwords are harder to crack.

Combining multiple words that you can remember, using at least 15 characters, creates a much stronger password than an arbitrary mix of uppercase letters and special symbols that you’ll struggle with.

The Ford Motor Company is another major company that is finding its own unique path to AI transformation.

Instead of taking an “AI will replace workers” approach, its doing the opposite: hiring 350 veteran engineers, specifically to make its AI systems better.

According to Ford’s VP of Vehicle Hardware Engineering, Charles P**n, “Artificial intelligence is a fantastic tool, but it’s only as good as the information you use to train it.”

The engineers they’ve hired run mandatory weekly peer design reviews to catch problems early, reprogram AI tools that aren’t performing well, and mentor junior engineers.

The payoff has been substantial: Ford jumped from 10th to 1st place among mainstream brands in J.D. Power’s initial Quality Survey.

Recalls have also dropped noticeably, and the company expects roughly $1 billion in warranty savings this year, after annual recall costs had ballooned to $4.8 billion.

If you’re asking us, that paints a clear picture of the power and hype around artificial intelligence. It’s no doubt a powerful tool, but it works best when paired with veteran engineering talent.

Construction companies have allocated a significant portion of their budget to technology that helps with preconstruction tasks, such as faster estimating and faster bidding.

If you put too much emphasis on the front end, you may be relocating your bottlenecks.

If the takeoff data, estimates, and scoping information aren’t flowing into a coordinated system before the crews show up, small design discrepancies can become expensive field problems.

This is a data ingestion problem, not a software problem. Building Information Modeling (BIM) platforms only create value when the process of getting information in—cleanly and consistently—is built to carry that information forward from bid to construction.

Speed at the bid stage is becoming table stakes. The next differentiator may be whether your construction firm’s data survives the trip from proposal onward.

Do you feel your data is available at every stage of preconstruction and construction?

Most privacy regulations were built for a world where a human being had to sit down and access a sensitive record, but that way of working is going away fast.

An automated tool doesn’t query one customer file at a time—it queries thousands of them in seconds.

[] For all the hype around “AI agents”, they can also be a huge compliance liability, because they default to the broadest access available, not the minimum needed to do their work.

[] Large language models like Claude and ChatGPT can potentially re-identify “anonymized” data when they analyze information from separate databases.

Big companies are already facing penalties, and small businesses should take note.

Amazon paid a €746 million GDPR fine related to targeted-advertising data, and that was before agentic AI (systems that can act autonomously on a user’s behalf) was doing any work in their network.

What’s the takeaway? If your organization wants to use AI anywhere near customer or employee data, access governance can’t be an afterthought; it must be one of the first things you think about.

There’s a difference between a company that manages technology and a team that can provide the leadership a business needs to make the most of its technology.

Business fluency matters as much as technical fluency.

Our most successful client relationships are with companies that see us as strategic partners, not just technology vendors. What’s the difference?

The first is someone who understands your business.

A managed IT services provider (MSP) can’t “think outside the box” and help you creatively solve problems with technology if it doesn’t first understand your business.

That business fluency should extend to understanding your budget and how procurement, governance, risk management, and strategy all impact the management of your technology stack.

Relationships and communication are real differentiators. After more than 20 years, we have learned the art of explaining technology to diverse audiences without being condescending. That human connection is critical.

Email is the most common business communication technology, with 392 billion emails sent every day; it’s also a source of serious cybersecurity insecurity.

Why? The reason is simple: decades-old technology.

You see, most mainstream email providers, such as Gmail, iCloud Mail, Outlook, and Yahoo Mail, still don’t use end-to-end encryption (E2EE), a feature common to many messaging apps like WhatsApp or Apple iMessage.

Instead, the technology that email is built on is designed to be easy to access. Features such as spam and phishing protection, cloud storage integration, and searchability all depend on a provider being able to read the content of your messages.

For businesses that need the highest level of confidentiality, most commercial email providers can be configured to provide either true E2EE or something that comes very close, but the trade-offs are not always worth it.

Would your business benefit from stronger encryption?

An interesting tale from the trenches of AI deployment.

The system was designed to scan backroom storage shelves using iPad Pros equipped with computer vision, spatial computing, and augmented reality.

Just nine months later, Starbucks quietly killed the program, announcing its “retirement” via an employee newsletter.

The issues were more mundane than you might think. Reflective surfaces caused duplicate counts, while milk products were frequently miscategorized by brand/type.

This had a human dimension. Some baristas were blamed for AI errors, while managers described being pressured to discipline staff for “missed counts” that were the AI’s fault.

Most interesting to us as a team of computer nerds was that Starbucks’s inventory system still uses 1990s-era IBM AS/400 technology. The system struggled to feed clean, real-time data to the more modern systems.

What’s the takeaway? We feel that AI will be a transformative technology, but without the right technology to support adoption, you will run into unexpected challenges that may undermine your efforts.