Category: Uncategorized

By Tatev Blbulyan, Co-founder & CEO of Meettal

A candidate applies for a role and receives a rejection before they have even closed the job post. An employee’s performance review is prepared from meeting transcripts, productivity data, written feedback and goals recorded throughout the year. Someone experiencing a workplace conflict explains the situation to an HR chatbot because it is available immediately, while the HR team is busy or difficult to reach.

A few years ago, these situations would have sounded like predictions about the future of work. By 2026, they are becoming part of everyday HR operations.

AI can already write job descriptions, search for candidates, summarise interviews, analyse engagement surveys, draft policies, prepare performance reviews and answer employees’ routine questions. As the tools improve, companies will naturally use them in more parts of the employee journey because they save time, organise information quickly and reduce the amount of repetitive administrative work HR teams have been carrying for years.

The more difficult conversation begins when AI moves beyond supporting the process and starts shaping decisions that directly affect people’s careers, income, reputation and sense of security at work. This is exactly the conversation we run at our Beyond AI strategy lab on August 7 in Yerevan.

Judgment: the human contribution HR cannot outsource

HR works with information, processes and data, although much of its responsibility appears in situations where the available information is incomplete. A recruiter may need to understand why someone’s career took an unusual direction. A manager may need to recognise that an employee’s weak results were caused by constantly changing priorities. An HR professional may need to notice that a seemingly simple conflict is connected to a deeper issue of power, exclusion or fear.

These situations require context, judgment and the willingness to take responsibility for the final decision. They also require someone who understands that two people can experience the same workplace very differently, and that the information recorded in a system may represent only a small part of what actually happened.

For many years, HR professionals have said they want to spend less time on administration and become more strategic. AI may finally remove a meaningful part of that administrative burden. The opportunity is significant, although it also creates a new expectation: once the repetitive work is reduced, HR will need to show what its human contribution actually is.

That contribution begins with judgment.

AI in recruitment: where human oversight actually matters

In recruitment, AI can help clarify an unclear hiring request, organise applications, summarise CVs, identify relevant experience and prepare interview questions. It can also help recruiters manage high volumes of applications without spending hours manually opening every document.

The risk appears when the system’s first assessment becomes the only assessment that matters.

Careers are rarely as structured as job descriptions expect them to be. A strong candidate may have an unfamiliar title, experience from a company the system does not recognise, a career gap, an international background or skills described with different terminology. Some candidates are excellent at presenting their experience on paper, while others have much stronger abilities than their CV suggests.

An experienced recruiter asks questions when something does not immediately fit. They may recognise potential in a person whose experience looks unconventional, understand why a career decision made sense at a particular moment or notice that someone’s abilities are stronger than the keywords used in their application.

AI works with the information it has been given and the patterns it has learned to recognise. When recruiters rely too heavily on those patterns, companies may repeatedly select people whose careers and communication styles resemble the data the system already understands.

Meaningful human oversight requires more than approving an automatically generated shortlist. Recruiters need to understand how the shortlist was created, which criteria were prioritised and which types of candidates may have disappeared before a human ever saw them. A recruiter who only confirms the recommendation has very little influence over the decision, even if the company can technically say that a person remained involved.

AI in performance management: data without context is dangerous

The same challenge appears in performance management.

AI can gather goals, summarise feedback, review written communication and identify patterns across several months. Used responsibly, this can help managers remember achievements that happened earlier in the year and reduce the tendency to judge employees mainly by their most recent work.

At the same time, performance is influenced by far more than completed tasks and measurable outputs. An employee may miss a goal because priorities changed several times, because another team created delays or because their manager failed to provide clear direction. Someone may appear less productive because they spend a large amount of time supporting colleagues, preventing problems and carrying work that never becomes visible in a dashboard.

Data may show that a result was missed, while the reason behind that result may require a much deeper understanding of the team, the manager and the working environment.

This is where managers need to remain fully involved. Performance reviews should reflect what a manager has observed, discussed and understood throughout the year. AI can help organise that information and prepare a first draft, although the manager must still form an independent opinion, explain the reasoning behind it and take responsibility for the assessment.

When managers allow AI to complete most of the thinking, employees may receive polished feedback that sounds professional but feels disconnected from their actual experience. The review becomes another document generated by the organisation rather than a meaningful conversation with the person who is supposed to understand their work.

Employees also need a real opportunity to respond, provide missing context and challenge conclusions that feel inaccurate. Human involvement only has value when the person involved has enough authority, knowledge and time to reconsider the recommendation.

Employee relations: where automation must stop

Employee relations requires even greater care.

AI can answer questions about annual leave, benefits, internal processes, documentation and company policies. Employees often appreciate receiving immediate answers to straightforward questions, especially when HR teams are working across countries and time zones.

Serious workplace concerns belong in a different category.

When someone reports harassment, bullying, discrimination, retaliation or a conflict involving a manager, they are usually assessing whether the organisation is safe enough to trust. The first response can influence whether they continue speaking, withdraw the complaint or begin looking for another job.

An automated system may help collect initial information or explain the next steps, although a person should become involved quickly. The employee needs to feel that someone has understood the seriousness of the situation, is able to ask thoughtful questions and will remain accountable for what happens next.

The same applies to moments when a company communicates difficult decisions. A person who has completed five interview stages deserves more than a generic rejection generated within seconds. An employee returning from parental leave may need a conversation about workload, flexibility and confidence rather than a link to a policy. Someone being made redundant needs a person who can explain the decision, answer questions and stay present even when the conversation becomes emotional or uncomfortable.

AI makes it easier to create communication that is grammatically correct, consistent and legally cautious. Those qualities matter, although they do not automatically make the communication thoughtful. People usually remember whether they felt respected, whether someone listened and whether the person delivering the message seemed willing to take ownership of it.

Regulation is a floor, not a ceiling

The European Union’s AI framework already reflects many of these concerns by treating employment-related AI systems, including certain recruitment and worker-management tools, as high-risk. It also places strict limitations on the use of emotion-recognition technology in workplaces.

Regulation will continue to influence how companies introduce AI into HR, although legal compliance alone will not create a trustworthy employee experience. An organisation may disclose that it uses AI and still provide candidates with no realistic way to appeal a decision. It may keep a human formally involved while giving that person no authority to challenge the recommendation. It may publish a responsible AI policy that looks impressive and remains almost impossible for employees to understand.

Responsible use depends on the choices a company makes about power and accountability.

HR leaders need to know which data is being used, who selected it, what assumptions are built into the system and whether certain groups are consistently receiving worse outcomes. Employees and candidates should know when AI plays a meaningful role in a decision, and there should be a clear route to human review when the outcome has serious consequences.

Someone inside the organisation must also be willing to explain the final decision without hiding behind the technology.

How AI could weaken the next generation of HR

As AI takes over more repetitive work, HR teams will need stronger skills in critical thinking, process design, evidence evaluation, communication and ethical decision-making. Understanding how to use AI will become part of the job, although the ability to recognise weak reasoning, missing context and unfair outcomes will determine whether the technology improves HR or simply makes existing problems faster.

There is another challenge that companies should consider early: many HR professionals and managers develop judgment through experience, repetition and mistakes.

Junior recruiters learn by reading hundreds of CVs, comparing candidates, conducting interviews and receiving feedback when their assumptions were wrong. New managers learn through difficult conversations, unclear situations and decisions that force them to consider several perspectives. HR professionals become trusted advisers by seeing how policies work in real situations and how differently employees respond to the same decision.

When AI completes all the early analysis, younger professionals may lose important opportunities to build those skills. They may become very efficient at reviewing recommendations without developing the experience required to question them.

Companies need to design AI into learning rather than allowing it to replace learning. Junior employees can use AI while still being expected to form their own opinion, explain their reasoning and identify where the system may be wrong. Managers can use AI to organise feedback while remaining responsible for observing employees and holding regular conversations throughout the year. (This is one of the frameworks we work through inside the Meettal Academy HR Deep Dive Program.)

What the strongest HR teams will do in 2026

The strongest HR teams in 2026 will probably use AI extensively, although their advantage will come from understanding where its involvement should end.

Scheduling, reporting, document preparation, policy searches and repetitive coordination can be automated without removing anything meaningful from the employee experience. Decisions involving personal dignity, conflicting accounts, career consequences, trust and accountability need much stronger human involvement.

AI will continue to improve at processing information, identifying patterns and completing structured tasks. HR’s value will increasingly come from understanding the person behind the information, recognising what the system cannot see and making decisions that can be explained with honesty and defended with responsibility.

The future of HR will depend on how carefully organisations protect those moments. When someone’s career, confidence or livelihood is affected, they should know that a real person considered their situation, understood the context and was prepared to stand behind the decision.

That is the part of HR that should remain human.

Frequently asked questions: HR and AI in 2026

By Tatev Blbulyan, Co-founder and CEO of Meettal

Armenia is increasingly known for software engineering, QA, DevOps and AI talent. But international companies looking for production support engineers in Armenia often overlook one of the market’s strongest and least visible technical talent pools.

The talent exists. The issue is that companies search for it too literally.

They open LinkedIn, enter “Production Support Engineer,” add Armenia as the location and assume that the results represent the entire market. When only a limited number of profiles appear, they conclude that Armenia does not have enough relevant professionals.

In reality, many of the strongest candidates are already doing production support work. They simply have a different title.

The market is much bigger than the job title

Production support responsibilities in Armenia are spread across several professional communities.

The relevant person may currently be called an Application Support Engineer, Technical Support Engineer, Systems Engineer, Site Reliability Engineer, DevOps Engineer, Database Engineer, Network Operations Engineer, Middleware Engineer or Technical Support Specialist.

Some candidates work under software engineering titles while spending a significant part of their time investigating live-system problems, analysing logs, checking databases, reproducing customer issues and coordinating urgent fixes.

Others sit inside technical support teams but work daily with APIs, Linux, SQL, cloud infrastructure, networking and monitoring systems.

Current and recent Armenian job listings reflect exactly this fragmentation. Roles have appeared under titles including Production Support Engineer, Application Support Engineer, Technical Support Specialist and Site Reliability Engineer, often placed in completely different job categories.

Search only by title, and Armenia appears to have a small production support market.

Search by responsibilities, technologies and environment, and the pool becomes much larger.

What does a production support engineer actually do?

Production support engineers investigate technical problems affecting software that is already being used by customers or internal teams.

Depending on the product, they may:

• Analyse application and infrastructure logs
• Investigate failed transactions
• Run SQL queries
• Troubleshoot APIs and integrations
• Work with Linux systems
• Review monitoring alerts
• Identify networking or infrastructure problems
• Reproduce production incidents
• Coordinate escalations with engineering teams
• Communicate incident status to customers and stakeholders
• Document the issue and help prevent it from happening again

This is not ordinary customer service with a more technical title.

A strong production support engineer understands how a system behaves when something goes wrong. They can investigate beyond a standard troubleshooting guide, identify where the problem may be coming from and decide what should happen next.

At the same time, they must communicate clearly while the answer is still unknown.

That combination is difficult to find.

A technically brilliant engineer may dislike customer communication or struggle to explain an incident clearly. A strong communicator may not have enough technical depth to investigate the problem independently.

Production support sits between engineering, operations and communication. The best candidates are comfortable in all three.

Why Armenia is a strong market for production support

Armenia already has the kind of technology environment that develops these skills.

The local ecosystem includes startups, established product companies, financial platforms, enterprise software businesses, telecommunications companies and international engineering centres. Enterprise Armenia describes the country’s high-tech sector as a combination of startups and mature IT firms building software, AI and electronics.

These companies do not only build software. They operate it, maintain it and support the people who use it.

Current Yerevan-based vacancies also show international product companies hiring for technical and application support functions. Staff.am has recently listed a Technical Support Specialist role at ServiceTitan and an Application Support position at Siemens Industry Software.

This matters because production support maturity is usually developed inside real operating environments.

You learn it when a customer cannot complete a transaction.

When an integration fails after a release.

When the system is technically running, but something is clearly wrong.

When several teams are involved and nobody immediately knows who owns the problem.

When a stakeholder needs an update before engineering has found the root cause.

Professionals who have worked in these environments learn that solving the technical issue is only one part of the job. Someone must also understand the business impact, coordinate the response, keep people informed and make sure the incident does not disappear without proper documentation.

That experience already exists in Armenia.

It is simply distributed across different companies and job titles. Our Hire Software Engineers in Armenia page has more on the broader Armenian tech market.

Where the strongest candidates may be hiding

The right candidate pool depends on the product being supported.

For a data-heavy platform, relevant profiles may come from application support, database administration, backend engineering, data operations or financial technology.

For infrastructure or networking software, the best candidates may currently work in DevOps, systems engineering, network operations, cloud infrastructure or SRE.

For a customer-facing SaaS product, technical support engineers with strong experience in SQL, APIs, integrations and troubleshooting may be more relevant than someone with the exact production support title.

For enterprise software, implementation engineers and integration specialists may already be doing much of the required work.

QA engineers can also be relevant, especially when they have strong debugging skills and experience reproducing complex production issues.

Even developers may be open to the right production support opportunity if the role includes serious technical investigation, system ownership and a clear path towards reliability engineering, platform engineering or technical leadership.

This is why I would never start the search with the title.

I would start with the environment.

What kind of system is being supported? What normally goes wrong? How deeply should the person investigate before escalating? Do they need Linux, SQL, networking, cloud infrastructure, Kubernetes or scripting? Will they communicate directly with customers? Is this primarily incident response, application support, infrastructure reliability or a combination of all three?

Once those questions are answered, the Armenian market can be mapped properly.

The biggest hiring mistake: treating all support as the same

One reason companies struggle to hire production support engineers is that the role itself is often poorly defined.

“Support” can mean almost anything.

In one company, it means answering standard customer questions and forwarding technical tickets.

In another, it means investigating a distributed platform at two in the morning while enterprise customers are waiting for an update.

Those are not the same job. They should not have the same job description, interview process or salary expectations.

Before entering the Armenian market, companies need to define the technical depth of the role.

Will the person only collect information, or will they investigate independently? Will they have access to production systems? Can they run queries and review logs? Are they expected to understand the architecture? Will they speak directly with customers? Do they participate in an on-call rotation? Can they fix issues themselves, or only escalate them? What happens after a serious incident?

Candidates will evaluate these details carefully.

Technically strong professionals often hear the word “support” and assume the role will mean repetitive tickets, limited ownership and no engineering career path. Sometimes that assumption is correct.

Calling a position “Production Support Engineer” while expecting one person to operate as an SRE, database engineer, incident manager and customer success specialist will not make the role attractive.

Neither will advertising an engineering position and revealing during the interview that most of the work is basic ticket processing.

The strongest candidates want to understand what they will own, what they will learn and where the role can take them.

Armenia is cost-effective, but strong candidates are not cheap

Another common mistake is entering Armenia with unrealistic compensation expectations. I wrote about this at length in Armenia Is Cost-Effective. It Is Not Cheap.

Armenia can still be a cost-effective location compared with many US and Western European markets. But experienced engineers are connected to international companies, remote opportunities and global salary information.

The strongest production support candidates are rarely unemployed and waiting for any offer.

They may already be working in DevOps, SRE, infrastructure, application support, technical support or software engineering. Moving into a new role needs to make sense financially and professionally.

Compensation should also reflect the real requirements.

A daytime application support role with a clear escalation process is one thing.

A customer-facing production role involving night shifts, on-call responsibility, high-severity incidents and deep infrastructure knowledge is another.

Companies that price both roles the same will either struggle to hire or attract candidates who cannot perform at the required level.

The role may still cost less than hiring an equivalent professional in San Francisco, London or Amsterdam.

That does not make the person cheap.

How I would build a production support team in Armenia

I would normally begin with one strong senior hire or team lead.

That person should understand the product, define the incident process and establish what good production support looks like inside the company.

They should help answer questions such as:

• What should the team resolve independently?
• When should an issue move to engineering?
• What information must be collected before escalation?
• How should incidents be prioritised?
• Who communicates with customers?
• How are root causes documented?
• How will repeated incidents be tracked?

Hiring several junior support specialists before answering these questions often creates a larger team without creating a stronger function.

The team receives incidents, but it cannot resolve complex problems. The first senior hire sets the standard.

After that, the team can be built from a combination of experienced production support professionals and candidates from adjacent backgrounds.

Some may come from application support. Some may come from systems or networking. Some may come from QA, DevOps or backend engineering. Some may have strong customer-facing experience and need additional technical training.

Others may be technically excellent but need help developing incident communication skills. Product-specific knowledge can be taught.

Troubleshooting discipline, technical curiosity, ownership and calm communication are much harder to build quickly.

What should companies assess during interviews?

A production support interview should not rely only on theoretical technical questions. Candidates should be given a realistic incident. For example:

A customer reports that an important transaction has failed. There is no obvious application error, and the engineering team is not immediately available.

• What would the candidate check first?
• What questions would they ask?
• Which logs or systems would they review?
• How would they narrow down the problem?
• At what point would they escalate?
• What information would they send to engineering?
• What would they tell the customer while the investigation continues?

The goal is to understand how they think when the answer is unclear.

Strong production support engineers create structure inside uncertainty. They know how to investigate without jumping randomly between possibilities. They communicate what is known, what is not yet known and what will happen next.

That is the real skill companies need to assess.

The talent pool is there. Companies need to search differently.

Armenia does not have an enormous pool of people carrying one perfectly matching title.

It has something more useful: a technically mature market full of professionals who already understand different parts of production support.

Frequently asked questions about hiring production support engineers in Armenia