Skip to main content
x

Interview with Jens Hagel, Founder & Owner, kjello

Interview with Jens Hagel, Founder & Owner, kjello

(Featured) - The kjello founder explains how to choose licenses, secure devices, clean up access and deploy laptops remotely.

Edited by Jenn Moon, Editor at Featured News · Nov 22, 2026

As founder and owner of kjello, what does your work in computer networking involve, and what have more than 20 years supporting small businesses in Hamburg taught you about the IT challenges they face?

I take care of the IT for small to midsize companies. Most of them have between five and 50 employees. Microsoft 365 is the norm. I manage devices, user accounts, security and backups. kjello was made for that.

There is a lack of ownership of the IT department. It's always someone who handles it as a side job. For example, a 150-person construction company had its head of finance doing it as a “power user.” He told me that if he has to install Word or a PDF reader, he can't do his job.

What path led you from working in IT to founding kjello, and how has that journey shaped the way you help small businesses make technology dependable and manageable?

I founded hagel IT in 2004 as a one-man company working from home. In 2013, I told my team we were going cloud-first: Office 365 instead of a mail server in the cellar. One team member quit because of that. It was the best decision we ever made.

kjello was born out of the same thought process. A company orders a device through the portal, and the device is shipped to the new employee. When they turn it on for the first time, it configures itself: user account, company email, applications, access and encryption.

When you first review a small company’s Microsoft 365 setup, what is the clearest sign that it is paying for licenses or tools it does not actually need?

Enterprise licenses in small companies. They bought them at some point because they needed one function.

One haulage company in Hamburg with 15 employees was paying around €45 per user per month for an Enterprise license. With Business Premium, they had everything they needed for around €20. That means €25 saved per user per month: €375 per month or €4,500 per year. A big chunk of their monthly IT bill.

You have helped clients move from Enterprise licenses to Business Premium; how do you determine which Microsoft 365 license level fits a small business’s real needs?

I always try to figure out what they do day to day and then match it to the license.

A development company in the Hamburg area with a little over 20 users had Enterprise licenses. They used a virtual private network (VPN), a secure remote connection to their network, for most of their work. Once their files were on SharePoint, they no longer needed it. We switched them over to Business Premium. They used Planner instead of Project and Microsoft Defender instead of their antivirus software.

There was one mistake I made. I told them that a certain version of Defender was included with Business Premium. We checked with Microsoft and corrected our mistake.

What is one security capability included in Microsoft 365 that small businesses often overlook, and how have you helped a client put it to practical use?

Manage your devices with Intune. It comes with Business Premium. You can encrypt laptops and allow or disallow certain devices from accessing your company data.

We had one client who had provided output management services to big banks for over 20 years. They were running their company network and family Wi-Fi through the same router and were on a Microsoft 365 Family plan. One of their clients, a bank, sent them an information security contract. The owner admitted that he had been using AI to help him write his security concepts. We got him up and running with business licenses, Intune, BitLocker and 2FA in a few days. He was able to sign the contract with the bank.

For a business that already pays separately for antivirus software, how would you assess whether Microsoft Defender included with its Microsoft 365 plan can cover its needs?

What does your plan include? With Business Premium, you have Microsoft Defender for Business. So, if you’re running another antivirus, you may be paying twice. Ask yourself: What does my other antivirus do that Defender doesn’t? Who is monitoring those alerts? If the answer is no one, then your second license is probably not worth it.

It’s also important to know your suppliers. When the Federal Office for Information Security (BSI) released its warning about Kaspersky, we let our clients know on March 15, 2022.

Drawing on your work cleaning up growing organizations’ IT environments, what is the most useful first step a small business can take to regain control of user accounts and former employees’ access?

Take your current list of licenses you’re paying for, and compare it with the number of employees you currently have. You should be able to find discrepancies.

We had a client reach out to us after a phishing attack. They had fewer than five full-time employees. We found 11 paid licenses of various types, former employees who were contractors and 15 shared links, six of which had public access. We sat down with the owner and went through their invoice. We found their actual number of employees and provided them with a technical contact.

What practical policy or workflow have you found most effective for helping teams share files and links safely without creating confusion or exposing sensitive information?

Store your files in one location and allow teams to access them. Share files with specific people.

A company in Bremen acquired another company with 18 employees. The employees were working on their personal laptops with admin rights. They had terabytes of images stored on a network drive that had no backup. Within four weeks, they had everything on SharePoint with access controls. The MD told me he had no idea how close they were to losing everything.

When a small business has no technician on site, how do you set up a new laptop so that a new employee can securely start work with the right account, apps and protections from day one?

The device is shipped directly to the new employee and sets itself up using Microsoft’s Autopilot and Intune, which is what kjello does as well. The company selects the device in our portal, and the device configures the user account, company email, applications, access and encryption when it is turned on for the first time. There is no need for a technician to come out.

We have a machine builder client with 42 employees in Hamburg and Leipzig. They used to set up PCs using a checklist of 20 items. The MD told me he spent half a day doing that. Now they can set up a workstation in 15 minutes.

Thanks for sharing your knowledge and expertise. Is there anything else you'd like to add?

You are probably already paying for more than you need. As I always say, most companies have Microsoft 365 but don't utilize it to its fullest.

Don't overdo it. Turn on 2FA and start adding features to your device management.

About the expert

Jens Hagel

Jens HagelFounder & Owner at kjello

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

Interview with Jens Hagel, Founder & Owner, kjello

Interview with Jens Hagel, Founder & Owner, kjello

(Featured) - The Hamburg IT firm’s lessons include kill switches, spending caps, checks against silent failures and fewer alerts that people will actually read.

Edited by Jenn Moon, Editor at Featured News · Nov 19, 2026

As founder and owner of a Hamburg-based IT services firm, how would you describe the work your company does and the operational AI systems you run today?

hagel IT provides IT services to small and medium-sized businesses in northern Germany. We have 32 employees and over 150 customers.

AI agents handle many tasks on our websites, blog, social media and servers. I create these agents with AI coding agents. Our own software products are also created by AI coding agents. At night, on weekends, or when all lines are busy, an AI assistant answers our phones and books appointments in our technicians’ calendars. In case of an emergency, it calls our on-call technician. Between 8 a.m. and 6 p.m. on weekdays, you will always reach a human.

What path led you from computer networking and IT services into building and operating AI-driven automations for your own firm and clients?

I started the company by myself from my home in 2004. Since 2025, Philip Kraatz has been running the company as CEO. I am the founder and owner.

One customer had orders coming in as PDFs. For every order they received (30-40 per day), they had to enter it into their enterprise resource planning (ERP) system, which handles their orders and invoices. Each order took 15–20 minutes to process, and if someone made a typo in the part number, the wrong item was delivered. We built an AI system that can read the email and PDF and create a table and invoice. Now each order can be checked in under a minute.

When you first put agents into real workflows, what made you decide that prompt instructions alone were not a sufficient safeguard?

They will follow the rules most of the time. Most of the time isn’t enough when your agent is working with customers’ websites or sending out emails.

I moved those restrictions into code that checks what the agent wants to do before it does it. You want to give your agent restrictions that it can’t break — with code. A prompt is something it can forget.

How have you translated your non-negotiable boundaries — such as prohibiting deletions, account changes or unapproved external contact — into technical checks that other small firms could adopt?

Each task your agent wants to do has to be checked against something. If it’s something on your “Don’t Do” list, such as other people’s domains, deleting data, or changing accounts or subscriptions, your agent won’t be able to do it. Emails are only allowed to be sent through one channel, with a limit on the number of emails per day, and the agent always copies me.

You can implement this with little work if your company is small. Make a list of things that will hurt you. Make sure your agent can’t do them through code. Make sure all emails go through one channel that you can check. I tell my agents what I want them to do, give them context and restrictions, but not how to do it.

What does an effective kill switch look like in a small-team environment, and what did your experience with queued jobs teach you about what it must stop?

It should stop your agent immediately, without having to deploy anything, and your team should understand how it works. Our website internal linking tool will stop when one file is deleted.

It will also stop itself. If three articles drop by more than three spots on Google within 14 days, it will stop itself and email me. It is also limited to making only five changes a day across all our websites.

The queue is what I learned from. It will stop new jobs from being added to the queue, but it will not stop jobs that are already in the queue. So you would have to stop those as well.

For customer-facing tasks such as order or invoice processing, how do you decide where an agent can act independently and where a person must remain the final approver?

For our manufacturer customers, the AI will prepare everything, and a human will check it before it reaches the customer.

This check takes under a minute. They are able to send out their invoices immediately instead of waiting 24-48 hours. One of our customers’ CEOs initially doubted that it would work because their orders varied so much that they weren’t sure how AI could understand them.

What practical limits have helped you keep AI-agent spending predictable, particularly when workflows can run repeatedly or call several services?

Limit how long your agent can run. One job in September 2026 would have cost us around $7,500 if we hadn’t had this limit. Ninety-three percent of that cost would have been from sub-agents.

Now, our agents stop after around 200 steps and pass the work to a new agent. We do this with code as well. Our agents are told to wrap up at 180 steps and can’t use any tools after step 221.

After a model provider change left your social-media pipeline failing silently, what monitoring checks did you add to confirm that a workflow completed rather than merely started?

Our social media automation was using a free large language model (LLM) that our provider cut off in July 2026. The automation failed on the first AI step, but nothing looked wrong. It took us around 10 days to notice that we weren’t posting.

I now ask my checks whether the task ran today, instead of asking whether there has been any activity in the last x amount of time. I also added checks that stop our builds if they don’t create at least a certain number of pages.

What have you learned about designing alerts that a busy owner or operations team will actually notice, understand and act on?

Employ fewer of them. One of our website checks emailed us on 26 of 41 days. Only four were actual outages. One deployment would send us 19 emails every 10 minutes. Our server checks were sending us around 200 emails a week.

We now start an AI to fix our issues if they turn red. We send an email only if they are still red afterward, once per incident, and provide a weekly summary. If half your inbox is from your automation, you won't read the emails.

Thanks for sharing your knowledge and expertise. Is there anything else you'd like to add?

Tell your AI what day it is. One of our companies' automated LinkedIn posts was telling people that Windows 10 would end support soon. Support ended in October 2025. A business partner asked us if we were a year late. Our AI wasn't aware of what day it was, and we had a check in place for old information that was too strict.

We had to edit 182 passages in our 41 blog articles. We now tell our AI when something ends: “Since October 14, 2025” instead of “soon.”

About the expert

Jens Hagel

Jens HagelFounder & Owner at kjello

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

A client asked for a new firewall, and the firewall wasn't the problem

A client asked for a new firewall, and the firewall wasn't the problem

(Jens Hagel) - When a technology company near Hamburg called about a firewall, asking what was wrong with the old one changed the whole job.

Edited by Jenn Moon, Editor at Featured News · Nov 19, 2026

In 2025, the chief technology officer of an IT company based near Hamburg got in touch and requested a new firewall. At first, it seemed like a straightforward inquiry — he was looking for advice on what to buy and an indicative price.

He told me the company employs about 48 people and produces various pieces of technology to support monitoring of key assets and infrastructure. He was the head of product development at the company and was familiar with the major firewall offerings on the market.

I was in a position to provide an indicative price for that on the same day. However, I inquired what the problem with the existing firewall was.

A symptom of a wider problem

It transpired that much of his frustration was centered on the personnel surrounding the firewall. It had been provisioned by their external IT partner, and the configuration was not known by anyone internally. Should he want to enable, for example, a virtual private network (VPN) to allow an additional staff member access to the network, he had to contact that partner. He described the partner as slow to respond and providing limited assistance. As a result, minor IT-related issues were falling to the head of development to action — clearly not the best allocation of his time.

Replacing this with a different make of firewall would not have addressed his main frustration. Without changing how the firewall was managed, we risked running into the same problems again.

After our conversation, he was looking for a way to improve the IT function in his business to relieve him from these day-to-day tasks.

Why I ask this question

When I set up my managed service provider (MSP) business in 2004, I found that inquiries I received would typically be for a specific piece of kit. A server, firewall, end user devices, etc. — things that are easily identifiable to procure. The issue they're trying to overcome is less clear-cut, so this information rarely surfaces unless asked.

As a result, I always try to ascertain what's not working before recommending kit. Here we were able to build a solution using further Microsoft 365 services that the company was already leveraging.

The proposal

The following is a summary of the proposal we put together with him, in the sequence we planned to deliver it. A managed firewall to replace the existing one but which gives the company greater autonomy to, for example, provision additional VPN access. Microsoft Intune to apply policies and restrictions to the company's managed devices. Windows Autopilot to enable end user devices to be self-service-configured by the end user on first boot, removing the requirement for IT to configure each new device.

It is too early to provide any detail on the success of the device management services, as this will follow the firewall implementation. However, we have given him a roadmap to improve his IT services, starting with the firewall.

Is it ever inappropriate to ask?

Yes, of course. It is entirely reasonable to assume that if you request a firewall and are met with a discussion about endpoint device management, we are trying to sell you something. There will be occasions where a customer knows precisely what they want, and in these cases I will simply provide a price. I may even advise that we are not the best company to carry out that task. One example was a machinery distribution business looking to deploy Universal Print to roughly 125 locations across Europe. This was outside of our skill set, so we directed them to other members of our channel network.

However, the firewall issue did not seem to be that scenario. There was a degree of frustration from the customer about a lack of response and knowledge, and replacing the firewall alone was unlikely to alleviate that.

2 questions to ask before approaching your IT provider

I would suggest that if you are a small or medium-sized enterprise (SME) are and approaching an IT provider about a specific piece of kit, ask yourself the following questions:

  • How is that currently failing?

  • Who has to deal with it?

In the example above, the latter revealed to me that the head of development had to support his colleagues with this, and gave me a better indication of the issue than his mention of the various firewall brands.

There is, of course, a risk in doing this, and you will take longer to provide a quote — potentially losing that customer to a more responsive competitor. I think it's important to respond to customers promptly, and we have received positive feedback in the past from customers who noted we were the only provider that responded. I therefore try to seek that information on the initial call to limit that impact.

About the author

Jens Hagel

Jens Hagel is the founder and owner of hagel IT-Services in Hamburg, Germany, which he started alone from home in 2004. His team of 32 people looks after information technology for more than 150 midsize companies in northern Germany. He also builds kjello, a platform for managed workplace IT.

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

The next AI revolution may happen underground, not in Silicon Valley

The next AI revolution may happen underground, not in Silicon Valley

(Dela Rostami) - By Dela Rostami

Edited by Jenn Moon, Editor at Featured News · Nov 17, 2026

As we interact with our screens, read news, watch videos, scroll through social media and have conversations with each other, we are often reminded of the power of AI in the world today. We hear of the many jobs that will be taken over by AI in the near future and the need for more and more computing power to support AI. We also hear of the many data centers that will need to be built to support the growing industry of AI. However, there is a very important part of the conversation about AI that is often left out. It is the mining industry that supplies the materials AI infrastructure depends on.

This industry is responsible for mining the materials that are needed to make chips, batteries, servers and electrical systems around the world. For example, copper is needed for electrical infrastructure, silicon for semiconductors and lithium for batteries. These are just some of the materials that are mined from the earth in order to support the growing industry of AI. I had no idea about this before my brother, who is a mining engineer, got me involved with mining. He has spent many years working for various engineering consultancy companies and in his free time writes algorithms. Through our conversations, I have learned a lot about AI and the mining world and how they interconnect.

At first, I was mostly listening to him explain the technical problems, and there were plenty of them: geology, mathematics, simulation, economics, mine planning, engineering, all that stuff that I know nothing about. However, I didn’t expect to learn so much about people and how they interact with technology. All the while, in Silicon Valley, we are focused on making the most trivial interaction with a platform as seamless as possible. On the other hand, in the mining world, engineers are trying to navigate a plethora of complex software, disconnected systems and information across a number of different tools for highly consequential projects. But why does understanding one decision require moving between so many other systems? Why are there so many silos? Why are there so many groups, and why don’t people just talk to each other?

I also began wondering how AI could help people make decisions faster in industries such as mining or civil engineering or any kind of physical infrastructure where the decisions made have a value of billions of dollars. How can AI and technology help bridge the gap? Can the AI explore thousands of scenarios, analyze information and do repetitive work while still being accurate? What kind of infrastructure does it need? How can it connect different siloed networks? For example, a geologist would first model the deposit of minerals. A mining engineer would then use the model to create possible mine designs. A process engineer would then need to understand the material that will be extracted. Finally, the financial team would need to know the cost of each decision.

All of these groups need to be able to run scenarios to see how each decision will affect the other. In many mining workflows today, if you change one thing in a scenario, you have to go back and manually update much of the rest. It takes a long time. A 2024 S&P Global study found that U.S. mines for critical minerals take an average of nearly 29 years to go from discovery to production. If we want to build sustainable AI, faster AI, better AI, better batteries, etc., then we need reliable access to the necessary materials, but a new U.S. mine that supplies those materials can take nearly three decades to come online.

One thing I've learned from working in UX is that complicated technology doesn't necessarily need to be complicated to use. That doesn't necessarily mean simplifying the actual engineering or hiding important details; it means thinking more carefully about people, what they need, how they need it and how they're making decisions. What we need to realize is that we have already been working on this for decades in the area of consumer technology. An example of this is the iPhone and the many years of engineering that went into making the screen work the way it does. The complexity is still there but it is organized and displayed in a very simple way for the user.

Industrial software is very different from consumer software. A mining engineer is not ordering a coffee from a café or scrolling through videos on TikTok. A mining engineer is trying to solve technical problems, dealing with uncertainty, making decisions that could cost millions of dollars and trying to understand the assumptions behind any calculation. But why does that mean the software has to be unnecessarily difficult to navigate? I don't think it should be. In fact, I argue that the more complicated the system is, the more design matters. I see a lot of attention in Silicon Valley being paid to the creation of new things, new apps, new platforms, new AI, new technology. There is nothing wrong with this, but it seems to me that there are many great opportunities in industries that are not glamorous.

The mining industry is a great example of an industry that is thousands of years old and, in the modern world, is responsible for providing the materials that make up the infrastructure for the AI that we read about every day. Scattered information, complicated workflows, people struggling to understand the consequences of their decisions, and incredibly sophisticated but difficult-to-use technology are all problems that exist in these industries and can be addressed by designing some of the most meaningful applications of AI for them. These problems exist in many of the industries that are responsible for the physical infrastructure of the modern world.

Perhaps some of the most meaningful applications of AI will be designed for these industries. If someone had told me years ago that I’d spend my time thinking about mine planning, engineering modeling and mineral supplies, I’d probably have laughed. But the more I learn about the industry and how much opportunity there is to rethink how people interact with complexity, the more interesting I find it. We can’t keep building smarter AI, faster chips and bigger data centers without eventually figuring out where the materials are coming from. Maybe we should be paying more attention to the people and technology responsible for figuring this out.

About the author

Dela Rostami

Dela Rostami is the Chief Product Officer and Co-Founder of AIMINR, an AI platform that helps mining engineers plan mines faster. She spent years in Silicon Valley working on UX and human-computer interaction, and is the author of two books on AI and design. She focuses on how complex engineering data is presented for high-stakes decisions.

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

Interview with Jens Hagel, Founder & Owner, kjello

Interview with Jens Hagel, Founder & Owner, kjello

(Featured) - The Hamburg IT founder discusses integrating acquisitions, separating teams, documenting responsibilities, and using Microsoft 365, Intune and Autopilot.

Edited by Jenn Moon, Editor at Featured News · Nov 16, 2026

As founder and owner of a computer-networking firm in Hamburg, what kinds of business-change IT projects have given you the clearest view of what companies get right — and wrong — when their structure changes?

What I have learned from quick buyouts is this: A company that owned several properties in Hamburg had taken over 22 companies in the span of two years. Users had administrative rights, and two-factor authentication (2FA) was disabled.

We offered them a support package of around 20 hours a month to help them help themselves. Most of this time went towards configuring Autopilot and Intune, Microsoft’s way of deploying and managing devices. Once this was done, all users had 2FA enabled, and new starters could be configured in minutes.

What path led you to build kjello and become the IT partner companies call when they are acquiring, separating, merging or rapidly scaling their operations?

I started hagel IT from my living room in 2004. We migrated to the cloud in 2013. We moved away from hourly billing to a fixed monthly price in 2019. We are now a team of 32 supporting over 150 companies in northern Germany.

Whenever a company changes, staff join and leave. Devices and permissions need to follow suit. kjello was our answer to that. The company can choose a laptop from our portal, and it will be delivered to the employee and configured on first boot. If someone leaves, we can begin disabling their access through our system.

When a client buys another business, what is the first IT discovery step you take to reveal the risks that could disrupt the combined company?

I like to conduct a cyber risk assessment with the managing director and then review the company’s infrastructure to find out where they stand with their newly acquired company.

A company in Bremen bought a specialist company with 18 employees. All employees worked on private laptops with admin rights. The company had terabytes of data on network-attached storage (NAS) with no backups, which would have been difficult to recreate. We moved everything to SharePoint within four weeks and restricted access. The managing director stated that he did not know how close they were to disaster.

From your experience, what information should an acquiring company document before deciding which systems, networks and user accounts to integrate or retire?

Document who has admin rights, what licenses you're paying for and where data is stored. Having admin rights will cause you the most headaches.

An investment company in Hamburg was removed from the group IT department of a larger asset manager. Because the group could not give them access to their Microsoft tenant due to password restrictions, we created a new Microsoft tenant for them. Their IT is now around 40% cheaper than what the group charged them.

When a team is being spun out of a corporate group, how do you help preserve the team’s access to the tools it needs without leaving the former parent company exposed?

Have the company being spun out configured independently. If the two companies share logins, it will only cause headaches for both parties.

A consulting company that was taken out of a corporate group had employees working in Hamburg, Frankfurt and at home. All their laptops and email signatures were configured through the group’s IT department. We built them a serverless office using Microsoft 365, Intune and Autopilot. Even their private iPhones were included, and they pay us a monthly rate per person.

After a period of fast growth, what IT problem tends to remain hidden longest, and what review can a growing company perform now to uncover it?

Excessive admin rights and missing or untested backups. You may not know there's a problem until something goes wrong.

One of our clients in Hamburg had around four employees when we first started supporting them. They now have over 30. One of their employees did IT as their second job, and everyone had admin rights. Our client told me they did not know if their backups were working. We removed their unnecessary administrative rights, and no one noticed. Always check who has admin rights, and restore a backup from the previous week to see if it works.

Across acquisitions, carve-outs, mergers and growth cleanups, what is the one IT habit you would put in place early to make the next major business change far less risky?

2FA on all logins that support it. When we first start working with a company, they almost never have this enabled.

We started with a law firm that had 11 employees and did not have 2FA. It became one of their top security features. When you're buying a company, you will have multiple logins being created. Having 2FA can limit the damage if login credentials are compromised.

Thanks for sharing your knowledge and expertise. Is there anything else you'd like to add?

Know who is doing what when a group of companies buys a company and keeps its group IT involved. We support one of our clients that has group IT in another country. We are the first point of contact, but we do not have permanent admin rights. So, we had to pass any simple tasks on to the group IT. We felt like we were doing work for them.

We now have these responsibilities documented and have started using time-limited administrator credentials.

About the expert

Jens Hagel

Jens HagelFounder & Owner at kjello

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

Interview with Roman Martynenko, Full-Stack Software Engineer and Founding Engineer

Interview with Roman Martynenko, Full-Stack Software Engineer and Founding Engineer

(Featured) - The founding engineer explains how continuous user research, familiar interfaces and workflow tracing help turn AI experiments into dependable business tools.

Edited by Jenn Moon, Editor at Featured News · Nov 13, 2026

As a full-stack software engineer and founding engineer, how would you describe the problems you focus on solving and the perspective you bring to building AI-enabled products?

Our customers were generating decks for their deals mostly by hand. They had tons of spreadsheets and other documents in different formats, with no clear source of truth. They dealt with that chaos every time they needed a deck and spent a lot of time going back and forth over the information.

They also hired whole teams of analysts and designers to sort through those documents and produce the decks. I built an AI-powered deck generation engine that structured the data and then used it to generate high-quality decks. I also worked on map generation and the front end around it.

That’s the kind of problem I focus on: a task that takes people a lot of time because the information is messy and the process is manual. My full-stack background lets me work on both the generation engine and the interface customers use. In this project, I needed to address the data behind the deck as well as the deck itself.

What path led you to become a founding engineer working at the intersection of full-stack development, automation and AI-powered business workflows?

My journey started as a front-end engineer at PVD-Technology, my first position in the information technology (IT) field. Frameworks like React weren’t really popular back then, so my main stack was Hypertext Markup Language (HTML), Cascading Style Sheets (CSS), and vanilla JavaScript. Having that narrow focus allowed me to build solid front end expertise, which served as a foundation for my future career.

In my next position, as a full-stack developer at DecisionMapper, I started extending my skill set with backend technologies. I was fascinated by how the whole thing worked end to end. It helped me deepen my expertise and connect the dots in my understanding of building complex applications with heavy data processing and traffic.

My journey in the startup world began at Zocdoc. During that period, I started paying more attention to the business impact of the software I was building, as well as the importance of domain expertise and working with stakeholders. Later, I joined the founding team at Henry AI, where we built AI-driven solutions for commercial real estate. My work included a deck generation engine that helped customers structure their data and automate work they had previously done manually.

When you move from an AI demo to a real product feature, what is the first real user workflow you examine to determine whether AI can genuinely improve it?

Usually, I start with manual tasks where users spend most of their time and cognitive effort tediously going through documents, entering information, comparing it and reconciling differences. These are tasks where I see AI agents helping more and more.

What interests me is that models can do more than parse documents in various formats and consolidate the information. They can also reason about that information, detect possible inconsistencies and anomalies, suggest changes and, with the right permissions, make changes themselves.

Another task that takes a lot of effort for knowledge workers is research. It might mean searching through documents, doing online research or combining both. With Model Context Protocol (MCP) connectors and plugins, models can retrieve more context from third-party services as they work. Instead of following the whole thread of a complex, multi-hour search, the user can review the end result, its citations, and the sources and actions behind it.

From your experience building workflow-level AI systems, how do you narrow a promising but broad AI idea into a first use case with clear business value?

The first and most important thing is to make sure there is a shared understanding of user needs. That involves user research and a lot of communication with the end user. When I talk to customers, I do not try to sell them my product immediately or ask for their opinion about my idea. Usually, people in that situation compliment your intentions to make you feel good. Instead, I ask questions that help me explore their past experiences with their workflows and the main pain points they had to deal with.

It is also critical to verify whether the problem they are talking about is really a problem. Sometimes people get emotional about things that irritate them once in a while, but it turns out they already have a low-effort workaround that works well enough. Such problems may become worth solving at some point, but at the moment, they are unlikely to bring much business value.

When a real customer problem with the highest potential for business value is identified, it is time to scope the first release. It should cover the core user needs in a specific area, but stay narrow enough to be delivered quickly and improved through further iterations.

What design choices have helped you make AI outputs dependable enough for professionals to use in their day-to-day work rather than treat as a novelty?

Product design is crucial to turning an idea into a useful product, alongside the technical implementation. A lot depends on user research: understanding people’s current workflows, needs and habits. Beyond following interface design best practices, we need to know how technical our users are, how comfortable they are with new technology and how willing they are to change established ways of working.

For example, in the deck editor I built at Henry AI, tables were among the most-used components. They could contain financial data, comparison analyses or breakdowns of metrics. We initially introduced a new way to edit them, but our users found it difficult. So we developed an interface that followed familiar Excel mechanics. Our users already spent a lot of time in Excel, and that familiarity made them much more comfortable making the final corrections to presentations generated by the agentic workflow.

Another part of making AI dependable is showing users where the output came from. In my experience, citations make a big difference. Users can check where a particular number came from or where an image was sourced. If the model takes a number from a profit-and-loss (P&L) statement for the wrong period, that is still a mistake, but the source makes it easier to understand and correct. Without it, the user has no straightforward way to tell whether the number came from the wrong document or was invented by the model.

How do you decide where an AI system can act autonomously and where a human checkpoint is essential in a business workflow?

There are two sides to the decision. On one side, it is the user’s call whether they are comfortable with AI developing, improving and executing a decision for them. Usually, if the cost of a mistake is not very high or critical to the business, and users trust the system, they are more willing to delegate the whole process. Otherwise, they are more cautious and want to keep the final approval.

On the other side, the product itself has to be reliable and traceable. What I find critical in AI-driven systems is evals. These are basically tests for an AI system: We provide use cases and check whether the output meets predefined quality criteria. The results help us understand which tasks the system can handle reliably and where human review is still needed.

This testing system is not static. As the product evolves and we discover new use cases or mistakes, we include them in the tests. That way, our understanding of the system’s reliability develops along with the product, and we have a basis for deciding whether to give it more autonomy.

As a full-stack engineer, how have you built reliability, observability and fallback behavior into AI-powered workflows when model outputs or connected systems fail?

Generally, the techniques developed over decades of classic software engineering still apply to AI-powered systems. We need systems that can scale up and down with the load and have enough redundancy to keep operating if an underlying service fails. For observability, I want to trace a user action from input to output or failure. In an AI workflow, that includes the model’s inputs and outputs, calls to connected services, intermediate artifacts, errors, token usage and the cost of a particular run. These records help us understand where something went wrong, even when the workflow involves several systems.

Another useful technique is deploying agents that observe the system while it runs in production. For example, an agent can monitor an internal assistant’s conversations with users and flag possible anomalies or signs of frustration. It can alert a human while the conversation is still happening, giving them a chance to step in. Those signals can also feed into further improvements, including automatically creating evaluation cases from problematic conversations. I would test any resulting changes before applying them to the live system.

For founding teams looking to turn AI experimentation into a dependable product, what is one practical habit you would establish early to keep development grounded in customer value?

If I had to highlight the most valuable practical habit for keeping development grounded in customer value, it would be continuous user research. As I mentioned earlier, it is critical to understand the user’s problem, their past experiences, their pain points and how they try to resolve those pain points. With that understanding, we can dig into the real user problem and then come up with a solution using our ideas and our product.

However, initial research is not enough. It should be a continuous habit. We keep talking to our customers as we solve certain problems for them, but other problems emerge as their workflows evolve with our product and the other tools they use daily. It is still important to avoid simply asking for feedback on our product or idea and instead focus on their past experiences.

Another point is the importance of building trust. It has always been critical to build trust with your customers, but in the AI era, it is becoming increasingly important and difficult as more and more products are driven by large language models (LLMs) with unpredictable outputs that are hard to control. In that situation, we have to be transparent about the capabilities and limitations of the system we build for our users so they do not encounter unexpected results or face risks they are unaware of.

About the expert

Roman Martynenko

Roman MartynenkoFullstack Software Engineer, Founding Engineer

Disclosure: I previously worked at Henry AI as a Founding Engineer, where I helped develop the AI-powered tools discussed in this interview

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

Small offices on Macs taught me to stop treating them as exceptions

Small offices on Macs taught me to stop treating them as exceptions

(Jens Hagel) - Several IT providers had turned down one of our clients because its staff work on Macs, and looking after offices like that changed how I think about managing devices.

Edited by Jenn Moon, Editor at Featured News · Nov 12, 2026

A while back, a 20-person company got in touch with us. All Mac users. They had been to a few other IT companies before coming to us. One even told them straight up, they didn't want to work with Macs.

I get it, but I don't think it was a good decision. I have been working in the IT sector for small to medium-sized businesses in northern Germany since 2004. I have learned more about device management from Mac offices than from most of our Windows clients.

Why they say no

I think most IT companies use software, scripts, etc., that are tailored to Windows. We also did. We only have a limited amount of hardware/software that we use because that is how you give good answers when you have a small team.

Managing a Mac for a business can require additional setup. There are some things you have to change and different ways to update, etc. Someone has to know where these things are. I think for an IT company that only has a few Macs in their customer base, it is easier to say no.

These companies are most likely not even making an IT decision. These are probably agencies or companies that enjoy using Macs. You wouldn't want them to change their work style just because your IT company doesn't want to learn.

What an office full of unmanaged Macs looks like

We had an advertising agency that came to us. 14 people all using Macs. They stored their files on a Linux server, which was in a small server room that in the summer was 38 degrees Celsius. Around 100 degrees Fahrenheit. They used mobile air conditioners to keep it cool. Only one man knew how to manage the server, and he was 76 and almost retired. He stored backups on Blu-rays that he believed would last 100 years. The managing director said he just wanted to be able to go on vacation and not worry about the server breaking.

The Macs were not a big issue. They were all set up manually and treated as personal computers. Nobody knew what was managed on what machine or what would happen if a Mac was stolen.

We took their files and put them on Microsoft 365/SharePoint. Managed their Macs through Apple's business enrollment platform and Microsoft Intune. That server room is now a storage room.

Managing a Mac

Much like Windows, you have your company device that is stored in a central database. Then you have a program that can manage that device.

Apple Business holds the enrollment records for company Macs. Microsoft Intune can then manage those devices. If you purchase your Mac through Apple or an authorized reseller and have it assigned to your company in Apple Business, it can automatically enroll in your mobile device management (MDM) and receive its settings when first turned on. Microsoft has a great article on how to set up your Mac with Intune. Intune can also be configured to require FileVault encryption during setup on supported Macs.

This means your user can unbox a laptop. Sign into their work account and have email and programs ready to go. Your company can manage that device if it gets lost.

When we created kjello, we wanted to support Macs as a normal device. Just like a user can choose a Windows laptop, they can choose a Mac. Both will set up when first turned on.

When I will say no

There are some situations where I will tell my customers that Macs are not the best choice for them.

There is some software that only works on Windows. Accounting programs and some machine controls come to mind. If you need to run these programs on a Mac, you are going to have to find a workaround. Every workaround you add is another thing that can fail. I would not switch a company over to Macs just for IT.

You will need someone on your team that knows both systems, which for an IT company means more knowledge. I think it's worth it, but I can see why some may not want to take on that extra knowledge. Just don't turn away your customers that use Macs.

The last thing I will say is when purchasing your Macs, if you buy a Mac from a store one Saturday, it may not be automatically added to Apple Business. You can add it later, but it's more work. I have found that if you don't do it when you buy the laptop, it may never get added. Always tell your customers to purchase through a vendor that will add the device to your company.

About the author

Jens Hagel

Jens Hagel is the founder and owner of hagel IT-Services in Hamburg, Germany, which he started alone from home in 2004. He has spent more than 20 years running IT for small and midsize companies in northern Germany and now builds kjello, a platform for managed workplace IT.

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

What a Hotel Doorman Taught Me About Building Software

What a Hotel Doorman Taught Me About Building Software

(Daniel Haiem) - By Daniel Haiem

Edited by Jenn Moon, Editor at Featured News · Nov 10, 2026

I recently came across a concept called the doorman fallacy. Someone posted about it somewhere (I can't remember where exactly), but it was so interesting that I made time to look it up. So, advertising executive Rory Sutherland coined the term "doorman fallacy" in his 2019 book Alchemy, and he illustrates it with a hypothetical hotel.

Picture a hotel manager reviewing costs and realizing that the doorman draws a full salary for just opening a door. He figures that if he gets an automatic door, it will save him money. Obviously, the math looks reasonable. The thing is, the doorman was never just opening a door. He greeted customers, hailed taxis for those who needed one, recognized regulars, and kept an eye on who walked in. Yes, the hotel might have been able to save on payroll, but it could risk losing some regulars along the way because they no longer feel welcome, or because nobody is hailing taxis anymore or watching who walks in.

That's the concept of doorman fallacy. You define a role by its most visible task, then you automate it and eventually lose everything else the role was quietly doing. What's funny is that I've seen companies make the same mistake with software for years — even myself — and I didn't know there was a term for it. Now, with AI booming, that mistake is becoming even easier to make.

Ever since automation became a thing, I've heard countless versions of, "Can we just automate this?" Don't get me wrong, some things genuinely need to be automated to make our lives easier. But there are also times when the workflow they want to replace is rarely as simple as the process document says. There are small steps between the obvious ones that most of us overlook, and the worst part is we won't notice them until later, after the gap has already left dents in the whole process. Just like what happened at Klarna. In early 2024, the company said its AI assistant was doing the equivalent work of 700 full-time customer service agents, but by 2025, its CEO acknowledged that the cost-focused approach had hurt service quality, and Klarna began expanding human support again. They didn't abandon AI completely, though. They basically kept the assistant and added people back where quality had slipped. Simply put, AI could handle the visible task, but the quality of the service was the doorman.

I have been trying to avoid the same mistake for the past few years. Whenever automation is in question — because we love to automate — my team and I do a simple exercise where we list every outcome the role touches, not every task it performs.

The questions include:Who does this person talk to?What do they notice that nobody asked them to notice?Which complaints never reach leadership because this person handles them first?

If that list runs longer than the job description, you've found your doorman. And if you do, for the love of God, do not replace them.

Now when it comes to setting an automation for our clients or if the project involves automation, the first thing I do is evaluate it by pricing everything the role or step touches, not just what it costs overall. This leads me to questions like, "What happens to retention, error rates, escalations, and customer trust?" If nobody on the team can answer that, the savings on the spreadsheet aren't real yet, they are merely an attempt to guess.

Now, let me be clear, none of this is an argument against automation. Most of the software we build exists to remove manual work, and much of it should. I am simply reminding founders, developers, and whoever this may concern about the sequence. You have to understand the full job first, then decide which parts the machine should take.

About the author

Daniel Haiem

Daniel Haiem is the CEO at AppMakers USA, a Los Angeles-based mobile app development firm that has helped startups and enterprise brands design, build, and scale mobile and web products since 2014. Dan writes about product strategy, AI in apps, and what it takes to ship software that lasts.

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

Institutional knowledge is the data skill AI cannot replace

Institutional knowledge is the data skill AI cannot replace

(Vinit Rajguru) - The real risk isn’t automation. It’s assuming AI understands your data when nobody in your company does.

Edited by Jenn Moon, Editor at Featured News · Nov 10, 2026

One weekend I was sitting at my apartment playing with my kid, and I saw a video on TV saying artificial intelligence (AI) will replace everything and people are losing their jobs. The rumination started in my mind that I might also lose my job because of AI, and honestly I was scared and stressed at that time, even though I was in a good position at my current role. As always, the mind tries to protect you. Then I started remembering my early days, how I started my career and how I am using AI in my current role. When I got my hands dirty with it, I realized it will not take my job the way I was thinking.

I have been with the same client for almost five years. When I started, I was shadowing someone and they told me to eyeball the data. Then after one month, that person left the company, and no one was there to teach me. I thought I couldn’t handle it because I didn’t know anything about it. I was hired as an ops person to help the organization with day-to-day operations. I honestly wanted to quit. No one was there to help me, and I was getting pressure from different teams including campaign ops, marketing ops, products and engineering.

This incident happened when the team wanted to launch a campaign, and the goal was to send communications to users who don’t have a business account, but we ended up sending it to users who already had a business account. The reason was the type of identifiers (IDs) we were using to identify those accounts — no one knew what those IDs meant.

So I started collecting evidence. Searching Slack to see if anyone knew about certain tables or columns. Digging through Confluence pages. Trying my luck everywhere. The customer data platform (CDP) also had its own reliability issues when data volumes were too large. The data pipeline was set up to run every two hours, falling back to the next cycle whenever a failure occurred. But there was a hard catch: a strict limit of 200 million records. Exceeding that volume broke the pipeline entirely, leaving us with only two paths forward: Either cut down the data volume or contact the CDP support team in advance to give them a heads-up that a specific audience would breach the threshold.

Usually, whenever we received a request from marketing ops or campaign ops or the product team, we had to follow a few steps: Submit a ticket in our project management tool, wait for it to be assigned and then take it from there. Based on these requests, the average end-to-end turnaround time was approximately five days, which became a pain point for the marketing and campaign ops teams. While this timeline was generally acceptable for regular campaigns, ad hoc campaigns often required a faster turnaround.

To optimize this workflow, and after many trials, I engineered an AI-assisted tool (Field Identifier) that lets campaign ops and marketing ops identify which fields they can use for their upcoming campaigns without reaching out to the data team every time.

Did the changes actually make a difference? We let the data speak for itself. In the second half of 2025, our team of three completed 170 requests with an average five-day turnaround. After building and testing a prototype in January 2026, I launched it in February. Over the next six months, despite taking on a higher workload of 185 requests, we cut our turnaround time down to just two days.Field Identifier did a lot of the heavy work, but the speed boost came from a few moving parts working together — mainly automation, better Structured Query Language (SQL) queries and the team tweaking the workflow as we grew. This was our "aha" moment because the dashboard gave leadership visibility into the measurable improvements we had achieved.

Currently, I am architecting an AI agent to fully automate query execution and segment building, so marketers don’t even have to interact with the tool. They simply make a request, and the work gets done. Their only responsibility is to review the result and make sure everything looks correct.

The way I use AI is more like a conversation. I bring my problem, AI asks me clarifying questions, I answer and together we figure things out. But here is the thing: If I don’t know my own data, I can’t answer those questions. AI is only as good as what I bring to the conversation. Without years of embedded contextual knowledge, it is just guessing.

Early in my career, when I told my colleague how lost I felt and overwhelmed I was, she gave it to me straight: I had been thrown into the deep ocean and had to sink or swim. Initially I thought, how can someone be so straight forward? Later I realized that most data environments are exactly like that. Nobody hands you a life jacket. The individuals who survive are those who master the waters and figure out how to swim — and then build the life jacket so the next person doesn’t have to.

Through discussion with colleagues and friends who are interested in data analytics and data engineering, I have observed that some people are reluctant to enter into the domain thinking AI might limit the opportunities.

My advice is to target overlooked areas. AI can definitely handle tasks, but you need to train it with the right knowledge first. That knowledge and those insights need to come from someone who is actively engaged in the domain, has hand-on experience with the day-to-day work and understands the why behind it. Be that practitioner.

About the author

Vinit Rajguru

Vinit Rajguru is a Senior Solution Consultant specializing in marketing data infrastructure and audience strategy at a global technology firm. He helps organizations build data pipelines, implement customer data platforms, and use artificial intelligence driven targeting to reach the right customers. He writes about the intersection of AI, data engineering, and modern marketing technology.

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

Interview with Udaya Bhaskar Vemuri, Senior Application Security Analyst

Interview with Udaya Bhaskar Vemuri, Senior Application Security Analyst

(Featured) - Vemuri discusses approved AI models, evidence-based findings, architecture context and collaboration with application teams.

Edited by Haley Spracale, Editor at Featured News · Nov 3, 2026

As a Senior Application Security Analyst, how would you describe your work in application security and the perspective you bring to secure design reviews?

As an application security engineer/architect, my goal is always to build secure application designs and work toward a shift-left design model. Secure design reviews (SDR) help in designing secure architectures and identifying flaws early in the software development life cycle (SDLC) process. Designs should also be scalable to support future-oriented AI designs and other capabilities.

What path led you to become a Senior Application Security Analyst, and how did secure design reviews become an important part of your work?

I started my career in security, and I came to love the work I did. Secure design reviews became especially interesting to me because they are the foundation of any application design. If an application's architecture or design is secure, the application tends to withstand attacks better than it would by relying only on front-end protections such as network firewalls and web application firewalls.

When you first began using an LLM-based assistant in design-review preparation, what practical problem were you trying to solve?

It’s been more than a year now. I was trying to solve the problem of keeping application security in step with the speed at which applications need to grow and scale their features to meet customer needs, so that security would never become a bottleneck for development teams. This is where AI-based reviews with human intervention help speed up reviews, prevent security from becoming a bottleneck, and ultimately deliver a secure application.

How do you decide which application materials and review tasks are appropriate to share with an AI assistant, particularly when handling sensitive technical information?

The AI model should have gone through its due diligence with the technology risk assessment teams and been approved by the compliance and security teams before you start sharing details with it. Nowadays, many major vendors offer on-premises options on an isolated network for your company, where they don't use your data to train the AI model. Once you have an approved AI model, you can share architecture diagrams and other details to conduct a secure design review.

What structured context do you provide an AI assistant to get useful early insights into an application’s architecture, data flows, and likely threat areas?

So, once the application teams share the architecture diagrams, data flow diagrams, and C1, C2, and C3 diagrams, we can run them through the AI and ask it to provide evidence-based results and explain, for each vulnerability, why it considers the application vulnerable, what part of the design triggered the vulnerability, and why it assigned that severity. This way, you will get an evidence-based result from the AI, and then you can manually question each finding.

Can you describe how you validate AI-generated review questions or observations before they influence a security finding or risk decision?

The configuration I set, or the prompt I give the AI, is to do SDR reviews using industry standards for findings and risk determination. Then, I send a list of questions to my internal app team to find out what kind of data the app handles and whether it is internally or externally facing. These are some generic questions that help the AI determine the risk. At the end of this process, there will be human validation.

Where have you found AI assistance most valuable in helping uncover gaps around authentication, authorization, integrations, or data handling during a design review?

I have found AI most valuable when reviewing complex designs and comparing them with previous design releases, as it can relate the overall architecture, data flow, and API or other integrations. It helps uncover issues related to authorization, authentication, integrations, and data handling if you provide the right prompt, ask the right questions, and give the AI the exact details it needs to evaluate the design.

How do you keep AI-assisted reviews collaborative with engineering teams rather than allowing the tool’s output to become a substitute for technical discussion and evidence?

I treat AI as a supporting tool for speeding up reviews rather than as a final decision-maker. A human always makes the final decisions. We go back and forth with questions to application teams to address assumptions AI makes when highlighting a finding. In this way, neither AI nor I make assumptions; instead, we deliver findings based on evidence. This also helps us understand the actual risk.

For application security teams starting with AI-assisted secure design reviews, what is the first human-led process or guardrail you would put in place, and why?

As I’ve said throughout., evidence-based output and human validation are the first and last guardrails I would put in place. AI might not understand the complete architecture. For example, it might flag a GDPR issue, but the compliance team may have already addressed it. Rather than sending the finding, first check with the app teams to see whether they have that data or those rules in place. This helps ensure a better secure design review using AI.

Thanks for sharing your knowledge and expertise. Is there anything else you'd like to add?

Yes, AI is developing at a fast pace, and application development is too. We can integrate with TFS or Git, read the full code, and ask the AI itself to give you the C1, C2, and C3 diagrams. Then you can send the diagrams to the app team to validate. In this way, the app team's burden is reduced, and they will be happy to collaborate. Happy exploring AI.

About the expert

Udaya Bhaskar Vemuri

Udaya Bhaskar VemuriSenior Application Security Analyst

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

Subscribe to Technology